eTable 6. Critical assessment of methods*¹.

Study	Year	Major domains						Minor domains			Overall risk
		1. Recruitment procedure & follow-up (in cohort studies)	2. Exposure definition and measurement	3. Outcome. Source and validation	4. Confounding	5. Analysis method	6. Chronology	7. Blinding of assessors	8. Funding	9. Conflict of interest
Fernandes	(2018)	↓	↑	↓	↑	↑	↑	→	↑	↑	↓
Iosifidis	(2015)	↑	↑	↑	→	→	↑	→	→	→	↓
Kujala	(1994)	↑	↑	↑	↑	↑	↑	↑	↑	→	↑
Kujala	(1995)	↓	↑	↑	↑	↑	↑	→	↑	→	↓
Roos a)*2	(1994)	↑	↑	↑	↑	↑	↑	↑	↑	→	↑
Roos b)*2	(1994)	↑	↑	↑	→	→	↑	↑	↑	→	↓
Sandmark a)*3	(1999)	↑	↑	↑	↑	↑	↑	→	↑	→	↑
Sandmark b)*3	(1999)	↑	↑	↑	→	→	↑	→	↑	→	↓
Thelin	(2006)	↑	↑	↑	↑	↑	↑	↑	↑	→	↑
Tveit	(2012)	↑	↑	↓	↑	↑	↑	→	↑	↑	↓
Vrezas	(2010)	↑	↑	↑	↑	↑	↑	→	↑	↑	↑

↓ Low risk of bias, ?high risk of bias, → unclear risk of bias

*¹ Details of the assessment of the methods used in the individual studies can be requested from the authors.

*² The two different assessments for Roos et al. (1994) relate to different odds ratios (in some cases calculated by the review authors). In calculating the odds ratio a) (6.32 [1.94; 20.66]), age and sex was considered, so domains 4 and 5 are rated as low risk, leading to classification of the overall risk as low. For all other odds ratios b), it is not clear whether the age was similar in the exposure and comparison groups, so domains 4 and 5 are rated as unclear risk and the overall risk is classified as high.

*³ The same applies to the different assessments for Sandmark & Vingard (1999). For one of the odds ratios a) (2.0 [1.4; 2.8]) calculated by the authors, adjustment was made for age among other factors. Moreover, only men were included in the study, so the variable “sex” was also considered. Domains 4 and 5 were thus rated as low risk, leading to classification of the overall risk as low. For the other odds ratios b), it is not clear whether the age was similar in the exposure and comparison groups, so domains 4 and 5 are rated as unclear risk and the overall risk is classified as high.