AI-Assisted Recognition of Misdiagnosed Paroxysmal Nocturnal Hemoglobinuria

AI-Assisted Recognition of Misdiagnosed Paroxysmal Nocturnal Hemoglobinuria | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article AI-Assisted Recognition of Misdiagnosed Paroxysmal Nocturnal Hemoglobinuria Adam Bowen, Shannon Pierce, Claire Bischel, Maha Bayya, Alfarooq Alshaikhli, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9383995/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 7 You are reading this latest preprint version Abstract Background Paroxysmal nocturnal hemoglobinuria (PNH) is often misdiagnosed, delaying effective therapy. We systematically reviewed delayed or misdiagnosed PNH cases and evaluated whether large language models could recognize PNH within the differential diagnosis using standardized clinical vignettes. Results A systematic review identified 63 publications describing 68 patients with confirmed PNH and diagnostic delay. Each case was converted into a standardized vignette and tested across 3 large language models under complete and feature-removal conditions, with diagnostic rank as the primary outcome. One model outperformed the others, identifying PNH in 89.7% of vignettes and ranking it first in 70.6%, exceeding the approximately 81% top-five rate of the other models. Rank performance remained stable without thrombosis or marrow-failure features but declined when both clinical hemoglobinuria-hematuria and laboratory hemolysis cues were absent, establishing hemolysis as the minimal diagnostic signal. Testing was typically triggered by unexplained hemolysis, 45.6%, or thrombosis, 36.8%, while frequent mislabels included iron-deficiency anemia, aplastic anemia, myelodysplastic syndromes, and urinary tract disease. Patients were younger than registry cohorts, median 30 versus 46 years, and had prolonged delays, median 24 months, IQR 12–60. Race and ethnicity were reported in 23.5% of cases. Limitations include reliance on case literature and vignette testing. Conclusions Large language models reliably recognized PNH when hemolysis features were present and failed when absent, mirroring clinical pitfalls. Earlier hemolysis-directed evaluation and future integration of large language model-assisted screening or electronic medical record evaluation may help shorten diagnostic timelines and expedite hematology referral. paroxysmal nocturnal hemoglobinuria diagnostic delay misdiagnosis rare disease large language models hemolysis Figures Figure 1 Figure 2 Figure 3 Background Paroxysmal nocturnal hemoglobinuria (PNH) is a rare, acquired clonal hematopoietic stem cell disorder caused by PIGA mutations that impair complement regulation and make erythrocytes susceptible to intravascular hemolysis, the hallmark of the disease. 1 , 2 The global incidence is estimated at 1–1.8 cases per million, highlighting the diagnostic challenge posed by its rarity. 3 Thrombosis develops in up to 40% of patients and remains the leading cause of morbidity and mortality, frequently at hepatic, portal, or cerebral sites. 1 Diagnostic delays contribute to irreversible organ injury and diminished treatment response. 4 By contrast, early treatment with complement inhibitors such as eculizumab or ravulizumab can reduce hemolysis, transfusion needs, and thrombotic risk, improving survival and quality of life. 5 Although high-sensitivity flow cytometry enables definitive diagnosis, delays remain common. Clinical features, including anemia, cytopenias, abdominal pain, renal dysfunction, and hemoglobinuria, often mimic more prevalent conditions such as aplastic anemia, myelodysplastic syndromes, autoimmune hemolytic anemia, or urologic disease. 6 While individual reports and registry data have noted these diagnostic pitfalls, they have not been systematically compiled. Recent advances in artificial intelligence (AI) raise the possibility of improving recognition of rare disorders. Machine-learning models can integrate diverse clinical features and may identify conditions that clinicians overlook. 7 Whether LLMs can aid in the earlier recognition of PNH has not been evaluated. We conducted a systematic review of published case reports and series describing delayed or misdiagnosed PNH. Presenting features, initial diagnostic labels, and triggers for confirmatory testing were extracted and used to generate standardized vignettes. These vignettes were tested with contemporary LLMs to determine whether PNH was included among the leading differential diagnoses. By combining clinical synthesis with AI evaluation, our objective was to highlight recurrent diagnostic pitfalls, assess the ability of LLMs to recognize rare disease, and generate insights to reduce diagnostic delay in PNH. Clinically, such tools could expedite the workup by prompting earlier hematology referral, drawing attention to diagnostic gaps or anchoring errors, and triggering timely laboratory evaluation to mitigate cognitive biases. Methods Study registration and design This systematic review was conducted in accordance with PRISMA 2020 guidelines and Joanna Briggs Institute (JBI) methodology for case reports and case series. The protocol was prospectively registered with PROSPERO (CRD420251104533). 8 No amendments were made after registration. Information sources and search strategy PubMed, Embase, and Google Scholar were searched from inception to July, 31, 2025 (Fig. 1 ). The search combined controlled vocabulary and keywords for paroxysmal nocturnal hemoglobinuria, case reports, case series, diagnostic delay, and misdiagnosis (Appendix 1). Eligibility criteria We included case reports and case series describing patients with confirmed PNH, diagnosed by high-sensitivity flow cytometry (glycosylphosphatidylinositol-deficient clones) or, in older studies, by the Ham or sucrose lysis test. Eligible cases required documentation of diagnostic delay, defined as ≥ 12 months between symptom onset and confirmed diagnosis or an explicit author statement of delay/misdiagnosis. Reports initially attributed to other conditions (eg, aplastic anemia, autoimmune hemolytic anemia, myelodysplastic syndromes, urologic disorders) but later revised to PNH were eligible. Exclusions were non-human or laboratory studies, reviews, editorials, clinical trials without individual case detail, and reports lacking extractable timelines. For duplicate publications, the most complete or recent version was retained. Study selection All references were imported into Rayyan (Qatar Computing Research Institute) for de-duplication and blinded screening. Two reviewers independently assessed titles/abstracts and full texts. Disagreements were resolved by consensus or third-party adjudication. Study flow is summarized in the PRISMA 2020 diagram (Fig. 1 ). Data extraction and management Data were extracted in duplicate using a piloted Excel form. Variables included publication details, demographics (age, sex, race/ethnicity, comorbidities), presentation (symptoms, laboratory and imaging findings), diagnostic course (initial misdiagnosis, treatments before diagnosis, delay duration, thrombosis, clinical trigger, confirmatory test). Evidence of marrow failure was also recorded. Discrepancies were resolved by consensus. Quality assessment Methodological quality was appraised with JBI Critical Appraisal Checklists for case reports and case series. Two reviewers rated each study as high, moderate, or low quality; disagreements were resolved through discussion. Full item-level ratings are provided in Appendix 2. 9–72 AI-based diagnostic evaluation Standardized clinical vignettes were generated from extracted case data and tested across GPT-4.5, Claude-Sonnet 4, and Gemini-2.5 (versions current May–July 2025). Each vignette included demographics, presenting features, laboratory and imaging findings, prior treatments, initial misdiagnoses, comorbidities, and duration of diagnostic delay (Appendix 3). Models were prompted with a standardized query requesting the five most likely diagnoses, and the rank position of PNH (1–6 scale) was recorded as the primary outcome. All model queries were performed as single-pass, deterministic evaluations using default settings without temperature adjustment or repeated sampling, and each vignette was structured using a predefined template with consistent variable ordering (Appendix 3 and 4). To assess diagnostic robustness, vignettes were re-run after systematic removal of predefined diagnostic domains: (A) hemoglobinuria/hematuria, (B) thrombosis, (C) laboratory hemolysis, and (D) bone-marrow failure. Figure 2 provides a visual summary of the systematic diagnostic domain ablation workflow used to test model robustness across all 16 vignette states. Each omission scenario was executed independently to prevent model memory carry-over. All models were evaluated in default configuration (no fine-tuning or post-processing). Full prompt text is provided in Appendix 4, and the full information-reduction protocol is detailed in Appendix 5. Ethics statement This study used only published, de-identified data extracted from case reports and case series. No human participants were directly involved, no identifiable private information was collected, and institutional review board approval was not required. Statistical analysis The primary outcome was the diagnostic rank assigned to PNH by each model on a 1–6 ordinal scale (1 = top diagnosis; 2–5 = within top five; 6 = not listed). Because each vignette was evaluated by all three models, paired, within-case comparisons of ranks were conducted using the Wilcoxon signed-rank test, with effect size r reported. Secondary outcomes included median diagnostic rank, performance under feature-removal scenarios, and inter-model comparisons. Continuous variables were summarized as medians with interquartile ranges (IQRs), categorical variables as counts and percentages. Feature-removal scenarios were unpaired and assessed with Mann–Whitney U tests (effect size Cliff’s delta, δ). Bonferroni corrections were applied for multiple testing. Hierarchical clustering was used to visualize performance across omission scenarios. Analyses were conducted in Python 3.13.5; descriptive tables were generated in Excel with the Real Statistics add-on. A two-sided adjusted p < 0.05 was considered statistically significant. Results Patients in the misdiagnosed cohort were younger than those in the International PNH Registry (median age 30 vs 46 years) 73 , while sex distribution was comparable (Table 1 ). Reports spanned all WHO regions, though representation from low- and middle-income countries was sparse. Race or ethnicity was documented in only 23.5% of cases, limiting assessment of disparities. In contrast, registry data report that 78% of patients are Caucasian. 73 Limited demographic reporting in case literature constrains evaluation of disparities in diagnostic delay. Table 1 Clinical comparison of the misdiagnosed PNH cohort and the International PNH Registry Key demographic and diagnostic-delay characteristics of the misdiagnosed PNH cohort (N = 68) compared with the International PNH Registry (Schrezenmeier et al., 2020); values shown as medians (IQR) or percentages. Variable Cohort (N = 68) International PNH Registry (Schrezenmeier et al., 2020) Demographic characteristics Median age, years (IQR) 30 (24–49) 46 (27–61) Sex, % female 45.6 46–55 Geographic distribution, % † EUR 34; AMR 29; WPR 13; SEAR 13; EMR 7; AFR 3 82.3 from EUR and AMR Race/Ethnicity reported, % 23.5 recorded vs 76.5 missing 78 Caucasian Diagnostic delay 5 years 15 (22.1) 24 Median delay (IQR) 24 months (12–60) Range 0.5–288 months 22.5 months (mean 1.88 year) Footnotes : † WHO regions: EUR = Europe, AMR = Americas, WPR = Western Pacific, SEAR = South-East Asia, EMR = Eastern Mediterranean, AFR = Africa. Diagnostic-delay data were available for 65 of 68 cases; three reports described an initial misdiagnosis but did not specify the elapsed time to confirmed diagnosis. Percentages are calculated using the full cohort (N = 68). Unexplained hemolysis (45.6%) and thrombosis (36.8%) were the leading triggers for PNH testing, together accounting for more than three-quarters of cases (Supplementary Table S1 ). Other triggers included incidental laboratory or marrow abnormalities, hemoglobinuria, and cytopenias. Initial misdiagnoses most often involved iron-deficiency anemia, marrow failure syndromes, or autoimmune hemolysis. Hemoglobinuria was frequently attributed to urinary tract pathology rather than recognized as evidence of hemolysis. Detailed lists of less common misdiagnoses are provided in the footnotes. Median time to diagnosis was 24 months (IQR 12–60), closely approximating registry data (median 22.5 months). 73 However, a larger share of patients in this cohort were diagnosed between one and five years after symptom onset (Table 1 ). While central tendencies were comparable, the distribution in our cohort reflected a greater proportion of diagnoses in the one-to-five-year interval. The most prolonged delays occurred when patients were initially labeled with autoimmune/systemic inflammatory disease (median 132 months) or marrow failure syndromes such as aplastic anemia (median 41.5 months) (Supplementary Table S2). In contrast, patients presenting with atypical thrombosis were diagnosed earliest (median 3 months). Intermediate delays were observed with more common mislabels such as iron deficiency anemia and renal/urologic disorders, reflecting how prevalent diagnoses can obscure recognition of rare disease. These findings highlight that misattribution to chronic or systemic conditions prolongs diagnosis, whereas atypical thrombotic presentations prompt earlier testing. On the 1–6 rank scale, GPT-4.5 achieved superior performance on complete vignettes. In paired, within-case comparisons of rank, GPT-4.5 outperformed Gemini-2.5 (Wilcoxon signed-rank, p < 0.0001, r = 0.30) and Claude-Sonnet 4 (p = 0.0068, r = 0.18). As secondary summaries, GPT-4.5 included PNH within the top five in 89.7% of cases and ranked it first in 70.6%, exceeding Claude-Sonnet 4 and Gemini-2.5 (both ~ 81% top-five) (Table 2 ). Rank stability was greater with GPT-4.5 and Gemini-2.5 than with Claude-Sonnet 4. Pairwise testing confirmed this hierarchy: GPT-4.5 outperformed Gemini-2.5 with a moderate effect size and Claude-Sonnet 4 with a smaller effect, while Claude-Sonnet 4 had only a modest edge over Gemini-2.5 (Table 2 ). Pairwise Wilcoxon signed-rank tests were applied because each vignette was evaluated by all three models, allowing direct within-case comparison of performance on the full dataset, with full vignettes being evaluated, similar to a clinical setting with full information. These results confirm that the performance hierarchy observed in descriptive analyses remained statistically robust when tested across paired vignettes. Table 2 Model Performance and Pairwise Comparisons of AI Diagnostic Accuracy in Misdiagnosed PNH Vignettes Primary outcome is per-case diagnostic rank of PNH (1 = top diagnosis; 2–5 = within top five; 6 = not listed in top five). Values shown are median [IQR] rank for each model on the same 68 vignettes. Paired Wilcoxon signed-rank tests compare GPT-4.5 vs Claude-Sonnet 4, GPT-4.5 vs Gemini-2.5, and Claude-Sonnet 4 vs