1.3.1. ROR method.

This method evaluates the strength of association between a specific drug and an AE by calculating the ratio of reported events. According to standardized FDA criteria, a statistically significant potential safety signal is identified when the lower bound of the 95% confidence interval (95% CI) for the ROR is greater than 1, and the number of reports exceeds three.

1.3.2. PRR (based on the MHRA comprehensive criteria) method.

The PRR method assesses drug-specific risk by comparing the proportion of a specific AE reported for the target drug with that reported for all other drugs. According to UK Medicines and Healthcare products Regulatory Agency (MHRA) criteria, a safety signal is considered present when the PRR is greater than 2, the χ² statistic exceeds 4, and the number of reports is at least three.

1.3.3. MGPS method.

MGPS uses Bayesian statistical modeling to quantify associations between drugs and AEs. Signal strength is measured using the Empirical Bayes Geometric Mean (EBGM). A statistically significant safety signal is defined when the lower bound of the 95% CI of EBGM (EBGM05) is greater than 2.

1.3.4. BCPNN method.

Based on Bayesian inference, the BCPNN method evaluates associations between drugs and AEs by calculating the Information Component (IC). According to established criteria, a potential safety signal is identified when the lower bound of the 95% CI of IC (IC–2SD) is greater than 0.

2.5.1. Basic information and demographic characteristics of AE reports in the EV database.

The EV database included 11,906,424 background patients, accounting for 50,987,250 reported AEs. Among these, 34,920 patients had reports involving isotretinoin, corresponding to 124,146 AE occurrences. The overall data volume was comparable to that of the FAERS database.

Regarding sex distribution, females accounted for 50.87% of cases, males for 43.33%, and unspecified gender for 5.80%, closely mirroring FAERS data. Most reports involved individuals aged 18–64 years (49.45%), followed by those younger than 18 years (23.62%). This age distribution aligned with observations in the FAERS database, indicating that reported AEs were predominantly concentrated among adolescents and young adults. Detailed data are presented in Table 6.

Download:

• PNG
  larger image
• TIFF
  original image

Table 6. Basic information of AE reports related to isotretinoin in the EV database.

https://doi.org/10.1371/journal.pone.0329161.t006

3.4.1. Isotretinoin and AE signals related to psychiatric disorders.

Currently, the association between isotretinoin and psychiatric disorders remains controversial. Studies suggest that isotretinoin may elevate risks of anxiety and depression by reducing serotonin levels, increasing neuronal apoptosis, regulating gene expression, affecting neural plasticity, or activating the HPA axis [19]. However, other research indicates that isotretinoin is primarily prescribed for moderate-to-severe acne, a condition inherently associated with a higher prevalence of psychiatric disorders [20], particularly among adolescents and young adults. Patients often experience depression, anxiety, and suicidal tendencies due to changes in physical appearance and social anxiety [21]. Thus, patients prescribed isotretinoin inherently face a higher risk of psychiatric disorders. The influence of acne itself must be comprehensively considered when investigating psychiatric side effects related to isotretinoin [22]. Recent large-scale cohort studies and meta-analyses have not definitively confirmed a causal relationship [23].

It should be noted that psychiatric signals detected in this study represent statistical associations only and do not imply direct causation by isotretinoin. Acne itself and its associated psychosocial stress may serve as confounding factors for psychiatric symptoms. Future prospective cohort or case-control studies are required to further investigate causality.

Although existing evidence does not sufficiently establish causality between isotretinoin and psychiatric disorders, the association cannot be entirely ruled out. Monitoring of psychiatric AEs is recommended in both domestic and international prescribing information for isotretinoin. This study identified strong psychiatric AE signals based on extensive real-world data from isotretinoin users, warranting serious attention. Clinically, mental health screening prior to isotretinoin therapy is advised. Psychological status should be routinely monitored throughout treatment. Personalized management plans should be implemented for high-risk patients. Psychological support and regular follow-up are necessary during therapy to mitigate psychiatric risks. Education on medication for both patients and families must be strengthened. Patients displaying psychiatric symptoms should promptly seek medical care. Multidisciplinary management approaches should be improved, including collaboration with psychiatric or psychological specialists as needed.

