OnabotulinumtoxinA for chronic migraine: A real-life multicenter study of 479 patients.

Future research should focus on identifying neurobiological and imaging biomarkers, which may provide higher predictive accuracy than clinical data. Furthermore, our findings regarding the negative impact of multiple prior treatment failures call for a re-evaluation of current reimbursement criteria.

Chronic migraine (CM) is a complex, and difficult-to-treat neurological disorder affecting approximately 2–3% of the general population. 1 , 2 It is defined as the occurrence of headache on at least 15 days per month, for at least three months, where at least 8 days meet the criteria for migraine without aura or respond to migraine-specific treatment. 3 Furthermore, the overuse of acute medications, which often leads to the development of medication-overuse headache (MOH), significantly complicates clinical management and contributes to substantial disability. 4 , 5 Despite the expanding therapeutic landscape, a significant proportion of CM patients show only a partial response or remain refractory to preventive treatment. 6 , 7 Onabotulinumtoxin A (BoNT-A) is an established preventive therapy for CM; its safety and efficacy were demonstrated in the Phase III Research Evaluating Migraine Prophylaxis Therapy (PREEMPT) clinical trials 8 – 10 and subsequently confirmed in numerous real-world studies. 11 – 14 Mechanistically, BoNT-A exerts its therapeutic effect by inhibiting the release of pro-inflammatory and nociceptive neurotransmitters, most notably calcitonin gene-related peptide (CGRP), substance P, and glutamate, from peripheral trigeminal sensory nerve endings. By dampening this peripheral nociceptive input, BoNT-A indirectly reduces the subsequent development and maintenance of central sensitization, which is a core pathophysiological driver of migraine chronification. 1 , 2 , 8 However, some reports have indicated only modest efficacy for BoNT-A in certain populations. 15 , 16 As with other preventive strategies, substantial inter-individual variability in treatment response is observed in clinical practice. Therefore, assessing treatment outcomes and identifying potential prognostic factors is essential for a reliable evaluation of therapeutic benefits. This is particularly relevant given the recent emergence of novel CM therapies, such as anti-CGRP therapies. 17 , 18 The aim of the present study was to evaluate the effectiveness of BoNT-A in a cohort of Polish patients enrolled in a national reimbursed chronic migraine treatment program and to identify potential clinical predictors of treatment response.

The study was a retrospective observational analysis of 479 patients treated with BoNT-A as part of the reimbursement program for the treatment of CM in Poland, enrolled between 2022 and 2024. A total of 12 tertiary headache centers participated in the study. All patients were diagnosed with CM according to the third edition of the International Classification of Headache Disorders (ICHD-3) 3 and were eligible for treatment according to the regulatory principles of the reimbursement program in Poland. A prerequisite for receiving 3-5 cycles of BoNT-A treatment was prior ineffective or intolerable treatment with at least two of the following three oral preventive medications: amitriptyline, topiramate, and valproic acid. If BoNT-A treatment proves ineffective or intolerable, patients become eligible to receive monoclonal antibodies. 