Gemini-2.5. Model / Comparison % Top-5 Rank % Top-1 Rank Median Rank IQR p-value Effect Size (r) GPT-4.5 89.7 70.6 1 2 NA NA Claude 80.9 63.2 1 3 NA NA Gemini 80.9 55.9 1 2 NA NA GPT-4.5 vs Gemini NA NA NA NA < 0.0001 0.30 GPT-4.5 vs Claude NA NA NA NA 0.0068 0.18 Claude vs Gemini NA NA NA NA 0.0218 0.16 Footnotes : NA indicates not applicable. Rank coding: 1 = top diagnosis; 2–5 = within top five; 6 = not in top five (poorest rank). Although each model evaluated the same vignettes under complete information, information-reduction removal created non-identical vignette subsets, necessitating unpaired Mann–Whitney testing; under these conditions, GPT-4.5 retained a moderate advantage over Gemini-2.5, whereas differences versus Claude-Sonnet 4 were less stable (Table 3 ). These findings confirm that the performance hierarchy observed in Table 2 persisted when models were evaluated on vignette subsets with systematically reduced diagnostic information. Table 3 Feature-removal experiments on the rank outcome (1–6) across predefined omission domains. Between-model unpaired comparisons of diagnostic performance across systematically reduced vignette conditions, in which predefined diagnostic omission domains: A = hemoglobinuria/hematuria, B = thrombosis, C = laboratory hemolysis, D = bone-marrow failure were omitted to simulate incomplete clinical information. Two-sided p-values and effect sizes (Cliff's delta, δ) are shown, representing the magnitude of performance differences under information-reduction scenarios. Comparison p-value Cliff’s Delta (δ) GPT-4.5 vs Gemini < 0.0001 0.36 GPT-4.5 vs Claude 0.076 0.38 Claude vs Gemini 0.029 0.15 Footnotes : Rank coding: 1 = top diagnosis; 2–5 = within top five; 6 = not in top five. Across pairwise Bonferroni-adjusted comparisons of diagnostic rank under systematic feature removal, most scenarios yielded p ≈ 1.0, indicating stable performance when a single domain was omitted or when thrombosis (B) and marrow-failure (D) features were removed in combination. Significant differences emerged only when both hemolysis cues were absent. Compared with the complete-information baseline, rank performance worsened (higher rank values) in all scenarios omitting hemoglobinuria (A) and laboratory hemolysis (C) together (p values 0.001–0.012). The maximal omission scenario (No A + B + C + D) differed significantly from every comparator retaining either A or C. Within A-absent conditions, removal of C consistently produced significance, whereas removal of B or D alone did not. These results demonstrate that model accuracy was robust across incomplete inputs unless both clinical and laboratory hemolysis features were removed, which defined the threshold for performance loss. Figure 3 presents a hierarchical clustering heatmap of Bonferroni-adjusted Mann–Whitney U test results across feature-removal scenarios. Hierarchical clustering of pairwise comparisons reinforced these findings, grouping all scenarios that retained at least one hemolysis feature into a single branch and isolating those lacking both into a distinct cluster. The clustering dendrogram confirmed these patterns by grouping all scenarios that retained at least one hemolysis feature together and isolating those lacking both A and C into a distinct branch, reinforcing hemolysis as the critical diagnostic signal. Discussion This systematic review and AI-based evaluation provide new insights into why PNH is frequently diagnosed late and whether LLMs can support earlier recognition. Across 68 misdiagnosed cases, both clinicians and LLMs depended heavily on evidence of hemolysis. When either hemoglobinuria or laboratory markers were absent, rank performance fell sharply. Even minimal signals such as dark urine, elevated LDH, or reduced haptoglobin proved to be the most actionable features for timely recognition. 4 , 74 , 75 GPT-4.5 identified PNH in nearly 90% of vignettes and ranked it first in 70%, outperforming Claude-Sonnet 4 and Gemini-2.5. Accuracy remained stable when thrombosis or marrow-failure features were removed but deteriorated in the absence of hemolysis cues. This pattern parallels clinical practice, where autoimmune or inflammatory mislabels often delay recognition, while hepatic, portal, or cerebral thrombosis reliably prompts testing. 4 , 74 – 76 Nearly one in five patients were initially labeled with iron-deficiency anemia despite later confirmation of hemolysis, illustrating how common comorbidities can obscure recognition. 77 Using both Wilcoxon signed-rank tests on paired vignettes and Mann–Whitney U with Cliff’s delta on reduced, unpaired scenarios strengthened confidence in these findings by demonstrating robustness under complete and incomplete information. 78 Patterns of misdiagnosis in the literature reflected common clinical anchors. Iron-deficiency anemia frequently obscured recognition, consistent with reports that true iron deficiency is common in PNH. 77 Hemoglobinuria was often misattributed to urinary tract disease rather than hemolysis, a well-described diagnostic pitfall. 65 Autoimmune hemolytic anemia, particularly during pregnancy or in association with thrombosis, led to misleading positive DAT results and partial steroid responses. Dysplastic marrow features and cytopenias frequently resulted in attribution to myelodysplastic or myelofibrosis-related anemia, while incidental marrow abnormalities were often labeled as failure syndromes without concurrent hemolysis testing. 6 , 74 , 75 Renal-predominant presentations, though rarely reported, further complicated attribution of symptoms. 13 The longest diagnostic delays occurred in patients initially labeled with autoimmune or marrow-failure disorders, underscoring how these categories suppress hemolysis assessment. In contrast, thrombotic presentations were diagnosed earliest, reflecting the high specificity of hepatic, splanchnic, or cerebral venous thrombosis for PNH. 4 , 75 , 76 Compared with international registry data, diagnostic delays in this cohort (median 24 months) were similar overall but skewed toward the one-to-five-year range. 73 Only one in five patients were diagnosed within the first year, compared with 40% in the registry, likely reflecting publication bias toward prolonged or atypical cases. 73 , 79 Outliers with extreme delays of up to 288 months highlight the heterogeneity of diagnostic pathways and suggest that systematic reflex testing strategies may be needed to capture these rare but high-impact cases. 6 , 74 Patients in this series were also younger than those in registry cohorts, consistent with prior reports that PNH clones may remain clinically silent for years, especially in younger adults where cytopenias are easily attributed to alternative etiologies. 2 Sex distribution was comparable to registry data, but race and ethnicity were missing in more than three-quarters of reports. This absence not only prevents meaningful equity analyses but also raises concern that diagnostic delays may be underestimated in underrepresented populations. 80 The geographic skew toward Europe and the Americas further suggests under-ascertainment in regions with limited awareness and diagnostic access. These findings reinforce two priorities for practice. First, clinicians should maintain a low threshold to test for hemolysis in patients with unexplained cytopenias, dark urine, or atypical thrombosis, even when initial impressions favor more common conditions such as iron deficiency or autoimmune hemolysis. 74 , 75 A reflex panel including LDH, haptoglobin, DAT, and high-sensitivity PNH flow cytometry at the time of marrow-failure labeling may represent the single highest-yield systems change. 3 , 6 , 74 Second, LLMs may have potential as adjunctive diagnostic tools. Their rank performance on the 1–6 scale across incomplete vignettes suggests feasibility within real-world electronic medical records (EMR), where missing information is common. 7 Early applications of machine learning to PNH detection in primary care datasets have demonstrated feasibility, 7 and embedding automated alerts that combine hemolysis markers with atypical thrombosis could trigger earlier suspicion. 7 , 74 , 76 Beyond system-level alerts, such tools could help reduce cognitive biases by countering anchoring on common diagnoses, drawing attention to atypical features, and encouraging reconsideration when clinical and laboratory findings are discordant. By highlighting overlooked signals, LLMs may prompt nonspecialists to initiate earlier referral to hematology/oncology and expedite definitive work-up, thereby shortening time to recognition. Future work should validate these findings prospectively within clinical workflows, such as hematology referral triggers, diagnostic decision-support tools, or prospective registry studies, alongside electronic health record datasets, while also exploring prompt standardization across LLM architectures and assessing model interpretability to identify which clinical features drive AI recognition. Several limitations merit consideration. Case-report literature is biased toward unusual or prolonged diagnostic journeys, inflating delay estimates and underrepresenting milder cases. 79 , 81 Publication bias and selective reporting further constrain generalizability. Demographic detail was sparse, limiting equity analyses. 80 AI performance was tested on standardized vignettes rather than raw clinical records, so real-world validation will require testing within electronic health records, or within further real-world studies using LLMs, where noise and contradictions are common. 7 Finally, the wide distribution of delays indicates that a small subset of patients experiences extreme diagnostic journeys; targeted interventions for this group may yield disproportionate benefit even if median times remain unchanged. The increasing integration of large language model–based tools into electronic medical record platforms makes this a particularly timely area for validation. As health systems adopt AI-assisted documentation, summarization, and clinical decision-support systems, embedding standardized, diagnostic prompts within these workflows offers a practical and scalable pathway to evaluate real-world impact on diagnostic delay. Conclusions In this systematic review of misdiagnosed PNH, both clinicians and large language models depended primarily on hemolysis cues for recognition. Model performance remained robust when thrombosis or marrow-failure features were absent but declined when both clinical hemoglobinuria-hematuria and laboratory hemolysis features were removed, reinforcing hemolysis as the critical diagnostic signal. These findings support earlier hemolysis-directed evaluation in patients with unexplained cytopenias, dark urine, or atypical thrombosis, and support further validation of AI-assisted diagnostic support within real-world clinical workflows to shorten time to PNH recognition. Declarations Ethics approval and consent to participate This study used only published, de-identified data extracted from case reports and case series. No human participants were directly involved, no identifiable private information was collected, and institutional review board approval was not required. Consent for publication Not applicable. Availability of data and materials All data generated or analysed during this study are included in this published article and its supplementary information files. The supplementary appendices provide the complete database search strategies, JBI critical appraisal tables, the standardized data extraction template, the full structured prompts used for AI evaluation, and the information-reduction protocol applied to test diagnostic robustness. Additional clarifications about the analytic workflow, statistical code, or prompt implementation are available from the corresponding author on reasonable request. Competing interests The authors declare that they have no competing interests. Funding No specific funding was received for this work. Authors' contributions AMB conceived and designed the study, collected and verified data, performed statistical analysis, prepared figures and tables, drafted the manuscript, coordinated revisions, and serves as guarantor of the work. MB and AA contributed to study design, quality review, and manuscript editing. ML contributed to manuscript editing. DB participated in article screening, contributed to case review, data extraction, and manuscript editing. SP, BA, and CB provided clinical input and critical review of the manuscript. KQ contributed to study conceptualization, data interpretation, and critical revision of the manuscript for important intellectual content. DI supervised the project, advised on study design and analysis, and provided critical review of the final manuscript. All authors read and approved the final manuscript. Acknowledgements The authors thank the Department of Hematology/Oncology at Karmanos Cancer Institute at McLaren Greater Lansing Hospital for academic support and guidance. Portions of this work were previously presented as an abstract at the 2025 American Society of Hematology Annual Meeting. Authors' information Not applicable. References Shi JJ, Ozcan YM, Santos CIA, Patel H, Shammo J, Bat T. Current landscape of paroxysmal nocturnal hemoglobinuria in the era of complement inhibitors and regulators. Ther Adv Hematol. 2024;15:20406207241307500. 10.1177/20406207241307500 . Colden MA, Kumar S, Munkhbileg B, Babushok DV. Insights into the emergence of paroxysmal nocturnal hemoglobinuria. Front Immunol. 2022;12:830172. 10.3389/fimmu.2021.830172 . Spychalska J, Duńska M, Myślińska A, Majewska-Wierzbicka M, Klimczak-Jajor E, Głodkowska-Mrówka E. Diagnostic landscape of first-time cytometric screening for paroxysmal nocturnal hemoglobinuria in Poland in 2013–2022. Orphanet J Rare Dis. 2024;19(1):271. 10.1186/s13023-024-03283-x . Devalet B, Mullier F, Chatelain B, Dogné J, Chatelain C. Pathophysiology, diagnosis, and treatment of paroxysmal nocturnal hemoglobinuria: a review. Eur J Haematol. 2015;95(3):190–8. 10.1111/ejh.12543 . Hillmen P, Hall C, Marsh JCW, et al. Effect of eculizumab on hemolysis and transfusion requirements in patients with paroxysmal nocturnal hemoglobinuria. N Engl J Med. 2004;350(6):552–9. 10.1056/NEJMoa031688 . Babushok DV. When does a PNH clone have clinical significance? Hematology. 2021;2021(1):143–52. 10.1182/hematology.2021000245 . Worker A, Mahon H, Sams J, et al. A machine learning algorithm for the detection of paroxysmal nocturnal haemoglobinuria (PNH) in UK primary care electronic health records. Orphanet J Rare Dis. 2024;19(1):378. 10.1186/s13023-024-03406-4 . Bowen A, Baraka D. AI-Assisted Recognition of Delayed or Misdiagnosed Paroxysmal Nocturnal Hemoglobinuria. PROSPERO International Prospective Register of Systematic Reviews; 2025. Accessed October 2, 2025. https://www.crd.york.ac.uk/PROSPERO/display_record.php?ID=CRD420251104533 Pokhrel B, Gautam S, Khanal S, Pokhrel NB, Shrestha A. A rare and misdiagnosed entity paroxysmal nocturnal hemoglobinuria: a case report. Cureus Published online May. 2021;8. 10.7759/cureus.14902 . Elias NS, Riedl J, Stouten K, et al. Abdominal pain in combination with an unexplained hemolytic anemia are crucial signs to test for paroxysmal nocturnal hemoglobinuria: A case report. Clin Case Rep. 2019;7(1):175–9. 