3.4.2. Isotretinoin and AE signals related to IBD.

This study identified an exceptionally strong signal for inflammatory bowel disease (IBD), with a reporting odds ratio (ROR) of 579.14, ranking first among all detected signals and second in the number of reported cases, indicating an unusually high statistical strength. Given the extraordinary magnitude of the IBD signal, a detailed analysis of reporting patterns was conducted to better contextualize this finding.

In total, 5,278 IBD reports were identified, of which 4,483 (84.94%) were submitted by legal professionals. Temporal analysis further revealed a significant clustering of reports between 2009 and 2014, during which 4,526 cases (85.75%) occurred. In contrast, since 2018, fewer than 10 IBD reports have been submitted annually. This markedly skewed distribution in reporter type and reporting period strongly suggests non-random reporting dynamics.

To further elucidate the potential reasons for the strong isotretinoin-IBD association observed, the broader regulatory and legal context was examined. Following 2009, multiple high-profile lawsuits were filed in the United States against isotretinoin manufacturers, coinciding with a marked surge in IBD reports from 2009–2012. Previous investigations indicated that a substantial proportion of FAERS reports linking isotretinoin to IBD originated from law firms, highlighting potential litigation-driven stimulated reporting [24, 25]. The reporting patterns observed in the present study closely align with known timelines of litigation activity in the United States. These results are consistent with previous studies documenting attorney-driven waves of isotretinoin-associated IBD reports in FAERS. Collectively, these findings indicate a period of litigation-driven overreporting likely occurred after 2009, and the exceptionally strong IBD signal detected here may largely reflect reporting bias.

Importantly, accumulating high-quality clinical evidence from recent years has consistently failed to demonstrate an independent association between isotretinoin exposure and IBD development. Large-scale population-based studies conducted over the past decade, along with updated meta-analyses and contemporary clinical guidelines, consistently suggest that IBD is predominantly driven by genetic susceptibility, environmental exposures, and immune dysregulation. These studies emphasize that such confounding factors frequently affect these patient populations, and that a direct causal relationship between isotretinoin and IBD has not been established [8,26–29]. Notably, large cohort studies employing propensity score–matched designs have shown that isotretinoin treatment, compared with appropriate acne-control populations, does not increase the risk of incident IBD [25,29]. Furthermore, several studies have reported that acne itself, rather than isotretinoin, is more strongly associated with subsequent IBD development, highlighting the critical role of confounding by indication in earlier observational findings. Systematic reviews and meta-analyses incorporating recent cohort data and trial sequential analysis further confirm no statistically significant association between isotretinoin use and IBD [26].

Collectively, although an exceptionally strong statistical signal for IBD was observed in this study, its clinical and pharmacological implications must be interpreted cautiously. Clinicians and regulatory authorities should remain attentive to shifts in reporting patterns, continuously integrate emerging high-quality clinical evidence, and provide scientifically grounded clarifications in drug labeling as appropriate. Such measures are essential for informed clinical decision-making and accurate public understanding of AE signals identified through pharmacovigilance systems.

3.4.3. Teratogenicity of isotretinoin.

Isotretinoin can cross the placental barrier and is classified as pregnancy safety Grade X by the FDA. Clinical studies indicate that oral isotretinoin increases the risk of severe congenital malformations, spontaneous abortion, and premature birt [30]. This study found that isotretinoin ranked third among positive signals involving congenital, familial, and genetic disorders, mainly affecting the ears and labyrinth, central nervous system, cardiovascular system, musculoskeletal system, and eyes. This finding aligns with the prescribing information and existing literature [31,32]. The top ten congenital abnormalities identified were congenital middle ear abnormalities, anorectal malformations, congenital cerebellar hypoplasia, congenital optic nerve abnormalities, hemihypertrophy, congenital inner ear abnormalities, cerebellar dysplasia, anophthalmia, congenital ear malformations, and follicular keratosis. Special attention should be given to hemihypertrophy and follicular keratosis, which exhibited strong positive signals but are not listed in isotretinoin prescribing information. Current studies suggest potential associations between isotretinoin exposure and partial developmental asymmetry, unilateral abnormal growth, skin dysplasia, and keratinization disorders [33,34]. However, definitive correlations remain unconfirmed, and further research is necessary.

