Serum levels of stearic and dihomo-γ-linolenic acids can be used to diagnose cervical cancer and cervical intraepithelial neoplasia.

Serum samples were obtained from 50 patients with CC, 30 patients with CIN2/3, and 40 healthy donors (Table 1 ). Serum samples from CC and CIN2/3 patients were obtained at the Department of Gynecology and Obstetrics, Nihon University Hospital (Tokyo, Japan) between 2017 and 2024. Serum samples from healthy donors were collected at Nihon University and Keio University (Tokyo, Japan). All samples were stored at − 150 °C until use. Clinicopathological information of the patients was collected from clinical records and pathology reports. CC patients were classified from stage I to stage IV according to the clinical staging system of the International Federation of Gynecology and Obstetrics (FIGO) 2018 33 . Serum samples were collected randomly from consenting patients before removal of the tumor at the time of initial surgery. To avoid any potential bias, each serum sample was collected prior to any treatment. Additionally, cases treated with neoadjuvant chemotherapy were excluded from sample collection. We obtained written informed consent from patients for the use of clinical materials for research purposes. SCC-Ag levels in samples from healthy donors and CIN2/3 patients were measured using a commercial service (SRL, Tokyo, Japan). SCC-Ag levels in samples from CC patients were collected from clinical records and pathology reports. Prior to GC–MS analysis, 0.3 mL of PBS containing internal standards (100 ng of margaric acid) was added to 20 µL of serum and mixed by vortexing. FFAs were extracted from samples using an ISOLUTE SLE + column (Biotage, Vimpelgatan, Uppsala, Sweden) and dichloromethane. The lower organic region of the sample was collected and dried under a nitrogen stream. The residue was dissolved in 5 µL of pyridine and 30 µL of the reagent BSTFA + TMCS (99:1) (TS-38831, Thermo Fisher Scientific, Waltham, MA, USA) for trimethylsilylation. The derivatization reaction was performed for 30 min at 40 °C. GC–MS analysis was performed on a Shimadzu GC–MS QP2010 Ultra equipped with an AOC20i autoinjector and Rtx-5MS column (30 m, 0.25 mm, 0.25 µm df) in the 70 eV electron ionization mode. The oven temperature program was as follows: 150 °C for 1 min, 20 °C/minute to 250 °C, 5 °C/minute to 280 °C, hold for 5 min, then 20 °C/minute to 330 °C, hold for 3 min, after which the temperature was maintained for 10 min. Helium was the carrier gas with a constant flow speed of 42.0 cm/sec. One microliter was injected at a 5:1 split ratio with an injector temperature of 250 °C. The MS interface temperature was held at 280 °C. The target ion (m/z), retention time, limit of detection, and limit of quantification for the 19 FFAs evaluated are summarized in Table 2 . The expression levels of the fatty acid metabolizing enzymes ACACA, FASN, SCD1, ELOVL1, ELOVL3, and ELOVL6 in cancer and non-cancer patients were analyzed using the Gene Expression Profiling Interactive Analysis (GEPIA) database ( http://gepia.cancer-pku.cn ). Tumor tissues were obtained from patients with cervical squamous cell carcinoma and endocervical adenocarcinoma. GEPIA is a web-based tool that uses data from The Cancer Genome Atlas and Genotype-Tissue Expression projects. Unsupervised clustering and heatmap generation with sorted datasets were performed using MetaboAnalyst 5.0. Comparisons between two groups were assessed using unpaired or paired (for matched comparisons) two-tailed Student’s t-tests or non-parametric Mann–Whitney U-tests. Multiple comparisons were assessed by one-way analysis of variance (ANOVA), including Tukey’s or Bonferroni’s multiple comparisons tests. The diagnostic sensitivity, specificity, and AUC values were calculated for each FFA. To construct an optimal diagnostic model for CC and CIN2/3, samples were randomly divided into discovery and validation sets at a 2:1 ratio stratified by malignancy status (control/CC + CIN2/3). The discovery set was used to select lipid markers and construct a model, then the validation set was used to validate the model. Using the discovery set, multivariable analyses were conducted using the logistic regression model. Variables were first selected from univariate screening by comparing those of healthy controls and CC + CIN2/3, which yielded 11 FFAs with significant differences, followed by a combination of forward-selection and backward-elimination procedures. Correlations between serum concentrations of these lipid markers were calculated using Spearman’s rank correlation method to assess for multicollinearity. The variable selection procedure was set to a threshold of 0.05 for inclusion and 0.05 for exclusion. The Akaike Information Criterion was applied to determine the best model. The DI score was calculated using the intercept and coefficients rounded to one decimal place. Finally, the performance of the DI was assessed with the validation dataset. For comparison, a DI with SCC-Ag values was also calculated using the discovery dataset. For each DI, the diagnostic sensitivity, specificity, and AUC values were calculated. The Jump Pro 13.2.1 statistical software package (SAS Institute Inc, Cary, NC, USA), Prism software (version 9.0; GraphPad Software, La Jolla, CA, USA), and MetaboAnalyst 5.0 were used for data management and statistical analyses. Significance levels were set at P  < 0.05 for all tests. Data are presented as the mean ± standard deviation.