10.1002/ccr3.1771 . Perdigones N, Morales M, Mason P, Bessler M. Case Report: Paroxysmal nocturnal hemoglobinuria in a woman heterozygous for G6PD A-. F1000Research. 2014;3:194. 10.12688/f1000research.4980.2 . Eisa MS, Mohamed SF, Ibrahim F, et al. Paroxysmal nocturnal hemoglobinuria with glucose-6-phosphate dehydrogenase deficiency: a case report and review of the literature. Case Rep Oncol. 2019;12(3):838–44. 10.1159/000503817 . Keating S, Machchhar R, Jain U, Naronowicz G, Lipschutz J. An interesting case of paroxysmal nocturnal hemoglobinuria with renal involvement. Cureus Published online July. 2024;5. 10.7759/cureus.63917 . Farooq Q, Saleem MW, Khan ZU, Hadi N. Paroxysmal nocturnal hemoglobinuria: a diagnostic zero-sum-game. Cureus. Published online Dec. 2020;7. 10.7759/cureus.11956 . Liang HY, Xie XD, Jing GX, Wang M, Yu Y, Cui JF. Posthepatectomy jaundice induced by paroxysmal nocturnal hemoglobinuria: A case report. World J Clin Cases. 2021;9(32):10046–51. 10.12998/wjcc.v9.i32.10046 . Xiao M, Di Y, An Y, Cui Y. Acute Necrotizing Pancreatitis Complicated with Paroxysmal Nocturnal Haemoglobinuria: A Case Report. Clin Exp Gastroenterol 2024;Volume 17:141–5. 10.2147/CEG.S449594 Athota S, Gopalan S, Arthur P, Jayaram A, Satish A. Deciphering paroxysmal nocturnal hemoglobinuria: an unusual paradigm of hemolytic Anemia. Cureus Published online August. 2024;19. 10.7759/cureus.67194 . Lumori BAE, Muyanja D. Diagnostic and treatment challenges of paroxysmal nocturnal hemoglobinuria in Uganda. Case Rep Hematol. 2019;2019:1–5. 10.1155/2019/7897509 . Quinquenel A, Maestraggi Q, Lecoq-Lafon C, Régis PDL, Delmer A, Servettaz A. Atypical presentation of paroxysmal nocturnal hemoglobinuria treated by eculizumab: A case report. Med (Baltim). 2017;96(12):e6403. 10.1097/MD.0000000000006403 . Fattizzo B, Pasquale R, Bellani V, Barcellini W, Kulasekararaj AG. Complement mediated hemolytic anemias in the COVID-19 era: case series and review of the literature. Front Immunol. 2021;12:791429. 10.3389/fimmu.2021.791429 . Korkama ES, Armstrong AE, Jarva H, Meri S. Spontaneous remission in paroxysmal nocturnal hemoglobinuria—return to health or transition into malignancy? Front Immunol. 2018;9:1749. 10.3389/fimmu.2018.01749 . Burns E, Anand K, Acosta G, et al. Autosplenectomy in a patient with paroxysmal nocturnal hemoglobinuria (PNH). Case Rep Hematol. 2019;2019:1–5. 10.1155/2019/3146965 . Cheekati M, Siu K, Ochs R. Management of paroxysmal nocturnal hemoglobinuria in CALR mutated post-essential thrombocythemia myelofibrosis: A case report. eJHaem. 2024;5(2):392–6. 10.1002/jha2.892 . Park S, So MK, Cho MS, Kim DY, Huh J. Coexistence of primary myelofibrosis and paroxysmal nocturnal hemoglobinuria clone with JAK2 V617F, U2AF1 and SETBP1 mutations: a case report and brief review of literature. Diagnostics. 2021;11(9):1644. 10.3390/diagnostics11091644 . Gediz F, Payzin B, Cakmak O, Uzum Y, Ernur D, Sahin F. Paroxysmal nocturnal hemoglobinuria in the differential diagnosis of thrombocytopenia. Hematol Rep. 2017;9(1):6862. 10.4081/hr.2017.6862 . Figueroa-Jiménez LA, González-Márquez AL, Alicea-Guevara R, et al. A Hispanic female patient with heartburn: A rare presentation of Paroxysmal Nocturnal Hemoglobinuria. Bol Asoc Med P R. 2015;107(3):9–12. Choi H, Yim H, Lee MJ. Hemosiderin induced acute kidney injury requiring hemodialysis in patients with paroxysmal nocturnal hemoglobinuria: A case report. Med (Baltim). 2023;102(40):e35412. 10.1097/MD.0000000000035412 . Balwani MR, Kute VB, Shah PR, Shah M, Varyani U, Trivedi HL. Manifestation of paroxysmal nocturnal hemoglobinuria as repeated acute kidney injury. J Nephropharmacology. 2016;5(2):116–8. Dambal A. Paroxysmal nocturnal haemoglobinuria masquerading as malaria: a case report. J Clin Diagn Res Published online. 2015. 10.7860/JCDR/2015/13465.6329 . Gervasi F, D’Amelio L, Trizzino A, et al. Cerebral stroke in a teenage girl with paroxysmal nocturnal hemoglobinuria. Hematol Rep. 2017;9(2):7012. 10.4081/hr.2017.7012 . Raghupathy R, Derman O. Response of paroxysmal nocturnal hemoglobinuria clone with aplastic anemia to rituximab. Case Rep Hematol. 2012;2012:1–5. 10.1155/2012/106182 . Rayas J, Hassan M, Hock RA, et al. Attack of the clones: a patient with untreated aplastic anemia presenting with classical paroxysmal nocturnal hemoglobinuria. Cureus Published online January. 2023;23. 10.7759/cureus.34093 . De Boysson H, Chèze S, Chapon F, Le Mauff B, Auzary C, Geffray L. Eosinophilic fasciitis with paroxysmal nocturnal hemoglobinuria. Joint Bone Spine. 2013;80(2):208–10. 10.1016/j.jbspin.2012.07.008 . Shakir A, Narula S, LaGrow A, Nusrat S. Two roads diverge: treatment choice in coexisting severe aplastic anemia and paroxysmal nocturnal hemoglobinuria. J Med Cases. 2020;11(6):182–4. 10.14740/jmc3478 . Memon A, Khan R, Rauf MU, Shafique K. Paroxysmal nocturnal hemoglobinuria presenting as cerebral venous sinus thrombosis: a case report. Int Arch Med. 2014;7(1):39. 10.1186/1755-7682-7-39 . Mandala E, Lafaras C, Goulis I, Tsioni K, Georgopoulou V, Ilonidis G. Treatment of a patient with classical paroxysmal nocturnal hemoglobinuria and Budd-Chiari syndrome, with complement inhibitor eculizumab: Case Report. Hippokratia. 2013;17(1):81–4. Brodsky A, Mazzocchi O, Sánchez F, Khursigara G, Malhotra S, Volpacchio M. Eculizumab in paroxysmal nocturnal hemoglobinuria with Budd-Chiari syndrome progressing despite anticoagulation. Exp Hematol Oncol. 2012;1(1):26. 10.1186/2162-3619-1-26 . Satish S, Rajesh R, Kurian G, Seethalekshmi N, Unni M, Unni V. Recurrent acute renal failure. Indian J Nephrol. 2010;20(4):217. 10.4103/0971-4065.73444 . Lauritsch-Hernandez LS, Kraehenmann F, Balabanov S, Kimmich N. Eculizumab application during pregnancy in a patient with paroxysmal nocturnal hemoglobinuria: A case report with review of the literature. Clin Case Rep. 2018;6(8):1582–7. 10.1002/ccr3.1634 . Patel PS, Singh PP, Archana A, Kumar O. A series of eight cases of pigment nephropathy: an obscured aspect of acute kidney injury. Cureus Published online July. 2024;10. 10.7759/cureus.64214 . Bais J, Pel M, Von Dem Borne A, Van Der Lelie H. Pregnancy and paroxysmal nocturnal hemoglobinuria. Eur J Obstet Gynecol Reprod Biol. 1994;53(3):211–4. 10.1016/0028-2243(94)90121-X . Santagostino A, Lombardi L, Dine G, Hirsch P, Misra SC. Paroxysmal nocturnal hemoglobinuria with a distinct molecular signature diagnosed ten years after allogenic bone marrow transplantation for acute myeloid leukemia. Case Rep Hematol. 2019;2019:1–3. 10.1155/2019/8928623 . Bhat LSR, Taduri G, Guditi S, Herur S, Bukka VC, Murugan P. An unusual case of relapsing and remitting acute kidney injury. Indian J Nephrol. 2024;34:665–6. 10.25259/ijn_437_23 . De-la-Iglesia S, Luzardo H, Lemes A, et al. Positive impact of eculizumab therapy on surgery for budd- chiari syndrome in a patient with paroxysmal nocturnal hemoglobinuria and a longterm history of thrombosis. Hematol Rep. 2016;8(3):34–6. 10.4081/hr.2016.6562 . Migdady Y, Olszewski AJ. Paroxysmal nocturnal hemoglobinuria presenting with a left intraventricular thrombus in a patient with prior thymoma and aplastic anemia. Clin Med Insights Oncol. 2011;5. 10.4137/CMO.S8442 . CMO.S8442. Vogel SJ, Reinhard EH. Paroxysmal nocturnal hemoglobinuria associated with infectious mononucleosis. Blood. 1979;54(2):351–3. Hegenbart U, Niederwieser D, Forman S, et al. Hematopoietic cell transplantation from related and unrelated donors after minimal conditioning as a curative treatment modality for severe paroxysmal nocturnal hemoglobinuria. Biol Blood Marrow Transpl. 2003;9(11):689–97. 10.1016/S1083-8791(03)00264-7 . Cheng KL, Brody J, Warshall CE, Sloand EM, Allen SL. Paroxysmal nocturnal hemoglobinuria following alemtuzumab immunosuppressive therapy for myelodysplastic syndrome and complicated by recurrent life-threatening thrombosis despite anticoagulation: Successful intervention with eculizumab and fondaparinux. Leuk Res. 2010;34(4):e85–7. 10.1016/j.leukres.2009.09.024 . Palmieri G, Selleri C, Montella L, et al. Thymoma followed by paroxysmal nocturnal hemoglobinuria: A unique clinical association in the context of multiorgan autoimmunity with a potential role for CD8 + T lymphocytes. Am J Hematol. 2006;81(10):774–8. 10.1002/ajh.20699 . Kawamoto M, Murakami Y, Kinoshita T, Kohara N. Recurrent aseptic meningitis with PIGT mutations: a novel pathogenesis of recurrent meningitis successfully treated by eculizumab. BMJ Case Rep. 2018;2018:bcr–2018. 10.1136/bcr-2018-225910 . Jackson GH, Noble RS, Maung ZT, Main J, Smith SR, Reid MM. Severe haemolysis and renal failure in a patient with paroxysmal nocturnal haemoglobinuria. J Clin Pathol. 1992;45(2):176–7. 10.1136/jcp.45.2.176 . Liu H, He C, Zhu H, et al. A paroxysmal nocturnal haemoglobinuria progress with waldenström macroglobulinemia along with t cell monoclonal expansion. Indian J Hematol Blood Transfus. 2014;30(S1):227–31. 10.1007/s12288-014-0337-9 . Alsalman Z, Alsalman M, Albesher M, et al. Primary myelofibrosis with concurrent paroxysmal nocturnal haemoglobinuria presenting with erectile dysfunction. Oxf Med Case Rep. 2022;2022(5):omac047. 10.1093/omcr/omac047 . de Ruiter ED, Baggen MGA, Middelkoop MP. A rare cause of haemolytic anaemia: paroxysmal nocturnal haemoglobinuria in an elderly patient. Neth J Med. 2003;61(5):174–6. Pande N, Bhat R, Singh BMK, Vivek G. Haemolytic anaemia—PNH type II cells presenting a diagnostic dilemma. BMJ Case Rep. 2014;2014:bcr2014204174. 10.1136/bcr-2014-204174 . Chandran AV, Nair R, Chandran PKG. The case | acute renal injury in a young burmese immigrant. Kidney Int. 2014;85(3):719–20. 10.1038/ki.2013.429 . Cheng ZJ, Shen YY, Warsame IM, Dai TM, Tu JL. Moyamoya syndrome caused by paroxysmal nocturnal hemoglobinuria. Chin Med J (Engl). 2018;131(23):2874–6. 10.4103/0366-6999.246065 . Mita M, Shichishima T, Noji H, Takahashi H, Nakamura K, Ikezoe T. Independent paroxysmal nocturnal hemoglobinuria and myelodysplastic syndrome clones in a patient with complete bone marrow failure. HemaSphere. 2018;2(5):e142. 10.1097/HS9.0000000000000142 . Townsley DM, Young NS. Blood consult: paroxysmal nocturnal hemoglobinuria and its complications. Blood. 2013;122(16):2795–8. 10.1182/blood-2013-07-360081 . Sica M, Pellecchia A, De Angioletti M, et al. Complement-mediated oxidative damage of red cells impairs response to eculizumab in a G6PD-deficient patient with PNH. Blood. 2020;136(26):3082–5. 10.1182/blood.2020007780 . Abideen ZU, Jafar MS, Hameed N, Malik A. Haemoglobinuria and portal venous thrombosis in a young male. J Ayub Med Coll Abbottabad JAMC. 2017;29(2):353–4. Arzour PH, Hadjami DD, Haddoum PF. Paroxysmal nocturnal hemoglobinuria (PNH): so much for a first diagnosis: a case report. East Afr Sch J Med Sci. 2024;7(09):386–8. 10.36349/easms.2024.v07i09.003 . Mancuso S, Sucato G, Carlisi M, et al. Paroxysmal nocturnal hemoglobinuria: When delay in diagnosis and long therapy occurs. Hematol Rep. 2018;10(1). 10.4081/hr.2018.7523 . Sandeep, Sharma P, Ahluwalia J, et al. Primary bone marrow T-cell/histiocyte-rich large B-cell lymphoma: a diagnostic challenge. Hematology. 2013;18(2):85–8. 10.1179/1607845412Y.0000000037 . Veerreddy P. Hemoglobinuria misidentified as hematuria: review of discolored urine and paroxysmal nocturnal hemoglobinuria. Clin Med Insights Blood Disord. 2013;6. 10.4137/CMBD.S11517 . CMBD.S11517. Parlier V, Tiainen M, Beris Ph, Miescher PA, Kimuutila S, Bellomo MJ. Trisomy 8 detection in granulomonocytic, erythrocytic and megakaryocytic lineages by chromosomal in situ suppression hybridization in a case of refractory anaemia with ringed sideroblasts complicating the course of paroxysmal nocturnal haemoglobinuria. Br J Haematol. 1992;81(2):296–304. 10.1111/j.1365-2141.1992.tb08223.x . Kirkizlar O, Kendir M, Karaali Z, et al. Acute renal failure in a patient with severe hemolysis. Int Urol Nephrol. 2007;39(2):651–4. 10.1007/s11255-006-9096-3 . Henderson C, Lo M, Massey G. Pediatric Paroxysmal Nocturnal Hemoglobinuria Presenting as Acute Kidney Injury. J Pediatr Hematol Oncol. 2021;43(4):e543–5. 10.1097/MPH.0000000000001847 . Misra UK, Kalita J, Bansal V, Nair PP. Paroxysmal nocturnal haemoglobinuria presenting as cerebral venous sinus thrombosis. Transfus Med. 2008;18(5):308–11. 10.1111/j.1365-3148.2008.00886.x . Bart JB. Paroxysmal nocturnal hemoglobinuria: report of a case with refractory megaloblastic-hypoplastic bone marrow, bizarre erythrocyte morphology, and a postive coombs’ test. Arch Intern Med. 1967;120(4):487. 10.1001/archinte.1967.04410010101015 . Yin XL, Zhou TH, Peng L, et al. A case report of concomitant paroxysmal nocturnal hemoglobinuria and heterozygous β-thalassemia. Ann Hematol. 2011;90(3):355–6. 10.1007/s00277-010-1011-4 . Ross JD, Rosenbaum E. Paroxysmal nocturnal hemoglobinuria presenting as aplastic anemia in a child. Am J Med. 1964;37(1):130–9. 10.1016/0002-9343(64)90217-7 . Schrezenmeier H, Röth A, Araten DJ, et al. Baseline clinical characteristics and disease burden in patients with paroxysmal nocturnal hemoglobinuria (PNH): updated analysis from the International PNH Registry. Ann Hematol. 2020;99(7):1505–14. 10.1007/s00277-020-04052-z . Parker CJ. Update on the diagnosis and management of paroxysmal nocturnal hemoglobinuria. Hematology. 2016;2016(1):208–16. 10.1182/asheducation-2016.1.208 . Brodsky RA. How I treat paroxysmal nocturnal hemoglobinuria. Blood. 2021;137(10):1304–9. 10.1182/blood.2019003812 . Kokoris S, Polyviou A, Evangelidis P, et al. Thrombosis in paroxysmal nocturnal hemoglobinuria (PNH): from pathogenesis to treatment. Int J Mol Sci. 2024;25(22):12104. 10.3390/ijms252212104 . Peng G, Yang W, Jing L, et al. Iron deficiency in patients with paroxysmal nocturnal hemoglobinuria: a cross-sectional survey from a single institution in China. Med Sci Monit. 2018;24:7256–63. 10.12659/MSM.910614 . Altman DG. Practical Statistics for Medical Research. 0 ed. Chapman and Hall/CRC; 1990. 10.1201/9780429258589 . Song F, Parekh S, Hooper L, et al. Dissemination and publication of research findings: an updated review of related biases. Health Technol Assess. 2010;14(8). 10.3310/hta14080 . Flanagin A, Frey T, Christiansen SL, AMA Manual of Style Committee. Updated guidance on the reporting of race and ethnicity in medical and science journals. JAMA. 2021;326(7):621. 10.1001/jama.2021.13304 . Gagnier JJ, Kienle G, Altman DG, Moher D, Sox H, Riley D. The CARE guidelines: consensus-based clinical case report guideline development. J Clin Epidemiol. 2014;67(1):46–51. 10.1016/j.jclinepi.2013.08.003 . Additional Declarations No competing interests reported. Supplementary Files PNHsupplement.