19 Secondary headache disorders were ruled out during the screening process in accordance with clinical guidelines, with the vast majority of patients having undergone unremarkable neuroimaging prior to enrollment. Comorbidities were identified based on clinical documentation and confirmed specialist diagnoses. Patients were administered 195 units of BoNT-A (39 injection sites) in accordance with the PREEMPT protocol at 3-month intervals. Efficacy was evaluated after 9 months, corresponding to three cycles of treatment based on standardized headache diaries. To minimize recall bias, patients prospectively completed daily paper-based or validated electronic headache diaries throughout the 9-month period, recording headache presence, intensity, characteristics, and acute medication use. Data collection was harmonized across all 12 participating centers using a standardized diary form. Each center reported the following demographic and clinical data: age, gender, reproductive status, presence of aura, allodynia, nausea, vomiting, photophobia, and phonophobia; age at onset and chronification of migraine; comorbidities; and data on current acute and preventive medications. Treatment outcomes included monthly headache days (MHD), monthly migraine days (MMD), monthly acute medication days (AMD), and the Migraine Disability Assessment (MIDAS) score. Treatment efficacy was defined as the mean reduction in monthly headache days (MHD) assessed after 9 months of treatment, based on data derived from headache diaries completed by patients. The primary effectiveness endpoint was defined as a ≥50% reduction in monthly headache days (MHD) assessed after 9 months of treatment (responders). Patients with a 25–49% reduction were categorized as partial responders, and those with a reduction of less than 25% were labeled as non-responders. Additionally, to provide deeper clinical context, the proportions of patients achieving stricter response thresholds, specifically ≥75% and 100% (complete remission) reductions in MHD, were evaluated as exploratory secondary outcomes. Approval for this study was obtained from the Bioethics Committee of Wroclaw Medical University (approval number: KB-40/2025; February 27, 2025). All participants provided written informed consent prior to inclusion. All procedures performed in this study were conducted in accordance with the ethical standards of the Declaration of Helsinki. A univariate analysis was performed using STATISTICA (StatSoft Inc., Tulsa, USA; version 13.0). The significance level was established as α = 0.05. Categorical variables were presented as numbers and percentages, and continuous ones as the mean and standard deviation (SD) or as the median and interquartile range. To detect differences between analyzed groups, Student’s t-test was used for normally distributed data, and the Mann-Whitney U test was used for other variables. The Chi-squared test for independence was performed to establish the significance of differences in variables measured at a nominal level. In order to establish predictors associated with the response to BoNT-A, we first performed univariable logistic regression analyses for each candidate variable. Variables that showed potential significance in univariate analyses (identified by p<0.1) or considered clinically relevant (age, sex) were subsequently included in a multivariable model. The multivariable model was developed using a generalized linear model (GLZ) with a binomial distribution and logit link function. Parameter estimation was performed by maximum likelihood estimation (MLE). Odds ratios (ORs) with corresponding 95% confidence intervals (95% CIs) were calculated for each predictor. Model performance was assessed by examining model fit indices (Akaike Information Criterion, AIC), discrimination (area under the receiver operating characteristic curve, AUC), and calibration (Hosmer-Lemeshow goodness-of-fit test). To evaluate the internal validity and robustness of the model, a 10-fold stratified cross-validation was performed, and the average AUC across folds was reported. Potential influential outliers were assessed using deviance and Pearson residual plots.