We have previously reported that the FFA composition is altered in lung and ovarian cancer patient serum samples from the aberrant expression of fatty acid metabolic enzymes, particularly stearoyl-CoA desaturase 1 (SCD1), in tumor tissues 22 , 23 . Analysis using public databases revealed that the expression patterns of fatty acid synthases (ACACA, FASN), fatty acid desaturase (SCD1), and fatty acid elongase (ELOVL) 1, 3, and 6 are also upregulated in CC tissues compared with the patterns of those in normal cervical tissues (Figure S1). Therefore, we first comprehensively measured 19 FFAs in serum samples from healthy donors, CIN2/3 patients, and stage I–IV CC patients using gas chromatography-mass spectrometry (GC–MS) (Tables 1 and 2 ). The results showed that the serum FFA composition was dramatically altered in CC patients compared with that in healthy donors. Furthermore, the serum FFA composition changes occurred from stage I (Fig.  1 A and Figure S2A). The serum FFA composition of CIN2/3 patients also differed from that of healthy donors, although not to the same extent as CC patients (Fig.  1 A and Figure S2B). Specifically, the levels of 10 FFAs (sapienic, palmitoleic, linoleic, α-linolenic, dihomo-γ-linolenic, oleic, arachidonic, stearic, arachidic, and palmitic acids) were significantly higher, while those of docosapentaenoic acid were significantly lower, in CC samples compared with the findings in healthy donor samples (Fig.  1 B). Furthermore, FFA levels of 8 of the 11 species significantly altered in CC were significantly elevated from CIN2/3 (Fig.  1 B). On the other hand, there were eight FFAs that remained unchanged (Figure S3). These results strongly suggest that the serum FFA compositions in CC and CIN2/3 patients significantly differ from that of healthy donors. Table 1 Characteristics of participants whose serum free fatty acid concentrations were measured in this study. Characteristics healthy donors (n = 40) CIN patients (n = 30) CC patients (n = 50) Age  Median (range) 43 (21–62) 37 (28–71) 46 (29–73) BMI  Median (range) 21.2 (17.2–26.8) 21.0 (14.8–32.8) Tumor size, mm  Median (range) 45 (4–160) Histopathological subtypes, n (%)  Squamous cell carcinoma 36 (72.0%)  Adenocarcinoma 14 (28.0%) Pathological stage, n (%)  CIN2 3 (10%)  CIN3 27 (90%)  I 27 (54.0%)  II 2 (4.0%)  III 20 (40.0%)  IV 1 (2.0%) CIN Cervical intraepithelial neoplasia, CC Cervical cancer. Table 2 The gas chromatography-mass spectrometry (GC/MS) analyses of serum free fatty acids. Compound Target ion m/z Rt a (min) LOD b (ng/20μL) LOQ c (ng/20μL) Sapienic acid 311.2 5.720 0.13 0.32 Palmitoleic acid 311.2 5.750 0.13 0.32 Palmitic acid 313.2 5.833 0.09 0.22 γ-linolenic acid 335.2 6.569 1.89 4.96 Stearidonic acid 333.2 6.611 0.75 1.75 Linoleic acid 337.2 6.652 0.04 0.08 Oleic acid 339.2 6.669 0.09 0.24 α-linolenic acid 335.2 6.698 0.19 0.48 Vaccenic acid 339.2 6.703 0.09 0.24 Stearic acid 341.2 6.790 0.08 0.21 Arachidonic acid 361.2 7.540 2.58 6.58 Eicosapentaenoic acid 359.2 7.595 9.65 21.31 Dihimo-γ-linolenic acid 363.2 7.657 0.15 0.30 Arachidic acid 369.2 7.917 0.12 0.31 Docosahexaenoic acid 385.25 8.798 0.91 2.25 Adrenic acid 389.25 8.842 4.35 8.93 Docosapentaenoic acid 387.25 8.915 0.43 1.02 Nervonic acid 423.3 10.745 0.03 0.07 Lignoceric acid 425.3 10.938 0.01 0.02 Margaric acid 327.2 6.293 I.S I.S