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 15 May, 2026 Reviewers agreed at journal 15 Apr, 2026 Reviewers agreed at journal 14 Apr, 2026 Reviewers invited by journal 14 Apr, 2026 Editor assigned by journal 14 Apr, 2026 Submission checks completed at journal 14 Apr, 2026 First submitted to journal 10 Apr, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9383995","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":624551582,"identity":"8dcc35a1-4867-4f47-af95-e9eee60b6349","order_by":0,"name":"Adam Bowen","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA1klEQVRIiWNgGAWjYDACCQhZz8/A3MCQAOYkEKXFIkGygZE0LRUJBgeAWhiI0cI/u/mYxMcdEnnGxxtbNzzcY8fAz55jgN+SO8fSJGeekSg2O3Ow7UbCs2QGyZ43+LUw3Mgxu83bJsG47UYiUMsBZgaDGwRskQdp+QvUsnn+Q5CWegZ7QloMQFoY2yQSN0gwgrQcZjCQIKDF8M6x9J+9bRLGEmfADjvOI3HmWQFeLXK3mw8b/Gyrk+NvP3zs5o8D1UBG8ga8WjAAD2nKR8EoGAWjYBRgBQAXgE7cnxOsRgAAAABJRU5ErkJggg==","orcid":"","institution":"Mclaren Greater Lansing Hospital","correspondingAuthor":true,"prefix":"","firstName":"Adam","middleName":"","lastName":"Bowen","suffix":""},{"id":624551585,"identity":"75dd29a7-1f82-4b71-bd67-6105f98594ee","order_by":1,"name":"Shannon Pierce","email":"","orcid":"","institution":"Mclaren Greater Lansing Hospital","correspondingAuthor":false,"prefix":"","firstName":"Shannon","middleName":"","lastName":"Pierce","suffix":""},{"id":624551586,"identity":"691e71d7-6237-4793-885d-bab8062f99ee","order_by":2,"name":"Claire Bischel","email":"","orcid":"","institution":"Mclaren Greater Lansing Hospital","correspondingAuthor":false,"prefix":"","firstName":"Claire","middleName":"","lastName":"Bischel","suffix":""},{"id":624551587,"identity":"5249f59f-ad56-4227-808d-21e4a8b6a84a","order_by":3,"name":"Maha Bayya","email":"","orcid":"","institution":"Karmanos Cancer Institute","correspondingAuthor":false,"prefix":"","firstName":"Maha","middleName":"","lastName":"Bayya","suffix":""},{"id":624551588,"identity":"481fcf64-c42a-414c-b423-f5ea19e9d37b","order_by":4,"name":"Alfarooq Alshaikhli","email":"","orcid":"","institution":"Karmanos Cancer Institute","correspondingAuthor":false,"prefix":"","firstName":"Alfarooq","middleName":"","lastName":"Alshaikhli","suffix":""},{"id":624551589,"identity":"56177e2c-7ecc-487e-9957-5e59a873d5e9","order_by":5,"name":"Miller Lantis","email":"","orcid":"","institution":"Mclaren Greater Lansing Hospital","correspondingAuthor":false,"prefix":"","firstName":"Miller","middleName":"","lastName":"Lantis","suffix":""},{"id":624551590,"identity":"0c7d0e0a-90b6-4495-b06d-3ce809e3e692","order_by":6,"name":"Dania Baraka","email":"","orcid":"","institution":"Mclaren Greater Lansing Hospital","correspondingAuthor":false,"prefix":"","firstName":"Dania","middleName":"","lastName":"Baraka","suffix":""},{"id":624551591,"identity":"f92831a8-d270-4195-9492-6f029a89409e","order_by":7,"name":"Bilal Ali","email":"","orcid":"","institution":"Mclaren Greater Lansing Hospital","correspondingAuthor":false,"prefix":"","firstName":"Bilal","middleName":"","lastName":"Ali","suffix":""},{"id":624551593,"identity":"802d7b2a-7bdf-4570-9fbe-935c36e11a39","order_by":8,"name":"Khaleel Quaseem","email":"","orcid":"","institution":"Mclaren Greater Lansing Hospital","correspondingAuthor":false,"prefix":"","firstName":"Khaleel","middleName":"","lastName":"Quaseem","suffix":""},{"id":624551594,"identity":"6f39cafe-791a-4348-a6fe-306e0d109be8","order_by":9,"name":"Daniel Isaac","email":"","orcid":"","institution":"Karmanos Cancer Institute","correspondingAuthor":false,"prefix":"","firstName":"Daniel","middleName":"","lastName":"Isaac","suffix":""}],"badges":[],"createdAt":"2026-04-11 02:54:14","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9383995/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9383995/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107616894,"identity":"fac7c443-d150-4dbf-a22d-2a04cbfa7d03","added_by":"auto","created_at":"2026-04-23 09:16:31","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":67979,"visible":true,"origin":"","legend":"\u003cp\u003ePRISMA 2020 flow diagram summarizing study selection for systematic review of misdiagnosed and delayed paroxysmal nocturnal hemoglobinuria (PNH) cases.\u003c/p\u003e\n\u003cp\u003ePRISMA 2020 flow diagram summarizing inclusion of 68 misdiagnosed PNH cases from 63 publications.\u003c/p\u003e","description":"","filename":"Figure1PRISMA2020flowdiagram.png","url":"https://assets-eu.researchsquare.com/files/rs-9383995/v1/85e97f2724dea9aaa0b0e7b4.png"},{"id":107616896,"identity":"52ea27ef-6b3b-4d32-b963-06b25a4432d8","added_by":"auto","created_at":"2026-04-23 09:16:31","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":200602,"visible":true,"origin":"","legend":"\u003cp\u003eInformation Reduction Protocol for AI Models\u003c/p\u003e\n\u003cp\u003eVisual summary of the systematic diagnostic domain ablation workflow used to test model robustness across all 16 vignette states.\u003c/p\u003e","description":"","filename":"Figure2InformationReductionProtocolforAIModels.png","url":"https://assets-eu.researchsquare.com/files/rs-9383995/v1/9a864519a7eded6d4942098f.png"},{"id":107706204,"identity":"1880d7ac-147b-464b-8cc8-8bdea3cd9893","added_by":"auto","created_at":"2026-04-24 09:17:39","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":269621,"visible":true,"origin":"","legend":"\u003cp\u003eHierarchical clustering heatmap of pairwise Mann–Whitney U test p-values comparing AI diagnostic accuracy across feature-exclusion scenarios in paroxysmal nocturnal hemoglobinuria.\u003c/p\u003e\n\u003cp\u003eRows and columns are ordered by hierarchical clustering, and dendrograms depict similarity relationships among scenarios. Scenarios retaining at least one hemolysis feature (A or C) cluster together, while those lacking both A and C form a distinct branch with worse rank performance. Scenarios tested included omission of clinical hemoglobinuria (A), thrombosis (B), laboratory hemolysis (C), or bone marrow failure (D) and their combinations.\u003c/p\u003e","description":"","filename":"Figure3Hierarchicalclusteringheatmap.png","url":"https://assets-eu.researchsquare.com/files/rs-9383995/v1/369aa6e5f135ca2911fba85f.png"},{"id":107709162,"identity":"c5bf4e49-c450-49e5-b593-28ef3fe954f9","added_by":"auto","created_at":"2026-04-24 09:34:52","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":760376,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9383995/v1/4d2a3ce7-6c47-40f1-82e3-0447cd4fa3a5.pdf"},{"id":107616895,"identity":"5ee67fa9-1b70-4551-90be-30086b31cedb","added_by":"auto","created_at":"2026-04-23 09:16:31","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":33988,"visible":true,"origin":"","legend":"","description":"","filename":"PNHsupplement.docx","url":"https://assets-eu.researchsquare.com/files/rs-9383995/v1/9897c739cb159a669d91edb3.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"AI-Assisted Recognition of Misdiagnosed Paroxysmal Nocturnal Hemoglobinuria","fulltext":[{"header":"Background","content":"\u003cp\u003eParoxysmal nocturnal hemoglobinuria (PNH) is a rare, acquired clonal hematopoietic stem cell disorder caused by \u003cem\u003ePIGA\u003c/em\u003e mutations that impair complement regulation and make erythrocytes susceptible to intravascular hemolysis, the hallmark of the disease.\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e The global incidence is estimated at 1\u0026ndash;1.8 cases per million, highlighting the diagnostic challenge posed by its rarity.\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eThrombosis develops in up to 40% of patients and remains the leading cause of morbidity and mortality, frequently at hepatic, portal, or cerebral sites.\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e Diagnostic delays contribute to irreversible organ injury and diminished treatment response.\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e By contrast, early treatment with complement inhibitors such as eculizumab or ravulizumab can reduce hemolysis, transfusion needs, and thrombotic risk, improving survival and quality of life.\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eAlthough high-sensitivity flow cytometry enables definitive diagnosis, delays remain common. Clinical features, including anemia, cytopenias, abdominal pain, renal dysfunction, and hemoglobinuria, often mimic more prevalent conditions such as aplastic anemia, myelodysplastic syndromes, autoimmune hemolytic anemia, or urologic disease.\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e While individual reports and registry data have noted these diagnostic pitfalls, they have not been systematically compiled.\u003c/p\u003e \u003cp\u003eRecent advances in artificial intelligence (AI) raise the possibility of improving recognition of rare disorders. Machine-learning models can integrate diverse clinical features and may identify conditions that clinicians overlook.\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e Whether LLMs can aid in the earlier recognition of PNH has not been evaluated.\u003c/p\u003e \u003cp\u003eWe conducted a systematic review of published case reports and series describing delayed or misdiagnosed PNH. Presenting features, initial diagnostic labels, and triggers for confirmatory testing were extracted and used to generate standardized vignettes. These vignettes were tested with contemporary LLMs to determine whether PNH was included among the leading differential diagnoses. By combining clinical synthesis with AI evaluation, our objective was to highlight recurrent diagnostic pitfalls, assess the ability of LLMs to recognize rare disease, and generate insights to reduce diagnostic delay in PNH. Clinically, such tools could expedite the workup by prompting earlier hematology referral, drawing attention to diagnostic gaps or anchoring errors, and triggering timely laboratory evaluation to mitigate cognitive biases.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy registration and design\u003c/h2\u003e \u003cp\u003e This systematic review was conducted in accordance with PRISMA 2020 guidelines and Joanna Briggs Institute (JBI) methodology for case reports and case series. The protocol was prospectively registered with PROSPERO (CRD420251104533).\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e No amendments were made after registration.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eInformation sources and search strategy\u003c/h3\u003e\n\u003cp\u003ePubMed, Embase, and Google Scholar were searched from inception to July, 31, 2025 (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The search combined controlled vocabulary and keywords for paroxysmal nocturnal hemoglobinuria, case reports, case series, diagnostic delay, and misdiagnosis (Appendix 1).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eEligibility criteria\u003c/h3\u003e\n\u003cp\u003eWe included case reports and case series describing patients with confirmed PNH, diagnosed by high-sensitivity flow cytometry (glycosylphosphatidylinositol-deficient clones) or, in older studies, by the Ham or sucrose lysis test. Eligible cases required documentation of diagnostic delay, defined as \u0026ge;\u0026thinsp;12 months between symptom onset and confirmed diagnosis or an explicit author statement of delay/misdiagnosis. Reports initially attributed to other conditions (eg, aplastic anemia, autoimmune hemolytic anemia, myelodysplastic syndromes, urologic disorders) but later revised to PNH were eligible. Exclusions were non-human or laboratory studies, reviews, editorials, clinical trials without individual case detail, and reports lacking extractable timelines. For duplicate publications, the most complete or recent version was retained.\u003c/p\u003e\n\u003ch3\u003eStudy selection\u003c/h3\u003e\n\u003cp\u003eAll references were imported into Rayyan (Qatar Computing Research Institute) for de-duplication and blinded screening. Two reviewers independently assessed titles/abstracts and full texts. Disagreements were resolved by consensus or third-party adjudication. Study flow is summarized in the PRISMA 2020 diagram (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\n\u003ch3\u003eData extraction and management\u003c/h3\u003e\n\u003cp\u003eData were extracted in duplicate using a piloted Excel form. Variables included publication details, demographics (age, sex, race/ethnicity, comorbidities), presentation (symptoms, laboratory and imaging findings), diagnostic course (initial misdiagnosis, treatments before diagnosis, delay duration, thrombosis, clinical trigger, confirmatory test). Evidence of marrow failure was also recorded. Discrepancies were resolved by consensus.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eQuality assessment\u003c/h2\u003e \u003cp\u003eMethodological quality was appraised with JBI Critical Appraisal Checklists for case reports and case series. Two reviewers rated each study as high, moderate, or low quality; disagreements were resolved through discussion. Full item-level ratings are provided in Appendix 2.\u003csup\u003e9\u0026ndash;72\u003c/sup\u003e\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eAI-based diagnostic evaluation\u003c/h3\u003e\n\u003cp\u003eStandardized clinical vignettes were generated from extracted case data and tested across GPT-4.5, Claude-Sonnet 4, and Gemini-2.5 (versions current May\u0026ndash;July 2025). Each vignette included demographics, presenting features, laboratory and imaging findings, prior treatments, initial misdiagnoses, comorbidities, and duration of diagnostic delay (Appendix 3). Models were prompted with a standardized query requesting the five most likely diagnoses, and the rank position of PNH (1\u0026ndash;6 scale) was recorded as the primary outcome.\u003c/p\u003e \u003cp\u003eAll model queries were performed as single-pass, deterministic evaluations using default settings without temperature adjustment or repeated sampling, and each vignette was structured using a predefined template with consistent variable ordering (Appendix 3 and 4).\u003c/p\u003e \u003cp\u003eTo assess diagnostic robustness, vignettes were re-run after systematic removal of predefined diagnostic domains: (A) hemoglobinuria/hematuria, (B) thrombosis, (C) laboratory hemolysis, and (D) bone-marrow failure. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e provides a visual summary of the systematic diagnostic domain ablation workflow used to test model robustness across all 16 vignette states.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eEach omission scenario was executed independently to prevent model memory carry-over. All models were evaluated in default configuration (no fine-tuning or post-processing). Full prompt text is provided in Appendix 4, and the full information-reduction protocol is detailed in Appendix 5.\u003c/p\u003e\n\u003ch3\u003eEthics statement\u003c/h3\u003e\n\u003cp\u003eThis study used only published, de-identified data extracted from case reports and case series. No human participants were directly involved, no identifiable private information was collected, and institutional review board approval was not required.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eThe primary outcome was the diagnostic rank assigned to PNH by each model on a 1\u0026ndash;6 ordinal scale (1\u0026thinsp;=\u0026thinsp;top diagnosis; 2\u0026ndash;5\u0026thinsp;=\u0026thinsp;within top five; 6\u0026thinsp;=\u0026thinsp;not listed). Because each vignette was evaluated by all three models, paired, within-case comparisons of ranks were conducted using the Wilcoxon signed-rank test, with effect size r reported. Secondary outcomes included median diagnostic rank, performance under feature-removal scenarios, and inter-model comparisons. Continuous variables were summarized as medians with interquartile ranges (IQRs), categorical variables as counts and percentages.