A total of 497 patients were initially screened, of whom 479 completed the full 9-month therapeutic protocol. Eighteen patients (3.6%) dropped out during the study period. Importantly, no patient discontinued the therapy due to adverse effects or tolerability issues. The specific reasons for discontinuation were a perceived lack of clinical efficacy after 1–2 cycles in 13 patients (72.2%) and logistic or administrative protocol non-compliance in 5 patients (27.8%). The study group consisted of 53 men (11.1%) and 426 women (88.9%). The mean age was 43.6±11.6 years (range 18-80), while the mean age of migraine onset was 19.7±8.9 years (range 5-49). A total of 142 (29.6%) patients had had at least four failed attempts of migraine prophylaxis in the past. The basic demographic and clinical data of the study group are presented in Table 1 . Table 1. The basic demographic and clinical data of the study group. Characteristics Total (n=479) Sex: Women 426 (88.9%) Men 53 (11.1%) Reproductive status (women) Active 232 (54.5%) Perimenopause 87 (20.4%) Menopause 107 (25.1%) Age (mean, SD, range) 43.6±11.6 (range 18-80) Center Wrocław - 87 (18.2%) Gdańsk A - 78 (16.3%) Gryfice - 67 (14%) Poznań - 45 (9.4%) Gdańsk B - 35 (7.3%) Zabrze - 34 (7.1%) Bydgoszcz - 31 (6.5%) Szczecin - 28 (5.8%) Lublin - 23 (4.8%) Rzeszów - 20 (4.2%) Opole - 17 (3.5%) Proszowice - 14 (2.9%) Type of migraine : MwA 138 (28.8%) MwoA 341 (71.2%) Allodynia 204 (42.6%) Headache characteristics : Unilateral 393 (82%) Nausea/vomiting 391 (81.6%) Photophobia/Phonophobia 459 (95.8%) Age of onset 19.7±8.9 (range 5-49) Age of chronification 35.5±11 (range 10-68) Comorbidities: ​ Anxiety disorder 96 (20%) Depressive disorder 91 (19%) Other psychiatric disorders 6 (1.2%) Arterial hypertension 61 (12.7%) Hypothyroidism 59 (12.3%) Graves’ disease 2 (0.4%) Obesity 27 (5.6%) Cardiovascular diseases 25 (5.2%) Epilepsy 4 (0.8%) History of AIS/TIA 1 (0.2%) Multiple sclerosis 3 (0.6%) SLE 2 (0.4%) MCTD 3 (0.6%) RA 4 (0.8%) Fibromyalgia 2 (0.4%) Sarcoidosis 2 (0.4%) Diabetes 2 (0.4%) Endometriosis 5 (1%) PCOS 2 (0.4%) SIBO 4 (0.8%) IBS 2 (0.4%) Asthma 8 (1.7%) Family history of migraine 243 (50.7%) Prior drug ineffectiveness: Topiramate 399 (83.3%) Amitryptyline 327 (68.3%) Valproic acid 145 (30.3%) Beta-blockers 119 (24.8%) Candesartan 35 (7.3%) Flunarizine 142 (29.6%) Iprazochrome 168 (35.1%) Abbreviations: SD-standard deviation, MwA - migraine with aura, MwoA - migraine without aura, AIS -acute ischemic stroke, TIA-transient ischemic attack, SLE - systemic lupus erythomatosus, MCTD - mixed connective tissue disease, RA - reumathoid arthritis, PCOS - polycystic ovary syndrome, SIBO - small intestinal bacterial overgrowth, IBS - irritable bowel syndrome. The basic demographic and clinical data of the study group. Abbreviations: SD-standard deviation, MwA - migraine with aura, MwoA - migraine without aura, AIS -acute ischemic stroke, TIA-transient ischemic attack, SLE - systemic lupus erythomatosus, MCTD - mixed connective tissue disease, RA - reumathoid arthritis, PCOS - polycystic ovary syndrome, SIBO - small intestinal bacterial overgrowth, IBS - irritable bowel syndrome. At baseline (M0), the mean MHD in the study group was 19.1 ± 4.4, the mean MMD was 15.1 ± 5.2, the mean AMD was 15.9 ± 5.5, and the mean baseline MIDAS score was 107.2 ± 57.2. When evaluated using non-parametric distributions to match longitudinal analyses, the median (IQR) values at baseline were 18 (16–21) days for MHD, 15 (11–18) days for MMD, 16 (12–18) days for AMD, and 99 (63–136) points for the MIDAS score, respectively ( Table 2 ). Among acute medication types, the mean triptan use was 9.8 ± 7.5 days, and the mean NSAID use was 7.4 ± 7.5 days. Table 2. Treatment outcomes at specific timepoints. Values presented as median (interquartile range). ​ M0 M3 p * M6 p * M9 p * Monthly headache days (MHD) 18 (16-21) 15 (11-19) <0.01 14 (10-18) <0.01 14 (8-17) <0.01 Monthly migrane days (MMD) 15 (11-18) 11 (8-16) <0.01 10 (7-15) <0.01 11 (6-15) <0.01 Acute medication days (ADM) 16 (12-18) 12 (8-16) <0.01 11 (7-15) <0.01 11 (6-16) <0.01 MIDAS 99 (63-136) 63 (40-100) <0.01 50 (27-81) <0.01 45 (21-80) <0.01 *Wilcoxon signed-rank test. Bold values indicate statistically significant results. Treatment outcomes at specific timepoints. Values presented as median (interquartile range). *Wilcoxon signed-rank test. Bold values indicate statistically significant results. At 3 months (M3), 64 patients (13.4%) achieved a ≥50% reduction in MHDs, while 131 (27.3%) were identified as partial responders (25-49% reduction). At 6 (M6) and 9 (M9), the proportions of responders and partial responders were 20.7% and 31.3%, and 27.3% and 23.2%, respectively ( Figure 1 ). Figure 1. Distribution of responders, partial responders, and non-responders across the treatment cycles. Distribution of responders, partial responders, and non-responders across the treatment cycles. In terms of mean improvements at 9 months, the mean reduction in MHD was 5.9 ± 5.4 days, in MMD 4.7 ± 4.8 days, and in MIDAS 52.7 ± 48.3 points. Correspondingly, the median (IQR) values at the 9-month follow-up significantly decreased to 14 (8–17) days for MHD, 11 (6–15) days for MMD, 11 (6–16) days for AMD, and 45 (21–80) points for the MIDAS score (p < 0.01 for all comparisons; Table 2 ), ( Figure 2 ). Figure 2. Median number of monthly headache days (MHD), monthly migraine days (MMD) and acute medication days (AMD) before treatment and across treatment cycles. Abbreviations: MHD, monthly headache days; MMD, monthly migraine days; ADM, acute medication days. Median number of monthly headache days (MHD), monthly migraine days (MMD) and acute medication days (AMD) before treatment and across treatment cycles. Abbreviations: MHD, monthly headache days; MMD, monthly migraine days; ADM, acute medication days. At enrollment, 339 (70.8%) individuals were diagnosed with concomitant MOH. Of these, 141 (41.6%) no longer met the criteria for MOH after 9 months of treatment. In the comparison between responders and the remaining study group, it was found that individuals who achieved at least a 50% improvement had fewer days of acute medication at M0 (median 15 vs. 16, p=0.026), were less likely to have a positive family history of migraine (39.2% vs. 55.3%, p=0.002) and were less likely to have had at least four preventive medication failures in the past (21.4% vs. 32.8%, p=0.015). The comparisons between groups are summarized in Table 3 . Table 3. Comparisons between responders and the remaining study group. ​ Responders (n=131) Non-responders (n=348) p-value Age 43 (33-51) 44 (36-50.5) 0.714 Age of onset 18 (14-24) 18 (14-23) 0.291 Age of chronification 33 (28-41) 36 (27-43) 0.606 MHD M0 18 (16-21) 18 (16-21) 0.913 MMD M0 15 (10-17) 15 (11-18) 0.176 AMD M0 15 (12-18) 16 (13-19) 0.026 Sex distribution Women 120 (91.6%) 306 (87.9%) 0.253 