a Retention time. b Limit of detection. c Limit of quantification. Fig. 1 The serum free fatty acid (FFA) compositions of cervical cancer (CC) and cervical intraepithelial neoplasia (CIN)2/3 patients differ from that of healthy donors (HD). The levels of 19 FFAs (shown in Table 2 ) were measured by gas chromatography-mass spectrometry (GC–MS) in serum samples from HD (n = 40), CIN2/3 patients (n = 30), stage I CC patients (n = 27), and stage II/III/IV CC patients (n = 23). ( A ) Heatmap of the average serum levels of the FFAs. ( B ) Comparison of the serum FFA levels between sample types. The FFAs that were significantly altered in CC and CIN2/3 patients compared with those in healthy donors are shown. Statistical analysis was performed by one-way ANOVA with Bonferroni's multiple comparison test. P -values were determined by comparing the healthy donors with each subject group. * P  < 0.05, ** P  < 0.01, *** P  < 0.001, **** P  < 0.0001. CC Pt, cervical cancer patient. Characteristics of participants whose serum free fatty acid concentrations were measured in this study. CIN Cervical intraepithelial neoplasia, CC Cervical cancer. The gas chromatography-mass spectrometry (GC/MS) analyses of serum free fatty acids. a Retention time. b Limit of detection. c Limit of quantification. The serum free fatty acid (FFA) compositions of cervical cancer (CC) and cervical intraepithelial neoplasia (CIN)2/3 patients differ from that of healthy donors (HD). The levels of 19 FFAs (shown in Table 2 ) were measured by gas chromatography-mass spectrometry (GC–MS) in serum samples from HD (n = 40), CIN2/3 patients (n = 30), stage I CC patients (n = 27), and stage II/III/IV CC patients (n = 23). ( A ) Heatmap of the average serum levels of the FFAs. ( B ) Comparison of the serum FFA levels between sample types. The FFAs that were significantly altered in CC and CIN2/3 patients compared with those in healthy donors are shown. Statistical analysis was performed by one-way ANOVA with Bonferroni's multiple comparison test. P -values were determined by comparing the healthy donors with each subject group. * P  < 0.05, ** P  < 0.01, *** P  < 0.001, **** P  < 0.0001. CC Pt, cervical cancer patient. After identifying 11 FFAs that were altered in CC patient serum samples, we next investigated if these FFAs are useful for diagnosing this disease. Receiver operating characteristic (ROC) curve analysis of healthy donors and CC patients showed that all fatty acids had high diagnostic accuracy (area under the curve (AUC) values > 0.7) (Fig.  2 A and B), especially nine FFAs that had very high diagnostic abilities (AUC > 0.8) (Fig.  2 A). These results suggest that 11 serum FFAs may be useful as diagnostic markers for CC. Interestingly, similar results were obtained in the ROC analysis for healthy donors and only stage I CC patients (Figure S4A and B), suggesting that these serum FFAs vary in a clinical stage-independent manner. Thus, the early diagnosis of CC may also be possible using serum FFAs. Fig. 2 Serum free fatty acid (FFA) levels are useful as diagnostic markers for stage I–IV cervical cancer (CC). Receiver operating characteristic (ROC) curves for detecting stage I–IV CC patients using serum FFAs in Fig.  1 B. ( A ) Area under the ROC curve (AUC) values of “AUC > 0.8” for the FFAs. ( B ) AUC values of “0.8 > AUC > 