\u003c/p\u003e \u003cp\u003eFeature-removal scenarios were unpaired and assessed with Mann\u0026ndash;Whitney U tests (effect size Cliff\u0026rsquo;s delta, δ). Bonferroni corrections were applied for multiple testing. Hierarchical clustering was used to visualize performance across omission scenarios. Analyses were conducted in Python 3.13.5; descriptive tables were generated in Excel with the Real Statistics add-on. A two-sided adjusted \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003ePatients in the misdiagnosed cohort were younger than those in the International PNH Registry (median age 30 vs 46 years)\u003csup\u003e\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e\u003c/sup\u003e, while sex distribution was comparable (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Reports spanned all WHO regions, though representation from low- and middle-income countries was sparse. Race or ethnicity was documented in only 23.5% of cases, limiting assessment of disparities. In contrast, registry data report that 78% of patients are Caucasian. \u003csup\u003e73\u003c/sup\u003e Limited demographic reporting in case literature constrains evaluation of disparities in diagnostic delay.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u003cb\u003eClinical comparison of the misdiagnosed PNH cohort and the International PNH Registry\u003c/b\u003e Key demographic and diagnostic-delay characteristics of the misdiagnosed PNH cohort (N\u0026thinsp;=\u0026thinsp;68) compared with the International PNH Registry (Schrezenmeier et al., 2020); values shown as medians (IQR) or percentages.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCohort (N\u0026thinsp;=\u0026thinsp;68)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eInternational PNH Registry\u003c/p\u003e \u003cp\u003e(Schrezenmeier et al., 2020)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003eDemographic characteristics\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMedian age, years (IQR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30 (24\u0026ndash;49)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e46 (27\u0026ndash;61)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSex, % female\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e45.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e46\u0026ndash;55\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGeographic distribution, % \u0026dagger;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEUR 34; AMR 29;\u003c/p\u003e \u003cp\u003eWPR 13; SEAR 13;\u003c/p\u003e \u003cp\u003eEMR 7; AFR 3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e82.3 from EUR and AMR\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRace/Ethnicity reported, %\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e23.5 recorded vs 76.5\u003c/p\u003e \u003cp\u003emissing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e78 Caucasian\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDiagnostic delay\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;1 year\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e14 (20.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e40\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u0026ndash;5 years\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e36 (52.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e36\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026gt;\u0026thinsp;5 years\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e15 (22.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e24\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMedian delay (IQR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e24 months (12\u0026ndash;60)\u003c/p\u003e \u003cp\u003eRange 0.5\u0026ndash;288 months\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e22.5 months\u003c/p\u003e \u003cp\u003e(mean 1.88 year)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"3\"\u003e\u003cb\u003eFootnotes\u003c/b\u003e:\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"3\"\u003e\u0026dagger; WHO regions: EUR\u0026thinsp;=\u0026thinsp;Europe, AMR\u0026thinsp;=\u0026thinsp;Americas, WPR\u0026thinsp;=\u0026thinsp;Western Pacific, SEAR\u0026thinsp;=\u0026thinsp;South-East Asia, EMR\u0026thinsp;=\u0026thinsp;Eastern Mediterranean, AFR\u0026thinsp;=\u0026thinsp;Africa.\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"3\"\u003eDiagnostic-delay data were available for 65 of 68 cases; three reports described an initial misdiagnosis but did not specify the elapsed time to confirmed diagnosis. Percentages are calculated using the full cohort (N\u0026thinsp;=\u0026thinsp;68).\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eUnexplained hemolysis (45.6%) and thrombosis (36.8%) were the leading triggers for PNH testing, together accounting for more than three-quarters of cases (Supplementary Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). Other triggers included incidental laboratory or marrow abnormalities, hemoglobinuria, and cytopenias. Initial misdiagnoses most often involved iron-deficiency anemia, marrow failure syndromes, or autoimmune hemolysis. Hemoglobinuria was frequently attributed to urinary tract pathology rather than recognized as evidence of hemolysis. Detailed lists of less common misdiagnoses are provided in the footnotes.\u003c/p\u003e \u003cp\u003eMedian time to diagnosis was 24 months (IQR 12\u0026ndash;60), closely approximating registry data (median 22.5 months).\u003csup\u003e\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e\u003c/sup\u003e However, a larger share of patients in this cohort were diagnosed between one and five years after symptom onset (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). While central tendencies were comparable, the distribution in our cohort reflected a greater proportion of diagnoses in the one-to-five-year interval.\u003c/p\u003e \u003cp\u003eThe most prolonged delays occurred when patients were initially labeled with autoimmune/systemic inflammatory disease (median 132 months) or marrow failure syndromes such as aplastic anemia (median 41.5 months) (Supplementary Table S2). In contrast, patients presenting with atypical thrombosis were diagnosed earliest (median 3 months). Intermediate delays were observed with more common mislabels such as iron deficiency anemia and renal/urologic disorders, reflecting how prevalent diagnoses can obscure recognition of rare disease. These findings highlight that misattribution to chronic or systemic conditions prolongs diagnosis, whereas atypical thrombotic presentations prompt earlier testing.\u003c/p\u003e \u003cp\u003eOn the 1\u0026ndash;6 rank scale, GPT-4.5 achieved superior performance on complete vignettes. In paired, within-case comparisons of rank, GPT-4.5 outperformed Gemini-2.5 (Wilcoxon signed-rank, p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001, r\u0026thinsp;=\u0026thinsp;0.30) and Claude-Sonnet 4 (p\u0026thinsp;=\u0026thinsp;0.0068, r\u0026thinsp;=\u0026thinsp;0.18). As secondary summaries, GPT-4.5 included PNH within the top five in 89.7% of cases and ranked it first in 70.6%, exceeding Claude-Sonnet 4 and Gemini-2.5 (both ~\u0026thinsp;81% top-five) (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Rank stability was greater with GPT-4.5 and Gemini-2.5 than with Claude-Sonnet 4. Pairwise testing confirmed this hierarchy: GPT-4.5 outperformed Gemini-2.5 with a moderate effect size and Claude-Sonnet 4 with a smaller effect, while Claude-Sonnet 4 had only a modest edge over Gemini-2.5 (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Pairwise Wilcoxon signed-rank tests were applied because each vignette was evaluated by all three models, allowing direct within-case comparison of performance on the full dataset, with full vignettes being evaluated, similar to a clinical setting with full information. These results confirm that the performance hierarchy observed in descriptive analyses remained statistically robust when tested across paired vignettes.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u003cb\u003eModel Performance and Pairwise Comparisons of AI Diagnostic Accuracy in Misdiagnosed PNH Vignettes\u003c/b\u003e Primary outcome is per-case diagnostic rank of PNH (1\u0026thinsp;=\u0026thinsp;top diagnosis; 2\u0026ndash;5\u0026thinsp;=\u0026thinsp;within top five; 6\u0026thinsp;=\u0026thinsp;not listed in top five). Values shown are median [IQR] rank for each model on the same 68 vignettes. Paired Wilcoxon signed-rank tests compare GPT-4.5 vs Claude-Sonnet 4, GPT-4.5 vs Gemini-2.5, and Claude-Sonnet 4 vs Gemini-2.5.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel / Comparison\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e% Top-5 Rank\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e% Top-1 Rank\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMedian Rank\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eIQR\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003ep-value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eEffect Size (r)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGPT-4.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e89.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e70.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eClaude\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e80.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e63.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGemini\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e80.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e55.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGPT-4.5 vs Gemini\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.30\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGPT-4.5 vs Claude\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0068\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.18\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eClaude vs Gemini\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0218\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.16\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"7\"\u003e\u003cb\u003eFootnotes\u003c/b\u003e:\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"7\"\u003eNA indicates not applicable. Rank coding: 1\u0026thinsp;=\u0026thinsp;top diagnosis; 2\u0026ndash;5\u0026thinsp;=\u0026thinsp;within top five; 6\u0026thinsp;=\u0026thinsp;not in top five (poorest rank).\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eAlthough each model evaluated the same vignettes under complete information, information-reduction removal created non-identical vignette subsets, necessitating unpaired Mann\u0026ndash;Whitney testing; under these conditions, GPT-4.5 retained a moderate advantage over Gemini-2.5, whereas differences versus Claude-Sonnet 4 were less stable (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). These findings confirm that the performance hierarchy observed in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e persisted when models were evaluated on vignette subsets with systematically reduced diagnostic information.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u003cb\u003eFeature-removal experiments on the rank outcome (1\u0026ndash;6) across predefined omission domains.\u003c/b\u003e Between-model unpaired comparisons of diagnostic performance across systematically reduced vignette conditions, in which predefined diagnostic omission domains: A\u0026thinsp;=\u0026thinsp;hemoglobinuria/hematuria, B\u0026thinsp;=\u0026thinsp;thrombosis, C\u0026thinsp;=\u0026thinsp;laboratory hemolysis, D\u0026thinsp;=\u0026thinsp;bone-marrow failure were omitted to simulate incomplete clinical information. Two-sided p-values and effect sizes (Cliff's delta, δ) are shown, representing the magnitude of performance differences under information-reduction scenarios.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eComparison\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ep-value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCliff\u0026rsquo;s Delta (δ)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGPT-4.5 vs Gemini\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.36\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGPT-4.5 vs Claude\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.076\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.38\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eClaude vs Gemini\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.029\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.15\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"3\"\u003e\u003cb\u003eFootnotes\u003c/b\u003e:\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"3\"\u003eRank coding: 1\u0026thinsp;=\u0026thinsp;top diagnosis; 2\u0026ndash;5\u0026thinsp;=\u0026thinsp;within top five; 6\u0026thinsp;=\u0026thinsp;not in top five.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eAcross pairwise Bonferroni-adjusted comparisons of diagnostic rank under systematic feature removal, most scenarios yielded p\u0026thinsp;\u0026asymp;\u0026thinsp;1.0, indicating stable performance when a single domain was omitted or when thrombosis (B) and marrow-failure (D) features were removed in combination.\u003c/p\u003e \u003cp\u003eSignificant differences emerged only when both hemolysis cues were absent. Compared with the complete-information baseline, rank performance worsened (higher rank values) in all scenarios omitting hemoglobinuria (A) and laboratory hemolysis (C) together (p values 0.001\u0026ndash;0.012). The maximal omission scenario (No A\u0026thinsp;+\u0026thinsp;B + C\u0026thinsp;+\u0026thinsp;D) differed significantly from every comparator retaining either A or C. Within A-absent conditions, removal of C consistently produced significance, whereas removal of B or D alone did not.\u003c/p\u003e \u003cp\u003eThese results demonstrate that model accuracy was robust across incomplete inputs unless both clinical and laboratory hemolysis features were removed, which defined the threshold for performance loss.\u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e presents a hierarchical clustering heatmap of Bonferroni-adjusted Mann\u0026ndash;Whitney U test results across feature-removal scenarios. Hierarchical clustering of pairwise comparisons reinforced these findings, grouping all scenarios that retained at least one hemolysis feature into a single branch and isolating those lacking both into a distinct cluster.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe clustering dendrogram confirmed these patterns by grouping all scenarios that retained at least one hemolysis feature together and isolating those lacking both A and C into a distinct branch, reinforcing hemolysis as the critical diagnostic signal.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis systematic review and AI-based evaluation provide new insights into why PNH is frequently diagnosed late and whether LLMs can support earlier recognition. Across 68 misdiagnosed cases, both clinicians and LLMs depended heavily on evidence of hemolysis. When either hemoglobinuria or laboratory markers were absent, rank performance fell sharply. Even minimal signals such as dark urine, elevated LDH, or reduced haptoglobin proved to be the most actionable features for timely recognition.