Methodological Limitations

• Extremely small sample size (n=9) with no control group, severely limiting generalizability and statistical power. The authors acknowledge this but proceed to draw clinical conclusions regardless.
• Retrospective data collection with inherent recall bias. Patients were asked to retrospectively explain their original motivations for transitioning years later, after developing regret, which may distort memories to align with current narratives.
• Selection bias: The sample only includes detransitioners who returned to the same clinic for medical/legal services. The authors note that only 24-29% of detransitioners inform their clinic, meaning this study captures a highly unrepresentative minority of motivated patients.
• Exclusion of four patients from an already tiny sample (n=13→9), including one excluded for developing psychosis after GAT, which appears to arbitrarily remove a potentially relevant adverse outcome. Three others were excluded for seeking 'second-opinion evaluation' at a different hospital, which may introduce selection bias toward patients who remained within the same institutional system.
• Circular reasoning in conclusions: The authors conclude that 'evaluating and treating serious psychiatric illnesses first...might reduce cases of detransitioning,' yet all patients in this sample DID receive psychiatric evaluation before initial GAT, and the study provides no evidence that more evaluation would have changed outcomes.
• Conflation of correlation with causation: The high prevalence of psychiatric comorbidities and trauma is interpreted as evidence that these 'caused' gender dysphoria and transition desires, but no longitudinal or causal analysis is performed. Alternative explanations (e.g., minority stress, social marginalization causing both trauma and gender dysphoria) are not considered.
• Use of outdated diagnostic instruments (SCID-IV from 2005) during the detransition assessment phase, despite DSM-5 being published in 2013. The initial assessments used SCID-IV for DSM-IV diagnoses, creating diagnostic inconsistency over time.
• Potential observer/experimenter bias: The psychiatrist collecting retrospective data (K.K.) also appears to be the first author, and was involved in implementing clinic changes based on findings. The psychologist who assessed patients (L.L.) is also a co-author. No blinding or independent verification is described.
• Methodological inconsistency in regret classification: The authors state they use 'Pfäfflin's classification' for regret, but Pfafflin's original work focused specifically on post-surgical regret, while most patients in this sample did not have genital surgery. The classification may not be validly applied to hormone-only or partial medical transitions.
• Lack of standardized measures for key outcomes: 'Regret' was not assessed with validated instruments but through clinical judgment. The visual analog scales for gender were added only in 2018, not during initial assessments, preventing direct within-subject comparison.
• Confounding by indication: Patients who independently initiated hormones or had private surgeries before completing evaluation (3 patients) represent a systematically different group, yet are pooled with others without subgroup analysis.
• The authors' stated aim to 'critically examine the psychiatric evaluation process...to decrease the occurrence of serious complications such as wishes to detransition' suggests a predetermined conclusion that detransition represents evaluation failure, potentially biasing data interpretation.
• Inappropriate comparison to general transgender population: The authors compare psychiatric prevalence in their detransition sample to Heylens et al.'s study of transgender individuals, but these are fundamentally different populations at different time points (pre-transition vs. post-detransition), and the comparison ignores that psychiatric conditions may result from transition experiences rather than pre-existing causes.
• Missing data and inconsistent reporting: Table 3 shows 'n1 = All diagnoses at initial assessment' and 'n2 = New diagnoses at detransition,' but the sum of these does not clearly represent current diagnostic status, and some categories show impossible patterns (e.g., 'reported problematic use' of substances decreases from 4 to 0, yet new diagnoses are 0).
• The ECR-R attachment measure was only administered to 6 of 9 patients, with no explanation for missing data. The clinical significance of attachment styles in this context is speculative.
• Potential funding bias: The study was funded by Helsinki University Hospital (TYH2021128), the same institution where authors work and where clinic changes were implemented. While declared as 'no conflict of interest,' the institutional incentive to demonstrate quality improvement may influence interpretation.