0.7” for the FFAs. Serum free fatty acid (FFA) levels are useful as diagnostic markers for stage I–IV cervical cancer (CC). Receiver operating characteristic (ROC) curves for detecting stage I–IV CC patients using serum FFAs in Fig.  1 B. ( A ) Area under the ROC curve (AUC) values of “AUC > 0.8” for the FFAs. ( B ) AUC values of “0.8 > AUC > 0.7” for the FFAs. Because it is important to detect CIN, a precancerous lesion, to improve CC patient prognosis, we next examined if CIN2/3 could be diagnosed by examining serum FFA levels. SCC-Ag levels, a squamous cell carcinoma tumor marker, are known to not vary in patients with CIN 24 . Data from our cohort confirmed that SCC-Ag levels are not useful for diagnosing CIN2/3 (AUC = 0.5725) (Figure S5). However, nine serum FFAs displayed high diagnostic ability for CIN2/3 (AUC > 0.7), especially palmitic, palmitoleic, and dihomo-γ-linolenic acids (AUC > 0.8) (Fig.  3 A–C). These results suggest that these serum FFAs can be used to diagnose not only CC, but also CIN2/3. Fig. 3 Serum free fatty acid (FFA) levels are useful as diagnostic markers for cervical intraepithelial neoplasia (CIN)2/3. Receiver operating characteristic (ROC) curves for detecting CIN2/3 patients using serum FFAs in Fig.  1 B. ( A ) Area under the ROC curve (AUC) values of “AUC > 0.8” for the FFAs. ( B ) AUC values of “0.8 > AUC > 0.7” for the FFAs. (C) AUC values of “AUC < 0.7” for the FFAs. Serum free fatty acid (FFA) levels are useful as diagnostic markers for cervical intraepithelial neoplasia (CIN)2/3. Receiver operating characteristic (ROC) curves for detecting CIN2/3 patients using serum FFAs in Fig.  1 B. ( A ) Area under the ROC curve (AUC) values of “AUC > 0.8” for the FFAs. ( B ) AUC values of “0.8 > AUC > 0.7” for the FFAs. (C) AUC values of “AUC < 0.7” for the FFAs. Because we identified serum FFAs that are useful for diagnosing both CC and CIN2/3, we investigated if examining certain fatty acids in combination could improve the diagnostic accuracy. We aimed to establish a diagnostic model that could distinguish healthy donors from CIN2/3 and stage I–IV CC patients for the use of serum FFAs as a broad screening tool. First, using the discovery dataset (healthy donor: n = 24, CC and CIN2/3: n = 56), a multivariable logistic regression model was performed using the eleven FFA markers. With a combination of forward-selection and backward-elimination procedures, the combination of stearic acid and dihomo-γ-linolenic acid was identified to be statistically effective for distinguishing CIN2/3 and stage I–IV CC patients. ROC analysis was performed using these two markers, with the results showing that the AUC value was 0.9427 (diagnostic index (DI) = –5.631 + 0.030 × (Stearic acid) + 9.188 × (Dihomo-γ-linolenic acid)) (Fig.  4 A and B). The diagnostic performance of this model was assessed using the validation dataset (healthy donor: n = 16, CC and CIN2/3: n = 24), which revealed that the model was very accurate, with an AUC value of 0.9531 (Fig.  4 C). These results suggest the importance of combining serum FFA markers to improve the diagnostic accuracy of this approach. Furthermore, the DI using SCC-Ag values showed an AUC of 0.6165, indicating that the DI (Stearic acid, Dihomo-γ-linolenic acid) demonstrated high diagnostic ability (Fig.  4 D and E). Fig. 4 Development of a