3.4.4. Isotretinoin and pregnancy events.

This study identified 6,453 AEs related to pregnancy, puerperium, and perinatal conditions, of which 4,641 (71.92%) were specifically pregnancy-related. Among these, unexpected pregnancy demonstrated strong positive signals, ranking within the top 30 AE signals. These findings highlight the necessity for enhanced management of pregnancy events in clinical practice. Isotretinoin carries a high risk of severe birth defects and is contraindicated for pregnant women or those who may soon become pregnant. According to prescribing information in some countries, such as China, women of childbearing age or their spouses must use effective contraception from three months before isotretinoin initiation, during treatment, and for three months following cessation. The US FDA guidelines require contraception for at least one month before treatment, during treatment, and for three months after discontinuation.

Given the unequivocal teratogenic risk of isotretinoin, numerous pregnancy prevention programs (PPPs) have been implemented worldwide. Among these, the U.S. iPLEDGE program is widely considered one of the most stringent regulatory strategies for managing risk [35,36]. Comparative studies have demonstrated substantial heterogeneity among PPPs internationally regarding regulatory rigor, contraceptive requirements, frequency of pregnancy testing, and enforcement mechanisms. In particular, iPLEDGE imposes a significantly greater administrative burden on both patients and healthcare providers [37].

Despite these measures, real-world evidence indicates suboptimal effectiveness of PPPs. Although such programs have existed for many years, approximately 3.3%–6.5% of isotretinoin users still experience pregnancy during treatment [38]. Contributing factors likely include inadequate adherence to contraceptive regimens, insufficient patient education on pregnancy prevention, and inappropriate selection of contraception methods. Population-based studies and registry analyses from the United States and Europe further indicate that, in routine clinical practice, adherence to key PPP components, such as contraceptive use, regular pregnancy testing, and enrollment procedures, is frequently poor, with complete compliance rates often below 50% [37,39–41]. Moreover, previous studies has reported that some patients fail to fully recognize the profound teratogenic risks associated with isotretinoin before starting treatment. This lack of understanding, combined with poor contraceptive adherence, contributes to persistently high unintended pregnancy rates [38].

The limited success of PPPs is multifactorial. Patient-level factors include inadequate health literacy, limited access to contraceptive resources, and socioeconomic barriers. System-level challenges involve procedural complexity, pharmacy accessibility, and difficulties navigating digital platforms. Additionally, provider-level issues such as workflow burdens and inconsistent counseling quality further undermine effectiveness [39–42]. Recent real-world cohort studies also show notably lower adherence among younger women and socioeconomically disadvantaged groups [40,42]. Collectively, these findings indicate that while programs like iPLEDGE have significantly reduced fetal exposure risks, they remain insufficient to completely prevent pregnancies. This highlights the necessity for ongoing optimization of pregnancy prevention strategies informed by real-world adherence data.

Based on these findings, clinicians and pharmacists must consistently remind patients to rigorously use contraception during and after isotretinoin treatment. Patients should prioritize highly effective contraceptive methods or employ multiple contraceptive strategies. Pregnancy monitoring should be intensified, and strengthening patient education is critical for improving compliance. Regulatory agencies should further enhance and reinforce pregnancy prevention measures to ensure isotretinoin is used safely.