Men 11 (8.4%) 42 (12.1%) ​ Aura 34 (26%) 104 (29.9%) 0.397 Allodynia 61 (47.3%) 143 (41.7%) 0.274 Unilateral pain 108 (83.1%) 285 (83.1%) 0.997 Nausea/vomiting 110 (84%) 281 (80.8%) 0.417 Photo/phonophobia 123 (93.9%) 336 (96.6%) 0.195 Comorbidities Arterial hypertension 18 (13.9%) 43 (12.4%) 0.664 Obesity 10 (7.7%) 17 (4.9%) 0.237 Cardiovascular disease 10 (7.7%) 15 (4.3%) 0.139 Anxiety 27 (20.6%) 69 (19.9%) 0.830 Depression 21 (16.2%) 70 (20.2%) 0.320 Family history of migraine 51 (39.2%) 192 (55.3%) 0.002 At least 4 preventive medication failures 28 (21.4%) 114 (32.8%) 0.015 Bold values indicate statistically significant results. Comparisons between responders and the remaining study group. Bold values indicate statistically significant results. In the univariable analyses, a positive family history of migraine (OR = 0.52, 95% CI 0.35-0.79, p = 0.002), and having at least four preventive medication failures in the past (OR = 0.56, 95% CI 0.35-0.90, p = 0.016) were significantly associated with treatment response. These variables, together with two additional clinically relevant factors (age and sex), were included in the multivariable model. In the final model, a both positive family history of migraine and multiple prophylaxis failure remained independently associated with treatment response (OR = 0.55, 95% CI 0.36-0.84, p = 0.005 and OR = 0.61, 95% CI 0.37-0.98, p=0.042, respectively). The results of the univariable and multivariable analyses are shown in Table 4 . Table 4. Results of univariable and multivariable analyses of predictors of treatment response. Predictor variable Univariable OR (95% CI) p-value Multivariable OR (95% CI) p-value Sex (Male vs Female) 0.67 (0.33-1.34) 0.256 0.73 (0.36-1.49) 0.390 Age (per year) 1.00 (0.98-1.01) 0.665 1.00 (0.98-1.02) 0.865 Age of onset (per year) 1.01 (0.99-1.03) 0.478 ​ ​ Age of chronification (per year) 1.00 (0.98-1.01) 0.671 ​ ​ MHD M0 0.99 (0.95-1.04) 0.792 ​ ​ MMD M0 0.98 (0.94-1.02) 0.372 ​ ​ AMD M0 0.98 (0.95-1.01) 0.204 ​ ​ MIDAS M0 1.00 (0.99-1.00) 0.141 ​ ​ MOH M0 0.87 (0.56-1.35) 0.541 ​ ​ Aura 0.82 (0.52-1.29) 0.398 ​ ​ Allodynia 1.25 (0.84-1.88) 0.275 ​ ​ Unilateral pain 1.00 (0.58-1.71) 0.997 ​ ​ Nausea/vomiting 1.25 (0.73-2.14) 0.418 ​ ​ Photo/phonophobia 0.55 (0.22-1.38) 0.201 ​ ​ Comorbidities Arterial hypertension 1.14 (0.63-2.06) 0.664 ​ ​ Obesity 1.62 (0.72-3.64) 0.241 ​ ​ Cardiovascular disease 1.85 (0.81-4.23) 0.145 ​ ​ Anxiety 1.06 (0.64-1.74) 0.830 ​ ​ Depression 0.76 (0.45-1.30) 0.321 ​ ​ Family history of migraine 0.52 (0.35-0.79) 0.002 0.55 (0.36-0.84) 0.005 At least 4 preventive medication failures 0.56 (0.35-0.90) 0.016 0.61 (0.37-0.98) 0.042 Abbreviations: OR, odds ratio; CI, confidence interval; MHD, monthly headache days; MMD, monthly migraine days; ADM, acute medication days; MOH, medication-overuse headache. Multivariable model fit: AIC = 533.7; AUC = 0.605. Bold values indicate statistically significant results. Results of univariable and multivariable analyses of predictors of treatment response. Abbreviations: OR, odds ratio; CI, confidence interval; MHD, monthly headache days; MMD, monthly migraine days; ADM, acute medication days; MOH, medication-overuse headache. Multivariable model fit: AIC = 533.7; AUC = 0.605. Bold values indicate statistically significant results. The logistic regression model demonstrated an adequate overall fit (AIC=533.7). Model discrimination was poor, with an AUC of 0.605. Internal validation with 10-fold cross-validation confirmed the lack of predictive ability, with a mean AUC of 0.54. Calibration was acceptable, as indicated by the Hosmer-Lemeshow test (p = 0.40).