cervical cancer (CC) and cervical intraepithelial neoplasia (CIN)2/3 detection model using serum free fatty acid (FFA) levels. ( A–C ) Diagnostic performance of the diagnostic index (DI) using stearic acid and dihomo-γ-linolenic acid (DI (Stearic acid, Dihomo-γ-linolenic acid)). ( A ) The DI (Stearic acid, Dihomo-γ-linolenic acid) established by multivariate logistic regression. ( B, C ) Receiver operating characteristic (ROC) curve analysis of healthy donors and CC + CIN2/3 patients in the ( B ) discovery (healthy donor: n = 24, CC and CIN2/3: n = 56) and ( C ) validation data sets (healthy donor: n = 16, CC and CIN2/3: n = 24). ( D, E ) Diagnostic performance of the DI using squamous cell carcinoma antigen (DI (SCC-Ag)). ( D ) The DI (SCC-Ag) established by multivariate logistic regression. ( E ) ROC curve analysis of healthy donors and CC + CIN2/3 patients. ( F, G ) The cut-off values in the ( F ) DI (Stearic acid, Dihomo-γ-linolenic acid) and ( G ) DI (SCC-Ag) were set, then the true negative/positive rates were calculated. The cut-off value was set by calculating Youden’s index by ROC analysis. SCC, squamous cell carcinoma; adeno, adenocarcinoma; CC Pt, cervical cancer patient. Development of a cervical cancer (CC) and cervical intraepithelial neoplasia (CIN)2/3 detection model using serum free fatty acid (FFA) levels. ( A–C ) Diagnostic performance of the diagnostic index (DI) using stearic acid and dihomo-γ-linolenic acid (DI (Stearic acid, Dihomo-γ-linolenic acid)). ( A ) The DI (Stearic acid, Dihomo-γ-linolenic acid) established by multivariate logistic regression. ( B, C ) Receiver operating characteristic (ROC) curve analysis of healthy donors and CC + CIN2/3 patients in the ( B ) discovery (healthy donor: n = 24, CC and CIN2/3: n = 56) and ( C ) validation data sets (healthy donor: n = 16, CC and CIN2/3: n = 24). ( D, E ) Diagnostic performance of the DI using squamous cell carcinoma antigen (DI (SCC-Ag)). ( D ) The DI (SCC-Ag) established by multivariate logistic regression. ( E ) ROC curve analysis of healthy donors and CC + CIN2/3 patients. ( F, G ) The cut-off values in the ( F ) DI (Stearic acid, Dihomo-γ-linolenic acid) and ( G ) DI (SCC-Ag) were set, then the true negative/positive rates were calculated. The cut-off value was set by calculating Youden’s index by ROC analysis. SCC, squamous cell carcinoma; adeno, adenocarcinoma; CC Pt, cervical cancer patient. We subsequently assessed the properties of the DI (Stearic acid, Dihomo-γ-linolenic acid) system. The results showed that the true positive rates of the DI (Stearic acid, Dihomo-γ-linolenic acid) in stage I CC and CIN2/3 patients were 96.3% and 70.0%, respectively (Fig.  4 F). The true positive rates for the DI (SCC-Ag) in stage I CC and CIN2/3 patients were 52.2% and 16.7%, respectively (Fig.  4 G), indicating that the DI (Stearic acid, Dihomo-γ-linolenic acid) has superior early diagnostic ability. Furthermore, for histological type, the DI (Stearic acid, Dihomo-γ-linolenic acid) was highly sensitive (true positive rate: 100.0%) in detecting adenocarcinoma, which is not detectable using the DI (SCC-Ag) (true positive rate: 0.0%) (Fig.  4 F and G). These results strongly suggest that the DI (Stearic acid, Dihomo-γ-linolenic acid) can be used to diagnose CC and CIN2/3 independently of clinical stage and histological type.