3.4.5. Biological plausibility and context of novel AE signals.

Several AE signals identified in this study are not currently included in the product labeling for isotretinoin. Among the highest-ranked signals were nasal vestibulitis, anal papillary hypertrophy, neonatal neuroblastoma, diverticular hernia, and SAPHO syndrome.

From a mechanistic perspective, isotretinoin exerts therapeutic effects by modulating the retinoic acid signaling pathway. This modulation suppresses sebaceous gland activity and alters epithelial differentiation. Such effects are known to cause pronounced drying and thinning of the nasal mucosa [43], representing a plausible mechanism for isotretinoin-associated nasal vestibulitis. Clinical studies have demonstrated that systemic isotretinoin therapy can impair nasal mucociliary clearance and disrupt mucosal barrier integrity. These disruptions may increase susceptibility to microfissures and secondary bacterial colonization in the nasal vestibule [43, 44]. Although nasal vestibulitis is infrequently described as a discrete clinical endpoint, commonly reported treatment-related symptoms, including nasal dryness, irritation, and epistaxis, indirectly support inflammatory involvement of the nasal vestibule [43,44].

The biological plausibility of anal papillary hypertrophy may similarly reflect established regulatory effects of retinoic acid signaling on gastrointestinal epithelial proliferation and mucosal homeostasis [45, 46]. Experimental studies suggest that retinoic acid can exert trophic and remodeling effects on intestinal epithelium under specific conditions [45, 47]. Clinically, direct evidence linking isotretinoin to anal papillary hypertrophy remains limited. However, isotretinoin-induced mucosal dryness and altered epithelial turnover may plausibly increase susceptibility to local irritation and reactive hyperplasia.

Regarding neonatal neuroblastoma, biological considerations require careful interpretation. Neuroblastoma is an embryonal tumor originating from neural crest cells. Migration, differentiation, and survival of these cells during early embryogenesis depend heavily on tightly regulated retinoic acid signaling [48]. Experimental data suggest that excessive or dysregulated retinoic acid exposure during critical developmental periods may disrupt neural crest–related gene expression and apoptotic pathways. This disruption could theoretically predispose to developmental abnormalities [49]. However, current clinical and epidemiological evidence does not support a causal relationship between maternal isotretinoin exposure and neonatal neuroblastoma. Conversely, isotretinoin and other retinoids are widely employed as differentiation agents in treating high-risk neuroblastoma. Their antitumor activity primarily occurs through promoting cellular maturation and apoptosis rather than inducing tumorigenesis [50, 51]. Reports of isotretinoin embryopathy accompanied by congenital neuroblastoma exist only as isolated cases and are insufficient to establish causality [52].

The potential biological mechanism underlying diverticular hernia may involve the regulatory effects of retinoic acid signaling on extracellular matrix remodeling and collagen metabolism. These processes are critical for maintaining the structural integrity of the gastrointestinal wall [53]. Abnormalities in collagen composition and matrix turnover have been implicated in the development of diverticular disease and hernia formation [54].

SAPHO syndrome is currently not listed among adverse reactions in approved isotretinoin product labeling. Biologically, SAPHO syndrome is a rare autoinflammatory disorder characterized by dysregulated innate immune responses and abnormal bone remodeling. Retinoic acid signaling has well-established immunomodulatory effects and influences bone metabolism, including regulation of proinflammatory cytokines and osteogenic pathways. This provides mechanistic plausibility for retinoid-mediated modulation of SAPHO-related pathways in susceptible individuals [55–57]. Clinically, associations between isotretinoin and SAPHO syndrome have been primarily reported in sporadic case reports without systematic controlled studies. Notably, retinoids have also been used therapeutically in selected SAPHO syndrome patients, further complicating causal interpretation [58].

Overall, this study identified several previously unlabelled AE signals associated with isotretinoin that exhibited relatively strong disproportionality signals. However, the absolute number of reports for these events was generally small, potentially limiting the precision and robustness of the findings. Although biological plausibility exists for some associations, current clinical and epidemiological evidence remains insufficient to establish causality. Continued surveillance of pharmacovigilance databases, alongside well-designed clinical and mechanistic studies, is necessary to validate these signals further. The present findings provide preliminary guidance on safety issues warranting attention in clinical practice and future research.