\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e,\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eGPT-4.5 identified PNH in nearly 90% of vignettes and ranked it first in 70%, outperforming Claude-Sonnet 4 and Gemini-2.5. Accuracy remained stable when thrombosis or marrow-failure features were removed but deteriorated in the absence of hemolysis cues. This pattern parallels clinical practice, where autoimmune or inflammatory mislabels often delay recognition, while hepatic, portal, or cerebral thrombosis reliably prompts testing.\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan additionalcitationids=\"CR75\" citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e\u003c/sup\u003e Nearly one in five patients were initially labeled with iron-deficiency anemia despite later confirmation of hemolysis, illustrating how common comorbidities can obscure recognition.\u003csup\u003e\u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e77\u003c/span\u003e\u003c/sup\u003e Using both Wilcoxon signed-rank tests on paired vignettes and Mann\u0026ndash;Whitney U with Cliff\u0026rsquo;s delta on reduced, unpaired scenarios strengthened confidence in these findings by demonstrating robustness under complete and incomplete information.\u003csup\u003e\u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003ePatterns of misdiagnosis in the literature reflected common clinical anchors. Iron-deficiency anemia frequently obscured recognition, consistent with reports that true iron deficiency is common in PNH.\u003csup\u003e\u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e77\u003c/span\u003e\u003c/sup\u003e Hemoglobinuria was often misattributed to urinary tract disease rather than hemolysis, a well-described diagnostic pitfall.\u003csup\u003e\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e Autoimmune hemolytic anemia, particularly during pregnancy or in association with thrombosis, led to misleading positive DAT results and partial steroid responses. Dysplastic marrow features and cytopenias frequently resulted in attribution to myelodysplastic or myelofibrosis-related anemia, while incidental marrow abnormalities were often labeled as failure syndromes without concurrent hemolysis testing.\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e,\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e\u003c/sup\u003e Renal-predominant presentations, though rarely reported, further complicated attribution of symptoms.\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e The longest diagnostic delays occurred in patients initially labeled with autoimmune or marrow-failure disorders, underscoring how these categories suppress hemolysis assessment. In contrast, thrombotic presentations were diagnosed earliest, reflecting the high specificity of hepatic, splanchnic, or cerebral venous thrombosis for PNH.\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e,\u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eCompared with international registry data, diagnostic delays in this cohort (median 24 months) were similar overall but skewed toward the one-to-five-year range.\u003csup\u003e\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e\u003c/sup\u003e Only one in five patients were diagnosed within the first year, compared with 40% in the registry, likely reflecting publication bias toward prolonged or atypical cases.\u003csup\u003e\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e,\u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e79\u003c/span\u003e\u003c/sup\u003e Outliers with extreme delays of up to 288 months highlight the heterogeneity of diagnostic pathways and suggest that systematic reflex testing strategies may be needed to capture these rare but high-impact cases.\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e\u003c/sup\u003e Patients in this series were also younger than those in registry cohorts, consistent with prior reports that PNH clones may remain clinically silent for years, especially in younger adults where cytopenias are easily attributed to alternative etiologies.\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e Sex distribution was comparable to registry data, but race and ethnicity were missing in more than three-quarters of reports. This absence not only prevents meaningful equity analyses but also raises concern that diagnostic delays may be underestimated in underrepresented populations.\u003csup\u003e\u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e80\u003c/span\u003e\u003c/sup\u003e The geographic skew toward Europe and the Americas further suggests under-ascertainment in regions with limited awareness and diagnostic access.\u003c/p\u003e \u003cp\u003eThese findings reinforce two priorities for practice. First, clinicians should maintain a low threshold to test for hemolysis in patients with unexplained cytopenias, dark urine, or atypical thrombosis, even when initial impressions favor more common conditions such as iron deficiency or autoimmune hemolysis.\u003csup\u003e\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e,\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e\u003c/sup\u003e A reflex panel including LDH, haptoglobin, DAT, and high-sensitivity PNH flow cytometry at the time of marrow-failure labeling may represent the single highest-yield systems change.\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e\u003c/sup\u003e Second, LLMs may have potential as adjunctive diagnostic tools. Their rank performance on the 1\u0026ndash;6 scale across incomplete vignettes suggests feasibility within real-world electronic medical records (EMR), where missing information is common.\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e Early applications of machine learning to PNH detection in primary care datasets have demonstrated feasibility,\u003csup\u003e7\u003c/sup\u003e and embedding automated alerts that combine hemolysis markers with atypical thrombosis could trigger earlier suspicion.\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e,\u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e\u003c/sup\u003e Beyond system-level alerts, such tools could help reduce cognitive biases by countering anchoring on common diagnoses, drawing attention to atypical features, and encouraging reconsideration when clinical and laboratory findings are discordant. By highlighting overlooked signals, LLMs may prompt nonspecialists to initiate earlier referral to hematology/oncology and expedite definitive work-up, thereby shortening time to recognition. Future work should validate these findings prospectively within clinical workflows, such as hematology referral triggers, diagnostic decision-support tools, or prospective registry studies, alongside electronic health record datasets, while also exploring prompt standardization across LLM architectures and assessing model interpretability to identify which clinical features drive AI recognition.\u003c/p\u003e \u003cp\u003eSeveral limitations merit consideration. Case-report literature is biased toward unusual or prolonged diagnostic journeys, inflating delay estimates and underrepresenting milder cases.\u003csup\u003e\u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e79\u003c/span\u003e,\u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e81\u003c/span\u003e\u003c/sup\u003e Publication bias and selective reporting further constrain generalizability. Demographic detail was sparse, limiting equity analyses.\u003csup\u003e\u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e80\u003c/span\u003e\u003c/sup\u003e AI performance was tested on standardized vignettes rather than raw clinical records, so real-world validation will require testing within electronic health records, or within further real-world studies using LLMs, where noise and contradictions are common.\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e Finally, the wide distribution of delays indicates that a small subset of patients experiences extreme diagnostic journeys; targeted interventions for this group may yield disproportionate benefit even if median times remain unchanged.\u003c/p\u003e \u003cp\u003eThe increasing integration of large language model\u0026ndash;based tools into electronic medical record platforms makes this a particularly timely area for validation. As health systems adopt AI-assisted documentation, summarization, and clinical decision-support systems, embedding standardized, diagnostic prompts within these workflows offers a practical and scalable pathway to evaluate real-world impact on diagnostic delay.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eIn this systematic review of misdiagnosed PNH, both clinicians and large language models depended primarily on hemolysis cues for recognition. Model performance remained robust when thrombosis or marrow-failure features were absent but declined when both clinical hemoglobinuria-hematuria and laboratory hemolysis features were removed, reinforcing hemolysis as the critical diagnostic signal. These findings support earlier hemolysis-directed evaluation in patients with unexplained cytopenias, dark urine, or atypical thrombosis, and support further validation of AI-assisted diagnostic support within real-world clinical workflows to shorten time to PNH recognition.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eEthics approval and consent to participate\u003c/h2\u003e\n\u003cp\u003eThis study used only published, de-identified data extracted from case reports and case series. No human participants were directly involved, no identifiable private information was collected, and institutional review board approval was not required.\u003c/p\u003e\n\u003ch2\u003eConsent for publication\u003c/h2\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003ch2\u003eAvailability of data and materials\u003c/h2\u003e\n\u003cp\u003eAll data generated or analysed during this study are included in this published article and its supplementary information files. The supplementary appendices provide the complete database search strategies, JBI critical appraisal tables, the standardized data extraction template, the full structured prompts used for AI evaluation, and the information-reduction protocol applied to test diagnostic robustness. Additional clarifications about the analytic workflow, statistical code, or prompt implementation are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003ch2\u003eCompeting interests\u003c/h2\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003ch2\u003eFunding\u003c/h2\u003e\n\u003cp\u003eNo specific funding was received for this work.\u003c/p\u003e\n\u003ch2\u003eAuthors\u0026apos; contributions\u003c/h2\u003e\n\u003cp\u003eAMB conceived and designed the study, collected and verified data, performed statistical analysis, prepared figures and tables, drafted the manuscript, coordinated revisions, and serves as guarantor of the work. MB and AA contributed to study design, quality review, and manuscript editing. ML contributed to manuscript editing. DB participated in article screening, contributed to case review, data extraction, and manuscript editing. SP, BA, and CB provided clinical input and critical review of the manuscript. KQ contributed to study conceptualization, data interpretation, and critical revision of the manuscript for important intellectual content. DI supervised the project, advised on study design and analysis, and provided critical review of the final manuscript. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003ch2\u003eAcknowledgements\u003c/h2\u003e\n\u003cp\u003eThe authors thank the Department of Hematology/Oncology at Karmanos Cancer Institute at McLaren Greater Lansing Hospital for academic support and guidance. Portions of this work were previously presented as an abstract at the 2025 American Society of Hematology Annual Meeting.\u003c/p\u003e\n\u003ch2\u003eAuthors\u0026apos; information\u003c/h2\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eShi JJ, Ozcan YM, Santos CIA, Patel H, Shammo J, Bat T. Current landscape of paroxysmal nocturnal hemoglobinuria in the era of complement inhibitors and regulators. Ther Adv Hematol. 2024;15:20406207241307500. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/20406207241307500\u003c/span\u003e\u003cspan address=\"10.1177/20406207241307500\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eColden MA, Kumar S, Munkhbileg B, Babushok DV. Insights into the emergence of paroxysmal nocturnal hemoglobinuria. Front Immunol. 2022;12:830172. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fimmu.2021.830172\u003c/span\u003e\u003cspan address=\"10.3389/fimmu.2021.830172\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSpychalska J, Duńska M, Myślińska A, Majewska-Wierzbicka M, Klimczak-Jajor E, Głodkowska-Mr\u0026oacute;wka E. Diagnostic landscape of first-time cytometric screening for paroxysmal nocturnal hemoglobinuria in Poland in 2013\u0026ndash;2022. Orphanet J Rare Dis. 2024;19(1):271. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s13023-024-03283-x\u003c/span\u003e\u003cspan address=\"10.1186/s13023-024-03283-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDevalet B, Mullier F, Chatelain B, Dogn\u0026eacute; J, Chatelain C. Pathophysiology, diagnosis, and treatment of paroxysmal nocturnal hemoglobinuria: a review. Eur J Haematol. 2015;95(3):190\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1111/ejh.12543\u003c/span\u003e\u003cspan address=\"10.1111/ejh.12543\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHillmen P, Hall C, Marsh JCW, et al. Effect of eculizumab on hemolysis and transfusion requirements in patients with paroxysmal nocturnal hemoglobinuria. N Engl J Med. 2004;350(6):552\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1056/NEJMoa031688\u003c/span\u003e\u003cspan address=\"10.1056/NEJMoa031688\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBabushok DV. When does a PNH clone have clinical significance? Hematology. 2021;2021(1):143\u0026ndash;52. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1182/hematology.2021000245\u003c/span\u003e\u003cspan address=\"10.1182/hematology.2021000245\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorker A, Mahon H, Sams J, et al. A machine learning algorithm for the detection of paroxysmal nocturnal haemoglobinuria (PNH) in UK primary care electronic health records. Orphanet J Rare Dis. 2024;19(1):378. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s13023-024-03406-4\u003c/span\u003e\u003cspan address=\"10.1186/s13023-024-03406-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBowen A, Baraka D. AI-Assisted Recognition of Delayed or Misdiagnosed Paroxysmal Nocturnal Hemoglobinuria. PROSPERO International Prospective Register of Systematic Reviews; 2025. Accessed October 2, 2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.crd.york.ac.uk/PROSPERO/display_record.php?ID=CRD420251104533\u003c/span\u003e\u003cspan address=\"https://www.crd.york.ac.uk/PROSPERO/display_record.php?ID=CRD420251104533\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePokhrel B, Gautam S, Khanal S, Pokhrel NB, Shrestha A. A rare and misdiagnosed entity paroxysmal nocturnal hemoglobinuria: a case report. Cureus Published online May. 2021;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.7759/cureus.14902\u003c/span\u003e\u003cspan address=\"10.7759/cureus.14902\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eElias NS, Riedl J, Stouten K, et al. Abdominal pain in combination with an unexplained hemolytic anemia are crucial signs to test for paroxysmal nocturnal hemoglobinuria: A case report. Clin Case Rep. 2019;7(1):175\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/ccr3.1771\u003c/span\u003e\u003cspan address=\"10.1002/ccr3.1771\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePerdigones N, Morales M, Mason P, Bessler M. Case Report: Paroxysmal nocturnal hemoglobinuria in a woman heterozygous for G6PD A-. F1000Research. 