학술지, 기념비적 스웨덴 연구를 정정: 재분석에서 성별 수술의 정신 건강 이점은 없었음

핵심 연구 수정: "성별 확인" 수술이 정신 건강을 개선한다는 증거 없음

Authors: Society for Evidence-Based Gender Medicine (SEGM)
Journal: Society for Evidence-Based Gender Medicine (SEGM)
Year: 2020
Source: segm.org/ajp_correction_2020

Summary

A 2019 Swedish study by Bränström and Pachankis in the American Journal of Psychiatry was widely reported as showing that gender surgery improves mental health. People whose surgery was longer ago were less likely to be getting mental health care in 2015; hormones showed no such link. After critical letters, the journal asked statisticians to review it and asked the authors to compare 1,018 surgery patients with 1,018 similar people who had not had surgery. That comparison showed 'no advantage of surgery'. The 2020 correction called the pro-surgery conclusion 'too strong' and fixed a 48% vs 37.9% error. This SEGM (Society for Evidence-Based Gender Medicine) article explains the correction accurately. But the new comparison was an afterthought, not part of the original plan, the paper was not withdrawn, and SEGM is a side in the debate.

Key findings

2019년 원래 연구는 970만 명을 포괄하는 스웨덴 국가 건강 기록을 이용했으며, 2005년부터 2015년 사이 성별 불일치 진단을 받은 2,679명이 포함되어 있고, 정신 건강 치료를 2015년 한 해만 살펴보았습니다.
호르몬을 쓴 햇수와 정신 건강 치료 사이에는 관련이 없었지만, 마지막 수술 후 매년 치료 오즈가 약 8% 낮았습니다. 저자들은 이것이 수술 제공을 뒷받침한다고 말했습니다.
스웨덴, 노르웨이, 미국, 영국 연구자들의 비판 서한 후, 저널 편집자는 두 차례의 통계 검토를 지시하고 저자들에게 수술 환자 1,018명과 비슷한 비수술 환자 1,018명을 비교하라고 요청했습니다.
2020년 8월 정정문은 이 비교가 기분 또는 불안 진료, 처방, 자살 시도 후 입원에서 ‘수술의 이점 없음’을 보여주었다고 밝혔습니다.
정정문은 이 연구가 시간 순으로 추적하거나 치료를 무작위로 배정하지 않았기 때문에 수술에 우호적인 결론이 ‘너무 강했다’고 했습니다. 또한 수술을 받은 비율을 48%에서 37.9%로 바로잡았습니다.
논문은 철회되지 않았습니다. 추가 주석과 함께 다시 게시되었고, 답변에서 저자들은 수술 후 경과 기간 발견을 옹호하며 비수술 환자는 부적절한 비교 집단이라고 주장했습니다.
새 비교에서 수술 환자는 불안 치료를 받을 가능성이 다소 높았고, 자살 시도 후 입원이 13건 대 7건이었습니다. 이 차이는 우연일 가능성이 충분하며, SEGM은 이것이 해를 보여주는 것은 아니라고 말합니다.
대부분의 수술이 최근이었습니다. 1,018명의 75% 넘게가 2015년 기준 3년 이내였고 10년 이상 전이었던 것은 19명뿐이어서, 이 연구는 장기 결과에 대해 말해주는 것이 거의 없습니다.

Abstract

In October 2019, the American Journal of Psychiatry (AJP) published a study from the Karolinska Institute in Sweden, and the Yale School of Public Health which reported that “gender-affirming" surgeries for gender dysphoric patients are associated with improved mental health outcomes. Looking at mental health utilization in the year 2015, a retrospective analysis showed that the more time passed since surgery, the fewer mental services were utilized by patients, with an average 8% reduction in mental health utilization for each year following surgery. From this, the study concluded that surgery has a beneficial effect on mental health, and that benefits continue to accrue over time. However, following a reanalysis of the data, this conclusion has now been officially corrected to indicate that there is “no advantage of surgery.”

Conclusion

After reanalysis with a control group comparing gender dysphoric patients who received surgery to those who did not, no statistically significant difference was found in mental health utilization measures. The AJP correction stated the original conclusion that 'the longitudinal association between gender-affirming surgery and lower use of mental health treatment lends support to the decision to provide gender-affirming surgeries' was too strong. The Society for Evidence-Based Gender Medicine (SEGM) concluded that no mental health benefit of gender-affirming surgery was demonstrated in the corrected analysis.

Methodological limitations

Written by a group with a position: This record is SEGM's explanation, not the correction itself. SEGM argues for caution in gender medicine and notes that some of its advisers wrote the critical letters, so its framing should be read with that in mind.
New comparison was an afterthought: The surgery vs no-surgery comparison was not part of the original question. The correction notes that the groups may differ in several other ways, so the 'no difference' result is also not strong proof either way.
Comparison group mixes different people: The authors argued that people diagnosed with gender incongruence who had not had surgery include many who never wanted it. So finding no difference does not show surgery has no benefit for those who seek it, a point SEGM accepts.
Corrected, not withdrawn: The journal corrected the conclusion and reposted the article, but did not withdraw the time-since-surgery link, which the authors stood by. Saying 'the study was wrong' goes further than the formal correction.
One-year snapshot: Both analyses looked at mental health treatment in 2015 only, comparing different people at different times since surgery rather than following the same people over time. They cannot show change in any one person.
Indirect measures: Visits, prescriptions and hospital stays measure use of services, not symptoms or quality of life. Required mental health assessments before surgery may have pushed up service use around the time of surgery.
Deaths and moves abroad missed: Only people alive and living in Sweden in 2015 were included. If people died, including by suicide, after surgery, those with surgery long ago would look healthier than they were. The authors admitted this.
Patients changed over time: More recent patients were younger, more often born female and more often had other mental health problems. This alone could explain why people with older surgeries used fewer services, regardless of surgery.
Many tests, one hit: The single positive result in the original was only just clear of chance, and came alongside several other measures tested without allowing for the extra chance of a fluke, as letter writers pointed out.
Doubtful 'lost patients' claim: SEGM suggests about 90% of patients were 'lost', because only 19 had surgery before 2005 compared with 324 in an earlier Swedish study. But the study only included people diagnosed from 2005 to 2015, so people treated earlier were mostly outside the study rather than lost.
Too few suicide-related cases: Hospital stays after suicide attempts (13 vs 7) are too few to support conclusions either way, and the new comparison was too small to test this properly.

detrans.ai

Summary

Key findings

Abstract

Conclusion