The prevention and treatment of CC is a global public health issue 25 , 26 . The five-year survival rate of CC detected at stage I is 81% to 96%, but the survival rate drops dramatically after the cancer cells invade surrounding the tissues or metastasize 8 . Therefore, it is critical to identify new diagnostic markers for the early diagnosis of CC. In this study, we demonstrated that the sapienic, palmitoleic, linoleic, α-linolenic, dihomo-γ-linolenic, oleic, arachidonic, stearic, arachidic, palmitic, and docosapentaenoic acid levels were dramatically altered in CC patient serum samples compared with the levels in samples from healthy donors. Furthermore, some of these FFAs also displayed differential levels in CIN2/3 patient serum samples, suggesting that both CC and CIN2/3 can be diagnosed using specific serum FFAs. Finally, we developed an optimal diagnostic model for the combination of FFAs using statistical methods, finding that the DI using stearic acid and dihomo-γ-linolenic acid levels had higher sensitivity and specificity (discovery set: sensitivity 0.89, specificity 0.96; validation set: sensitivity 0.88, specificity 0.89). Furthermore, the DI (Stearic acid, Dihomo-γ-linolenic acid) was highly sensitive (true positive rate: 100.0%) in detecting adenocarcinoma. Thus, the DI (Stearic acid, Dihomo-γ-linolenic acid) can be used to detect CC in a clinical stage- and histological type-independent manner. We previously reported that palmitoleic, oleic, linoleic, α-linolenic, vaccenic, arachidic, docosahexaenoic, and lignoceric acid levels were altered in ovarian cancer patients 23 . In this study, palmitoleic, oleic, linoleic, α-linolenic, and arachidic acids were found to also be altered in CC patient serum samples, suggesting that these FFAs may display differential patterns in a relatively wide range of cancer types. In addition, these five FFAs would be suitable for very early disease screening to distinguish between cancer and non-cancer without determining the cancer type because their concentrations in serum are high and easily detected by GC–MS. Because the other six FFAs (sapienic, dihomo-γ-linolenic, docosapentaenoic, arachidonic, stearic, and palmitic acids) were altered in CC but not in ovarian cancer 23 , they may be CC-specific diagnostic markers. In this study, a DI using stearic acid and dihomo-γ-linolenic acid was constructed and determined to be the optimal diagnostic model for CC. In ovarian cancer, a DI using oleic and arachidic acid has been constructed 23 . The identification of different FFAs in cervical and ovarian cancers is very interesting because it suggests the possibility of a cancer type-specific diagnostic marker. The sensitive detection of CIN2/3 is most important in repeat non-responders to CC screening, a subgroup known to be at high-risk of developing CC 27 . In this study, we identified several FFAs that could be used to detect CIN2/3 in patients. In addition, true positive rates for DI (SCC-Ag) in CIN2/3 were 16.7%, while the DI (Stearic acid, Dihomo-γ-linolenic acid) constructed in this study detected not only CC but also CIN2/3 with high sensitivity (true positive rate: 70.0%). These findings suggest that serum FFA screening is a promising alternative for detecting CIN2/3 and could significantly improve the current CC prevention strategies. However, the mechanism by which these specific serum FFAs levels are altered in CIN2/3 is unknown. We previously reported that serum FFA levels are altered from changes in fatty acid metabolizing enzyme expression patterns, particularly of SCD1, in cancer tissues 23 . High SCD1 expression levels in tumor tissues and elevated serum concentrations of palmitoleic acid and oleic acid, SCD1-generated fatty acids, were observed in CC, consistent with the previous report 22 , 23 . However, in CIN2/3, changes in fatty acid metabolic properties in a very limited range of lesions are unlikely to be reflected in systemic changes in fatty acid metabolism, suggesting that other mechanisms are involved. Liquid factors, such as cancer cell-derived miRNAs and long non-coding RNAs (lncRNAs), can reportedly regulate the activity of fatty acid metabolizing enzymes, such as SCD1 28 , 29 . Cancer cells potentially release miRNAs and lncRNAs associated with fatty acid metabolic reprogramming to control systemic fatty acid metabolism. First, diagnostic accuracy should be further examined in a larger cohort because the cohort used in this study has a very small sample size. In addition, comparisons with benign gynecological diseases　(e.g. endometriosis and uterine fibroids) and other HPV-related cancers (e.g. HPV + oropharyngeal, anal, penile, vulvar and vaginal cancers) must be performed to verify the specificity of this diagnostic tool. Second, all the healthy donors and patients enrolled in this study were Japanese, and the study has not been validated in a cohort of low-and middle-income countries (LMIC), which have high CC morbidity and mortality rates. Clinical studies with LMIC cohorts should be conducted because of the different dietary cultures that may affect serum FFA levels in LMIC. Nevertheless, the significance of this study is noteworthy because serum FFAs screening is highly sensitive in detecting stage I CC, cervical adenocarcinoma, and CIN2/3, which are difficult to detect by SCC-Ag screening. Although CC screening programs are widely implemented, they have not been satisfactorily effective in LMIC because of a lack of pathologists and medical equipment to perform cytology for cancer screening. Therefore, 90% of CC-related deaths worldwide are in LMIC 26 , 30 , 31 . Blood sampling is commonly performed in clinical practice and is considered to be a convenient and non-invasive method. Furthermore, the serum FFAs test can be performed using any commercially available GC–MS with basic specifications. In addition, the patients can simply send the dried blood spots via regular mail to the laboratory for analysis 32 . Therefore, our serum FFA diagnostic model may be a new cancer screening tool that can be implemented in LMIC, potentially contributing to significantly increasing the number of cervical screening program recipients and improving CC patient prognosis.