2014;3:194. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.12688/f1000research.4980.2\u003c/span\u003e\u003cspan address=\"10.12688/f1000research.4980.2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEisa MS, Mohamed SF, Ibrahim F, et al. Paroxysmal nocturnal hemoglobinuria with glucose-6-phosphate dehydrogenase deficiency: a case report and review of the literature. Case Rep Oncol. 2019;12(3):838\u0026ndash;44. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1159/000503817\u003c/span\u003e\u003cspan address=\"10.1159/000503817\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKeating S, Machchhar R, Jain U, Naronowicz G, Lipschutz J. An interesting case of paroxysmal nocturnal hemoglobinuria with renal involvement. Cureus Published online July. 2024;5. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.7759/cureus.63917\u003c/span\u003e\u003cspan address=\"10.7759/cureus.63917\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFarooq Q, Saleem MW, Khan ZU, Hadi N. Paroxysmal nocturnal hemoglobinuria: a diagnostic zero-sum-game. Cureus. Published online Dec. 2020;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.7759/cureus.11956\u003c/span\u003e\u003cspan address=\"10.7759/cureus.11956\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiang HY, Xie XD, Jing GX, Wang M, Yu Y, Cui JF. Posthepatectomy jaundice induced by paroxysmal nocturnal hemoglobinuria: A case report. World J Clin Cases. 2021;9(32):10046\u0026ndash;51. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.12998/wjcc.v9.i32.10046\u003c/span\u003e\u003cspan address=\"10.12998/wjcc.v9.i32.10046\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiao M, Di Y, An Y, Cui Y. Acute Necrotizing Pancreatitis Complicated with Paroxysmal Nocturnal Haemoglobinuria: A Case Report. Clin Exp Gastroenterol 2024;Volume 17:141\u0026ndash;5. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2147/CEG.S449594\u003c/span\u003e\u003cspan address=\"10.2147/CEG.S449594\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAthota S, Gopalan S, Arthur P, Jayaram A, Satish A. Deciphering paroxysmal nocturnal hemoglobinuria: an unusual paradigm of hemolytic Anemia. Cureus Published online August. 2024;19. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.7759/cureus.67194\u003c/span\u003e\u003cspan address=\"10.7759/cureus.67194\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLumori BAE, Muyanja D. Diagnostic and treatment challenges of paroxysmal nocturnal hemoglobinuria in Uganda. Case Rep Hematol. 2019;2019:1\u0026ndash;5. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1155/2019/7897509\u003c/span\u003e\u003cspan address=\"10.1155/2019/7897509\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQuinquenel A, Maestraggi Q, Lecoq-Lafon C, R\u0026eacute;gis PDL, Delmer A, Servettaz A. Atypical presentation of paroxysmal nocturnal hemoglobinuria treated by eculizumab: A case report. Med (Baltim). 2017;96(12):e6403. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1097/MD.0000000000006403\u003c/span\u003e\u003cspan address=\"10.1097/MD.0000000000006403\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFattizzo B, Pasquale R, Bellani V, Barcellini W, Kulasekararaj AG. Complement mediated hemolytic anemias in the COVID-19 era: case series and review of the literature. Front Immunol. 2021;12:791429. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fimmu.2021.791429\u003c/span\u003e\u003cspan address=\"10.3389/fimmu.2021.791429\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKorkama ES, Armstrong AE, Jarva H, Meri S. Spontaneous remission in paroxysmal nocturnal hemoglobinuria\u0026mdash;return to health or transition into malignancy? Front Immunol. 2018;9:1749. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fimmu.2018.01749\u003c/span\u003e\u003cspan address=\"10.3389/fimmu.2018.01749\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBurns E, Anand K, Acosta G, et al. Autosplenectomy in a patient with paroxysmal nocturnal hemoglobinuria (PNH). Case Rep Hematol. 2019;2019:1\u0026ndash;5. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1155/2019/3146965\u003c/span\u003e\u003cspan address=\"10.1155/2019/3146965\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheekati M, Siu K, Ochs R. Management of paroxysmal nocturnal hemoglobinuria in CALR mutated post-essential thrombocythemia myelofibrosis: A case report. eJHaem. 2024;5(2):392\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/jha2.892\u003c/span\u003e\u003cspan address=\"10.1002/jha2.892\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePark S, So MK, Cho MS, Kim DY, Huh J. Coexistence of primary myelofibrosis and paroxysmal nocturnal hemoglobinuria clone with JAK2 V617F, U2AF1 and SETBP1 mutations: a case report and brief review of literature. Diagnostics. 2021;11(9):1644. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/diagnostics11091644\u003c/span\u003e\u003cspan address=\"10.3390/diagnostics11091644\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGediz F, Payzin B, Cakmak O, Uzum Y, Ernur D, Sahin F. Paroxysmal nocturnal hemoglobinuria in the differential diagnosis of thrombocytopenia. Hematol Rep. 2017;9(1):6862. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4081/hr.2017.6862\u003c/span\u003e\u003cspan address=\"10.4081/hr.2017.6862\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFigueroa-Jim\u0026eacute;nez LA, Gonz\u0026aacute;lez-M\u0026aacute;rquez AL, Alicea-Guevara R, et al. A Hispanic female patient with heartburn: A rare presentation of Paroxysmal Nocturnal Hemoglobinuria. Bol Asoc Med P R. 2015;107(3):9\u0026ndash;12.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChoi H, Yim H, Lee MJ. Hemosiderin induced acute kidney injury requiring hemodialysis in patients with paroxysmal nocturnal hemoglobinuria: A case report. Med (Baltim). 2023;102(40):e35412. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1097/MD.0000000000035412\u003c/span\u003e\u003cspan address=\"10.1097/MD.0000000000035412\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBalwani MR, Kute VB, Shah PR, Shah M, Varyani U, Trivedi HL. Manifestation of paroxysmal nocturnal hemoglobinuria as repeated acute kidney injury. J Nephropharmacology. 2016;5(2):116\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDambal A. Paroxysmal nocturnal haemoglobinuria masquerading as malaria: a case report. J Clin Diagn Res Published online. 2015. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.7860/JCDR/2015/13465.6329\u003c/span\u003e\u003cspan address=\"10.7860/JCDR/2015/13465.6329\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGervasi F, D\u0026rsquo;Amelio L, Trizzino A, et al. Cerebral stroke in a teenage girl with paroxysmal nocturnal hemoglobinuria. Hematol Rep. 2017;9(2):7012. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4081/hr.2017.7012\u003c/span\u003e\u003cspan address=\"10.4081/hr.2017.7012\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRaghupathy R, Derman O. Response of paroxysmal nocturnal hemoglobinuria clone with aplastic anemia to rituximab. Case Rep Hematol. 2012;2012:1\u0026ndash;5. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1155/2012/106182\u003c/span\u003e\u003cspan address=\"10.1155/2012/106182\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRayas J, Hassan M, Hock RA, et al. Attack of the clones: a patient with untreated aplastic anemia presenting with classical paroxysmal nocturnal hemoglobinuria. Cureus Published online January. 2023;23. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.7759/cureus.34093\u003c/span\u003e\u003cspan address=\"10.7759/cureus.34093\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDe Boysson H, Ch\u0026egrave;ze S, Chapon F, Le Mauff B, Auzary C, Geffray L. Eosinophilic fasciitis with paroxysmal nocturnal hemoglobinuria. Joint Bone Spine. 2013;80(2):208\u0026ndash;10. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jbspin.2012.07.008\u003c/span\u003e\u003cspan address=\"10.1016/j.jbspin.2012.07.008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShakir A, Narula S, LaGrow A, Nusrat S. Two roads diverge: treatment choice in coexisting severe aplastic anemia and paroxysmal nocturnal hemoglobinuria. J Med Cases. 2020;11(6):182\u0026ndash;4. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.14740/jmc3478\u003c/span\u003e\u003cspan address=\"10.14740/jmc3478\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMemon A, Khan R, Rauf MU, Shafique K. Paroxysmal nocturnal hemoglobinuria presenting as cerebral venous sinus thrombosis: a case report. Int Arch Med. 2014;7(1):39. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/1755-7682-7-39\u003c/span\u003e\u003cspan address=\"10.1186/1755-7682-7-39\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMandala E, Lafaras C, Goulis I, Tsioni K, Georgopoulou V, Ilonidis G. Treatment of a patient with classical paroxysmal nocturnal hemoglobinuria and Budd-Chiari syndrome, with complement inhibitor eculizumab: Case Report. Hippokratia. 2013;17(1):81\u0026ndash;4.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrodsky A, Mazzocchi O, S\u0026aacute;nchez F, Khursigara G, Malhotra S, Volpacchio M. Eculizumab in paroxysmal nocturnal hemoglobinuria with Budd-Chiari syndrome progressing despite anticoagulation. Exp Hematol Oncol. 2012;1(1):26. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/2162-3619-1-26\u003c/span\u003e\u003cspan address=\"10.1186/2162-3619-1-26\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSatish S, Rajesh R, Kurian G, Seethalekshmi N, Unni M, Unni V. Recurrent acute renal failure. Indian J Nephrol. 2010;20(4):217. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4103/0971-4065.73444\u003c/span\u003e\u003cspan address=\"10.4103/0971-4065.73444\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLauritsch-Hernandez LS, Kraehenmann F, Balabanov S, Kimmich N. Eculizumab application during pregnancy in a patient with paroxysmal nocturnal hemoglobinuria: A case report with review of the literature. Clin Case Rep. 2018;6(8):1582\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/ccr3.1634\u003c/span\u003e\u003cspan address=\"10.1002/ccr3.1634\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePatel PS, Singh PP, Archana A, Kumar O. A series of eight cases of pigment nephropathy: an obscured aspect of acute kidney injury. Cureus Published online July. 2024;10. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.7759/cureus.64214\u003c/span\u003e\u003cspan address=\"10.7759/cureus.64214\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBais J, Pel M, Von Dem Borne A, Van Der Lelie H. Pregnancy and paroxysmal nocturnal hemoglobinuria. Eur J Obstet Gynecol Reprod Biol. 1994;53(3):211\u0026ndash;4. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/0028-2243(94)90121-X\u003c/span\u003e\u003cspan address=\"10.1016/0028-2243(94)90121-X\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSantagostino A, Lombardi L, Dine G, Hirsch P, Misra SC. Paroxysmal nocturnal hemoglobinuria with a distinct molecular signature diagnosed ten years after allogenic bone marrow transplantation for acute myeloid leukemia. Case Rep Hematol. 2019;2019:1\u0026ndash;3. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1155/2019/8928623\u003c/span\u003e\u003cspan address=\"10.1155/2019/8928623\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBhat LSR, Taduri G, Guditi S, Herur S, Bukka VC, Murugan P. An unusual case of relapsing and remitting acute kidney injury. Indian J Nephrol. 2024;34:665\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.25259/ijn_437_23\u003c/span\u003e\u003cspan address=\"10.25259/ijn_437_23\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDe-la-Iglesia S, Luzardo H, Lemes A, et al. Positive impact of eculizumab therapy on surgery for budd- chiari syndrome in a patient with paroxysmal nocturnal hemoglobinuria and a longterm history of thrombosis. Hematol Rep. 2016;8(3):34\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4081/hr.2016.6562\u003c/span\u003e\u003cspan address=\"10.4081/hr.2016.6562\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMigdady Y, Olszewski AJ. Paroxysmal nocturnal hemoglobinuria presenting with a left intraventricular thrombus in a patient with prior thymoma and aplastic anemia. Clin Med Insights Oncol. 2011;5. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4137/CMO.S8442\u003c/span\u003e\u003cspan address=\"10.4137/CMO.S8442\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. CMO.S8442.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVogel SJ, Reinhard EH. Paroxysmal nocturnal hemoglobinuria associated with infectious mononucleosis. Blood. 1979;54(2):351\u0026ndash;3.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHegenbart U, Niederwieser D, Forman S, et al. Hematopoietic cell transplantation from related and unrelated donors after minimal conditioning as a curative treatment modality for severe paroxysmal nocturnal hemoglobinuria. Biol Blood Marrow Transpl. 2003;9(11):689\u0026ndash;97. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S1083-8791(03)00264-7\u003c/span\u003e\u003cspan address=\"10.1016/S1083-8791(03)00264-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheng KL, Brody J, Warshall CE, Sloand EM, Allen SL. Paroxysmal nocturnal hemoglobinuria following alemtuzumab immunosuppressive therapy for myelodysplastic syndrome and complicated by recurrent life-threatening thrombosis despite anticoagulation: Successful intervention with eculizumab and fondaparinux. Leuk Res. 2010;34(4):e85\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.leukres.2009.09.024\u003c/span\u003e\u003cspan address=\"10.1016/j.leukres.2009.09.024\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePalmieri G, Selleri C, Montella L, et al. Thymoma followed by paroxysmal nocturnal hemoglobinuria: A unique clinical association in the context of multiorgan autoimmunity with a potential role for CD8\u0026thinsp;+\u0026thinsp;T lymphocytes. Am J Hematol. 2006;81(10):774\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/ajh.20699\u003c/span\u003e\u003cspan address=\"10.1002/ajh.20699\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKawamoto M, Murakami Y, Kinoshita T, Kohara N. Recurrent aseptic meningitis with PIGT mutations: a novel pathogenesis of recurrent meningitis successfully treated by eculizumab. BMJ Case Rep. 2018;2018:bcr\u0026ndash;2018. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/bcr-2018-225910\u003c/span\u003e\u003cspan address=\"10.1136/bcr-2018-225910\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJackson GH, Noble RS, Maung ZT, Main J, Smith SR, Reid MM. Severe haemolysis and renal failure in a patient with paroxysmal nocturnal haemoglobinuria. J Clin Pathol. 