The present study evaluated the effectiveness of BoNT-A in a large cohort of Polish patients with CM treated within a national reimbursement program. Our findings confirm the real-world efficacy of BoNT-A; 27.3% of patients achieved the primary endpoint of a ≥50% reduction (responders), and 23.2% were partial responders after nine months of treatment. The demographic characteristics of our cohort are largely consistent with the typical profile of CM patients reported in other studies. The mean age of 43.6 years and the strong female predominance (88.9%) closely mirror the populations in the REPOSE (mean age 45.3, 85% female) and COMPEL (mean age 43.1, 84.7% female) studies. 11 , 13 However, a more granular analysis reveals that our patients represent a particularly long-standing and refractory phenotype. The mean age of migraine onset in our group was 19.7 years, which, given the current mean age of 43.6, suggests an average disease duration of approximately 24 years. This is notably longer than the duration reported in several European registries. For instance, in the Italian real-life study by Negro et al., the average history of migraine was approximately 18–20 years. 20 Notably, before the implementation of this drug program, no migraine-specific preventive treatment for chronic migraine was publicly reimbursed in Poland. Furthermore, the fact that nearly 30% of our patients had failed four or more prior preventive medications further distinguishes our cohort from those in earlier real-life studies, such the analysis by Ahmed et al., where the majority of patients had failed only 2–3 preventives. 13 This level of prior treatment failure is significantly higher than in the original PREEMPT trials, suggesting that the Polish cohort enters the advanced treatment program at a later, more clinically advanced stage of the disease. 8 – 10 The long interval between disease onset and the initiation of BoNT-A - often exceeding two decades - likely contributes to the high prevalence of concomitant MOH (70.7%) in our cohort. Compared to the PREEMPT clinical trials, where MOH was present in approximately 60% of participants, our population exhibits a significantly higher burden of medication overuse. 8 – 10 The effectiveness of BoNT-A observed in our Polish cohort is consistent with global real-world evidence, although certain differences in responder rates merit discussion. In our study, approximately 27.3% of patients achieved a ≥50% reduction in MHD within 9 months. This is slightly lower than the responder rates reported in the COMPEL study (∼49% at 108 weeks 39.5% at 24 weeks and 61.1% at 108 weeks). 11 However, our findings are remarkably consistent with other real-world analyses involving highly refractory populations. For instance, in a large prospective study conducted in the UK by Khalil et al., the responder rate at the first evaluation was nearly identical to ours at 32%. 21 Similarly, in the Italian RAMO study, reported a 50% responder rate of 38.8% after the first treatment cycle. 20 These comparisons strengthen the argument that in clinical practice, particularly within stringent reimbursement frameworks, a responder rate of approximately 25-30% represents a realistic benchmark for BoNT-A efficacy in difficult-to-treat CM patients. 22 , 23 The MOH resolution rate of 41.6% in our study represents a significant clinical outcome, comparable to the 54% reported in the Spanish real-world registry by Irimia et al. 24 and the 45% observed in Italian cohorts. 25 Our study identified a positive family history of migraine and a high number of prior preventive treatment failures (≥4) as independent predictors of a poorer response to BoNT-A. These findings are consistent with, and extend, the existing evidence derived from real-world evidence (RWE). The negative impact of multiple prior preventive treatment failures (OR = 0.61) and a positive family history (OR = 0.55) were independently associated with a poorer treatment response in our cohort. In clinical practice, multiple prophylaxis failures typically serve as an indicator of a highly treatment-resistant clinical phenotype. 26 – 28 Although the PREEMPT studies did not stratify results in detail according to the number of prior failures, our data confirm the observations of Negro et al. that patients who have failed fewer than three preventive trials generally achieve better outcomes than those with a more extensive treatment history. 8 – 10 , 20 Similarly, a positive family history may reflect a stronger genetic predisposition to higher baseline disease severity. 29 , 30 However, given the low discriminative performance of our multivariable model (AUC = 0.605, cross-validated AUC = 0.54), these statistical associations possess very limited predictive power and cannot be reliably used to guide clinical decisions or draw definitive neurobiological conclusions. The overall poor predictive ability of routine demographic and clinical variables strongly demonstrates that response to BoNT-A cannot be easily anticipated using standard clinical features alone, underscoring the limitations of relying solely on patient phenotypes for patient selection. Interestingly, our study did not confirm several predictive factors reported in other RWE studies, where parameters such as unilateral pain, younger age, or the presence of MOH were identified as predictors of BoNT-A response. 