Cervical cancer (CC) develops primarily through precancerous lesions, called cervical intraepithelial neoplasia (CIN), and is caused by persistent infection with high-risk human papillomavirus 1 . Despite the widespread implementation of screening programs, CC still has high incidence and mortality rates, especially in developing countries 2 – 6 . Low participation in these screening programs has led to their limited effectiveness, mainly because early-stage CC often has no clinical symptoms 6 . CC has a good prognosis when detected and treated at stage I, with a five-year survival rate of 81% to 96% 7 , 8 . Furthermore, CIN progression to CC can be prevented if it is detected and managed appropriately 9 . Conversely, the CC five-year survival rates for stages II, III, and IV are approximately 65% to 87%, 35% to 50%, and 15% to 20%, respectively, with a poor prognosis 7 , 8 , 10 . Therefore, a non-invasive and convenient diagnostic method, such as liquid biopsy, is needed to help identify CC in individuals not participating in the current screening programs. Squamous cell carcinoma antigen (SCC-Ag) is widely accepted in current clinical practice as a tumor marker for cervical squamous cell carcinoma. However, SCC-Ag is not a sufficient indicator of early-stage CC or CIN 8 . Additionally, there is no specific tumor marker for cervical adenocarcinoma. Molecular markers, such as Keratin 4/17 11 , 12 , Annexin A2 13 , and PTEN 14 , as well as certain microRNAs (miRNAs) 15 , 16 and urinary DNA methylation 17 , have been reported as candidate diagnostic markers for CC. However, these markers have not yet been used in clinical practice, nor has any validation study been conducted to demonstrate their ability to decrease CC patient morbidity and mortality rates. Therefore, there is an urgent need to develop new non-invasive, convenient, and highly sensitive early diagnostic technologies to improve the prognosis of CC patients. Cancer cells can reprogram their metabolic pathways to favor their own survival. Many studies have examined cancer cell-specific glucose and amino acid metabolism, which have been applied to therapeutics and diagnostics 18 – 21 . We have previously demonstrated that lung and ovarian cancers can also reprogram their fatty acid metabolic systems 22 , 23 . Furthermore, we reported that the composition of serum free fatty acids (FFAs) in ovarian cancer is significantly altered from early-stage because of this reprogramming 23 . Interestingly, the early diagnosis of ovarian cancer was found to be possible by examining specific serum FFAs 23 . Fatty acid metabolic reprogramming occurs in a wide range of cancer types and may expand the indications for cancer diagnosis using serum FFAs. However, no studies have examined the utility of serum FFAs for diagnosing CC and CIN. In this study, we identified several serum FFAs that display dramatically altered levels in patients with CC or CIN. Furthermore, we combined these FFAs to develop a highly sensitive and specific diagnostic model.

The authors declare that all other data supporting the findings of this study are available within the article and its supplementary information files, and from the corresponding author upon reasonable request.

Supplementary Information. Supplementary Information.