1992;45(2):176\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/jcp.45.2.176\u003c/span\u003e\u003cspan address=\"10.1136/jcp.45.2.176\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu H, He C, Zhu H, et al. A paroxysmal nocturnal haemoglobinuria progress with waldenstr\u0026ouml;m macroglobulinemia along with t cell monoclonal expansion. Indian J Hematol Blood Transfus. 2014;30(S1):227\u0026ndash;31. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s12288-014-0337-9\u003c/span\u003e\u003cspan address=\"10.1007/s12288-014-0337-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlsalman Z, Alsalman M, Albesher M, et al. Primary myelofibrosis with concurrent paroxysmal nocturnal haemoglobinuria presenting with erectile dysfunction. Oxf Med Case Rep. 2022;2022(5):omac047. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/omcr/omac047\u003c/span\u003e\u003cspan address=\"10.1093/omcr/omac047\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ede Ruiter ED, Baggen MGA, Middelkoop MP. A rare cause of haemolytic anaemia: paroxysmal nocturnal haemoglobinuria in an elderly patient. Neth J Med. 2003;61(5):174\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePande N, Bhat R, Singh BMK, Vivek G. Haemolytic anaemia\u0026mdash;PNH type II cells presenting a diagnostic dilemma. BMJ Case Rep. 2014;2014:bcr2014204174. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/bcr-2014-204174\u003c/span\u003e\u003cspan address=\"10.1136/bcr-2014-204174\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChandran AV, Nair R, Chandran PKG. The case | acute renal injury in a young burmese immigrant. Kidney Int. 2014;85(3):719\u0026ndash;20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/ki.2013.429\u003c/span\u003e\u003cspan address=\"10.1038/ki.2013.429\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheng ZJ, Shen YY, Warsame IM, Dai TM, Tu JL. Moyamoya syndrome caused by paroxysmal nocturnal hemoglobinuria. Chin Med J (Engl). 2018;131(23):2874\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4103/0366-6999.246065\u003c/span\u003e\u003cspan address=\"10.4103/0366-6999.246065\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMita M, Shichishima T, Noji H, Takahashi H, Nakamura K, Ikezoe T. Independent paroxysmal nocturnal hemoglobinuria and myelodysplastic syndrome clones in a patient with complete bone marrow failure. HemaSphere. 2018;2(5):e142. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1097/HS9.0000000000000142\u003c/span\u003e\u003cspan address=\"10.1097/HS9.0000000000000142\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTownsley DM, Young NS. Blood consult: paroxysmal nocturnal hemoglobinuria and its complications. Blood. 2013;122(16):2795\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1182/blood-2013-07-360081\u003c/span\u003e\u003cspan address=\"10.1182/blood-2013-07-360081\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSica M, Pellecchia A, De Angioletti M, et al. Complement-mediated oxidative damage of red cells impairs response to eculizumab in a G6PD-deficient patient with PNH. Blood. 2020;136(26):3082\u0026ndash;5. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1182/blood.2020007780\u003c/span\u003e\u003cspan address=\"10.1182/blood.2020007780\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAbideen ZU, Jafar MS, Hameed N, Malik A. Haemoglobinuria and portal venous thrombosis in a young male. J Ayub Med Coll Abbottabad JAMC. 2017;29(2):353\u0026ndash;4.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArzour PH, Hadjami DD, Haddoum PF. Paroxysmal nocturnal hemoglobinuria (PNH): so much for a first diagnosis: a case report. East Afr Sch J Med Sci. 2024;7(09):386\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.36349/easms.2024.v07i09.003\u003c/span\u003e\u003cspan address=\"10.36349/easms.2024.v07i09.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMancuso S, Sucato G, Carlisi M, et al. Paroxysmal nocturnal hemoglobinuria: When delay in diagnosis and long therapy occurs. Hematol Rep. 2018;10(1). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4081/hr.2018.7523\u003c/span\u003e\u003cspan address=\"10.4081/hr.2018.7523\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSandeep, Sharma P, Ahluwalia J, et al. Primary bone marrow T-cell/histiocyte-rich large B-cell lymphoma: a diagnostic challenge. Hematology. 2013;18(2):85\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1179/1607845412Y.0000000037\u003c/span\u003e\u003cspan address=\"10.1179/1607845412Y.0000000037\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVeerreddy P. Hemoglobinuria misidentified as hematuria: review of discolored urine and paroxysmal nocturnal hemoglobinuria. Clin Med Insights Blood Disord. 2013;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4137/CMBD.S11517\u003c/span\u003e\u003cspan address=\"10.4137/CMBD.S11517\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. CMBD.S11517.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eParlier V, Tiainen M, Beris Ph, Miescher PA, Kimuutila S, Bellomo MJ. Trisomy 8 detection in granulomonocytic, erythrocytic and megakaryocytic lineages by chromosomal in situ suppression hybridization in a case of refractory anaemia with ringed sideroblasts complicating the course of paroxysmal nocturnal haemoglobinuria. Br J Haematol. 1992;81(2):296\u0026ndash;304. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1111/j.1365-2141.1992.tb08223.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1365-2141.1992.tb08223.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKirkizlar O, Kendir M, Karaali Z, et al. Acute renal failure in a patient with severe hemolysis. Int Urol Nephrol. 2007;39(2):651\u0026ndash;4. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s11255-006-9096-3\u003c/span\u003e\u003cspan address=\"10.1007/s11255-006-9096-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHenderson C, Lo M, Massey G. Pediatric Paroxysmal Nocturnal Hemoglobinuria Presenting as Acute Kidney Injury. J Pediatr Hematol Oncol. 2021;43(4):e543\u0026ndash;5. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1097/MPH.0000000000001847\u003c/span\u003e\u003cspan address=\"10.1097/MPH.0000000000001847\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMisra UK, Kalita J, Bansal V, Nair PP. Paroxysmal nocturnal haemoglobinuria presenting as cerebral venous sinus thrombosis. Transfus Med. 2008;18(5):308\u0026ndash;11. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1111/j.1365-3148.2008.00886.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1365-3148.2008.00886.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBart JB. Paroxysmal nocturnal hemoglobinuria: report of a case with refractory megaloblastic-hypoplastic bone marrow, bizarre erythrocyte morphology, and a postive coombs\u0026rsquo; test. Arch Intern Med. 1967;120(4):487. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1001/archinte.1967.04410010101015\u003c/span\u003e\u003cspan address=\"10.1001/archinte.1967.04410010101015\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYin XL, Zhou TH, Peng L, et al. A case report of concomitant paroxysmal nocturnal hemoglobinuria and heterozygous β-thalassemia. Ann Hematol. 2011;90(3):355\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00277-010-1011-4\u003c/span\u003e\u003cspan address=\"10.1007/s00277-010-1011-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoss JD, Rosenbaum E. Paroxysmal nocturnal hemoglobinuria presenting as aplastic anemia in a child. Am J Med. 1964;37(1):130\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/0002-9343(64)90217-7\u003c/span\u003e\u003cspan address=\"10.1016/0002-9343(64)90217-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchrezenmeier H, R\u0026ouml;th A, Araten DJ, et al. Baseline clinical characteristics and disease burden in patients with paroxysmal nocturnal hemoglobinuria (PNH): updated analysis from the International PNH Registry. Ann Hematol. 2020;99(7):1505\u0026ndash;14. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00277-020-04052-z\u003c/span\u003e\u003cspan address=\"10.1007/s00277-020-04052-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eParker CJ. Update on the diagnosis and management of paroxysmal nocturnal hemoglobinuria. Hematology. 2016;2016(1):208\u0026ndash;16. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1182/asheducation-2016.1.208\u003c/span\u003e\u003cspan address=\"10.1182/asheducation-2016.1.208\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrodsky RA. How I treat paroxysmal nocturnal hemoglobinuria. Blood. 2021;137(10):1304\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1182/blood.2019003812\u003c/span\u003e\u003cspan address=\"10.1182/blood.2019003812\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKokoris S, Polyviou A, Evangelidis P, et al. Thrombosis in paroxysmal nocturnal hemoglobinuria (PNH): from pathogenesis to treatment. Int J Mol Sci. 2024;25(22):12104. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/ijms252212104\u003c/span\u003e\u003cspan address=\"10.3390/ijms252212104\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePeng G, Yang W, Jing L, et al. Iron deficiency in patients with paroxysmal nocturnal hemoglobinuria: a cross-sectional survey from a single institution in China. Med Sci Monit. 2018;24:7256\u0026ndash;63. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.12659/MSM.910614\u003c/span\u003e\u003cspan address=\"10.12659/MSM.910614\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAltman DG. Practical Statistics for Medical Research. 0 ed. Chapman and Hall/CRC; 1990. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1201/9780429258589\u003c/span\u003e\u003cspan address=\"10.1201/9780429258589\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSong F, Parekh S, Hooper L, et al. Dissemination and publication of research findings: an updated review of related biases. Health Technol Assess. 2010;14(8). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3310/hta14080\u003c/span\u003e\u003cspan address=\"10.3310/hta14080\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFlanagin A, Frey T, Christiansen SL, AMA Manual of Style Committee. Updated guidance on the reporting of race and ethnicity in medical and science journals. JAMA. 2021;326(7):621. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1001/jama.2021.13304\u003c/span\u003e\u003cspan address=\"10.1001/jama.2021.13304\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGagnier JJ, Kienle G, Altman DG, Moher D, Sox H, Riley D. The CARE guidelines: consensus-based clinical case report guideline development. J Clin Epidemiol. 2014;67(1):46\u0026ndash;51. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jclinepi.2013.08.003\u003c/span\u003e\u003cspan address=\"10.1016/j.jclinepi.2013.08.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"orphanet-journal-of-rare-diseases","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ojrd","sideBox":"Learn more about [Orphanet Journal of Rare Diseases](http://ojrd.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/ojrd/default.aspx","title":"Orphanet Journal of Rare Diseases","twitterHandle":"@bmc","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"paroxysmal nocturnal hemoglobinuria, diagnostic delay, misdiagnosis, rare disease, large language models, hemolysis","lastPublishedDoi":"10.21203/rs.3.rs-9383995/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9383995/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eParoxysmal nocturnal hemoglobinuria (PNH) is often misdiagnosed, delaying effective therapy. We systematically reviewed delayed or misdiagnosed PNH cases and evaluated whether large language models could recognize PNH within the differential diagnosis using standardized clinical vignettes.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eA systematic review identified 63 publications describing 68 patients with confirmed PNH and diagnostic delay. Each case was converted into a standardized vignette and tested across 3 large language models under complete and feature-removal conditions, with diagnostic rank as the primary outcome. One model outperformed the others, identifying PNH in 89.7% of vignettes and ranking it first in 70.6%, exceeding the approximately 81% top-five rate of the other models. Rank performance remained stable without thrombosis or marrow-failure features but declined when both clinical hemoglobinuria-hematuria and laboratory hemolysis cues were absent, establishing hemolysis as the minimal diagnostic signal. Testing was typically triggered by unexplained hemolysis, 45.6%, or thrombosis, 36.8%, while frequent mislabels included iron-deficiency anemia, aplastic anemia, myelodysplastic syndromes, and urinary tract disease. Patients were younger than registry cohorts, median 30 versus 46 years, and had prolonged delays, median 24 months, IQR 12\u0026ndash;60. Race and ethnicity were reported in 23.5% of cases. Limitations include reliance on case literature and vignette testing.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eLarge language models reliably recognized PNH when hemolysis features were present and failed when absent, mirroring clinical pitfalls. Earlier hemolysis-directed evaluation and future integration of large language model-assisted screening or electronic medical record evaluation may help shorten diagnostic timelines and expedite hematology referral.\u003c/p\u003e","manuscriptTitle":"AI-Assisted Recognition of Misdiagnosed Paroxysmal Nocturnal Hemoglobinuria","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-23 09:16:26","doi":"10.21203/rs.3.rs-9383995/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-05-15T10:39:46+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"176497495165660177389736368503865608770","date":"2026-04-15T07:33:28+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"195628699199534613905188670369795358557","date":"2026-04-15T01:08:28+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-14T21:09:44+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-14T20:29:01+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-14T10:09:30+00:00","index":"","fulltext":""},{"type":"submitted","content":"Orphanet Journal of Rare Diseases","date":"2026-04-11T02:44:39+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"orphanet-journal-of-rare-diseases","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ojrd","sideBox":"Learn more about [Orphanet Journal of Rare Diseases](http://ojrd.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/ojrd/default.aspx","title":"Orphanet Journal of Rare Diseases","twitterHandle":"@bmc","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"8d4d6b81-8ef1-4ec7-bbd0-297eb0bbaa28","owner":[],"postedDate":"April 23rd, 2026","published":true,"recentEditorialEvents":[{"type":"editorInvitedReview","content":"","date":"2026-05-15T10:39:46+00:00","index":19,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-04-23T09:16:26+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-23 09:16:26","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9383995","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9383995","identity":"rs-9383995","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}