25 , 28 , 31 In our analysis, age, sex, and the presence of MOH did not reach statistical significance. This inconsistency across the literature, combined with the low predictive power of our own model, further reinforces that the clinical response to BoNT-A is highly variable and cannot be accurately anticipated using routinely collected clinical features alone. This study possesses several significant strengths. First and foremost is the large sample size of 479 patients, representing one of the largest real-world cohorts of CM patients treated with BoNT-A in Central and Eastern Europe. This substantial population provides high statistical power for identifying potential predictors of treatment response in a clinical setting. However, because several candidate clinical predictors and comorbidities were relatively uncommon in our cohort and given the poor overall discriminative performance of our multivariable model (cross-validated AUC = 0.54), this sample size was still insufficient to establish a highly predictive clinical model. This underscores that even within large real-world registries, standard clinical features remain limited in their capacity to predict individualized treatment responses. Second, the study employed a multicenter design, involving 12 specialized headache centers nationwide. This ensures that the findings reflect a broad clinical perspective and the application of standardized treatment protocols across the country. Third, our study provides unique insights into a highly specific “difficult-to-treat” population. Unlike many clinical trials, our cohort consisted of patients who had failed multiple prior preventive therapies, reflecting the stringent criteria of the national reimbursement program in Poland. Consequently, these results offer a realistic assessment of BoNT-A efficacy in the most refractory cases encountered in tertiary headache centers. Despite its clinical relevance and the inclusion of a large cohort, several limitations must be acknowledged. First, the retrospective and observational nature of the analysis may introduce inherent biases, such as selection bias or incomplete data for certain clinical variables. Second, the study was conducted within a strict national reimbursement framework, which may limit the generalizability of the results to other healthcare settings. While this provides valuable data on a “difficult-to-treat” population, it may also result in a ceiling effect, where the high degree of treatment resistance in the cohort leads to lower responder rates compared to less refractory populations. Third, the follow-up period was limited to nine months (three treatment cycles). Although this is sufficient to evaluate the initial response to BoNT-A, some studies suggest that “late responders” may emerge only after the fourth or fifth cycle; however, research indicates a low probability of achieving a clinical response if a reduction of at least 30% is not reached after the first two cycles. 32 Therefore, our results might slightly underestimate the long-term effectiveness of the therapy in this specific cohort. Furthermore, our analysis was restricted strictly to completers, which reflects the programmatic requirements of the reimbursement framework but may introduce attrition bias. Finally, due to the retrospective nature of the study and the standard framework of the national registry, we did not prospectively collect granular sub-classifications of higher response rates across all centers, which limits our ability to fully characterize the cohort’s super-responders. Finally, the therapeutic efficacy of BoNT-A is inherently operator-dependent, relying heavily on the injector’s precision, accuracy of anatomical site localization, and adherence to the injection technique. In multi-center settings, variations in administration can introduce significant confounding bias. To minimize this factor, our study utilized a strictly standardized protocol across all 12 centers, where all patients received exactly 195 units across 39 fixed sites according to the PREEMPT paradigm. Additionally, because all procedures were performed exclusively by highly experienced specialists in tertiary headache clinics, the potential impact of operator-dependent variability was maximally controlled and minimized.

Our study confirms that onabotulinumtoxin A is an effective preventive treatment for chronic migraine in a real-world Polish cohort. Despite the high degree of treatment resistance and long disease duration in our population, BoNT-A provided a statistically significant reduction in monthly headache days (MHD), monthly migraine days (MMD), and acute medication intake. We identified a positive family history and a history of multiple prior prophylaxis failures (≥4) as independent predictors of a poorer treatment response. However, the low predictive power of our multivariable model indicates that clinical and demographic factors alone are insufficient to accurately identify responders before initiating therapy. This suggests that the response to BoNT-A is likely governed by complex neurobiological and genetic mechanisms rather than observable clinical phenotypes. A crucial observation from our study is the negative impact of multiple prior treatment failures on the efficacy of BoNT-A. This finding highlights a significant clinical and systemic challenge: the delayed initiation of advanced migraine therapies, which may limit the ultimate success of the treatment.