3.4.6. Analysis of external validation.

From an epidemiological baseline perspective, the EV and FAERS databases were highly comparable in background patient numbers, target drug report numbers, and patient sex and age distributions. These similarities strengthen the credibility of our findings. At the System Organ Class (SOC) level, results from the EV database were highly consistent with those from FAERS.

Specifically, three key aspects of concordance were observed. First, SOCs involved in overall AE reports and positive AE signals were identical between the two databases, confirming that isotretinoin-associated AEs affect multiple organ systems. Second, SOCs highly ranked in terms of signal counts and report frequencies in the FAERS database similarly ranked near the top in the EV database, with only minor ordering differences. In the EV database, psychiatric disorders, congenital, familial, and genetic disorders, and gastrointestinal disorders remained the top three SOCs ranked by signal count, confirming that AEs in these categories are among the most frequently encountered in clinical practice. Third, SOCs exhibiting cumulative positive signals in FAERS were also positive in the EV database. Notably, the EV database demonstrated an additional cumulative positive signal for congenital, familial, and genetic disorders. Further analysis revealed that this SOC was borderline positive in FAERS under the MGPS method, with an EBGM05 value of 1.99 (threshold >2), while the remaining three methods indicated positive signals.

Collectively, these findings further support the feasibility of the methodology and the robustness of the results. In the FAERS database, 469 positive AE signals associated with isotretinoin were identified, of which 396 (84.43%) overlapped with signals in the EV database. Among the top 30 signals, only Kleine–Levin syndrome and hemihypertrophy were not identified as positive in EV. This discrepancy likely results from the limited number of related reports in EV.

In addition, given the strong signal for inflammatory bowel disease (IBD) identified in EV, further analyses were performed using raw data. A total of 6,300 IBD reports were identified, with 6,112 submitted by non–healthcare professionals. Between 2009 and 2014, 5,547 IBD reports (88.05% of the total) were recorded. Since 2020, fewer than 10 IBD reports have been submitted annually. This temporal reporting pattern closely mirrors observations from FAERS, suggesting increased global reporting during a specific period. This spike may be attributed to heightened international awareness following events involving U.S. manufacturers.

Overall, the substantial overlap in positive signals between the two databases provides strong external validation, confirming the reliability and reproducibility of this study’s findings.

4.1.1. Methodological novelty and external validation.

We pioneered the systematic integration of four internationally recognized pharmacovigilance algorithms (ROR, PRR, BCPNN, MGPS). This integrated approach was applied comprehensively to 142,160 AE reports associated with isotretinoin over two decades (2004–2024) from the FAERS database. This approach significantly enhanced the robustness of signal detection and improved identification of rare events.

A key innovation was the inclusion of external validation using an independent pharmacovigilance database. AE signals identified for isotretinoin in FAERS were systematically replicated in the EudraVigilance database. The results showed high concordance in epidemiological characteristics, System Organ Class distributions, and signal detection outcomes. This dual-database framework enhances reproducibility, reduces database-specific bias, and strengthens the robustness and generalizability of the findings, supporting more reliable post-marketing safety assessments.

4.1.2. Expansion of the risk profile.

We established the most comprehensive AE profile for isotretinoin to date, identifying 469 consolidated safety signals. Notably, 8 previously unlisted signals (including nasal vestibulitis, anal papilla hypertrophy, and neonatal neuroblastoma) provide preliminary insights into safety concerns that warrant attention in clinical practice and future research.

4.1.3. Novel risk management findings.

Unintended pregnancy (ROR = 91.39) emerged as the strongest pregnancy-related signal. This finding provides critical empirical evidence for optimizing global Pregnancy Prevention Programs (e.g., iPLEDGE) and highlights potential vulnerabilities in current risk management protocols.