2. Characteristics of nonrandomised studies (considered for safety analysis).

Study	Methods	Participants	Interventions / Exposures	Outcomes (Psychiatric adverse effects)	Summary  Findings (Frequencies of psychiatric adverse effects)	Assessment of the risk of bias (Turner 2013)
Azoulay 2008	Design: Case‐control cross‐over Duration of the study: not provided Period of recruitment: from January 1, 1984, and December 31, 2003	Setting: Québec, Canada Initial cohort: 30,496 participants Cases (n = 126): age, mean ± SD (years): 28.1 ± 9.0 gender: 37.3% males (n = 47 Inclusion criteria: having been exposed to at least one oral isotretinoin prescription between January 1, 1984, and December 31, 2003 and diagnosed or hospitalised for depression, with an antidepressant prescription in the 30 days following the diagnosis or hospitalisation; being covered by the Régie de l' Assurance Maladie du Québec (RAMQ) drug plan for at least 12 months prior to the date of depression diagnosis (index date); having at least 1 diagnosis of acne vulgaris at any time prior to the index date * Exclusion criteria: having received an antidepressant prescription in the 12 months prior to index date	Exposure group: oral isotretinoin use in a five‐month period immediately prior to the index date Control group: oral isotretinoin use out of the 5‐month risk period (considering a 2‐month wash‐out period) prior to the index date	Incidence of a first diagnosis or hospitalisation for depression in each specific time window Dose response of oral isotretinoin on the incidence of depression Outcomes assessments occurred by linkage of diagnosis codes and antidepressant prescriptions with medical records	Adjusted relative risk of depression associated with oral isotretinoin use was 2.68 (95% CI 1.10 to 6.48). There were no statistically significant differences in incidence of depression before and after Canadian label change warning about isotretinoin possible psychiatric risks There was no significant association between a cumulative dose range and the occurrence of depression	Selection bias: High risk Although authors had reported an adjusted analysis for some covariates, the study did not consider either acne severity or time with disease as confounders Performance bias (all outcomes): High risk Comment: As the study had a retrospective design, probably personnel and participants were not blinded regarding the exposition/drug Detection bias (all outcomes): Unclear risk Comment: The study did not provide any information regarding blinding of the outcome assessors Attrition bias (all outcomes): Unclear risk Comment: There was no information about missing data or the potential for data to be missing to permit judgment. Selective reporting bias: Low risk Comment: no protocol available; however, there was an adequate report of outcomes listed in methods section Other bias: High‐risk The study used an instrument with a poor accuracy to measure outcomes: record linkage across electronic databases. Also, the study had presented only a case‐cross‐over analysis and this design had a potential risk of bias related to the possibility of carry‐over effect in participants of nonexposed groups, despite the 2‐month wash‐out period
Chia 2005	Design: Prospective cohort   Duration of the study: not clearly provided   Period of recruitment: between October 1998 and December 2001   Factors influencing the choice of treatment: history of previous treatment failure, participant/parent preference, out‐of‐pocket medical cost, ability/willingness to comply with treatment requirements for frequent visits and phlebotomy, participant/parental concerns about adverse drug effects, and objections to oral contraceptive therapy	Setting: United States, two centres at Missouri: Departments of Dermatology at Saint Louis University, Saint Louis, and at University of Missouri, Columbia. Both were outpatient clinics, one urban/hospital affiliated and the other suburban/community affiliated   Gender: Group 1: n = 59 (44  males and 15 females) Group 2: n = 73 (44 females and 29 males)   The study provided no information regarding duration of acne in enrolled participants   Inclusion criteria: male and female participants between the ages of 12 and 19; being presented for treatment of moderate to severe inflammatory and cystic acne   Exclusion criteria: history of or current DSM‐IV (Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition) Axis I diagnosis; prior use of or allergy to isotretinoin, and pregnancy	Group 1: oral isotretinoin at approximately 1 mg/kg per day, rounded to the nearest 20 mg   Group 2: conservative therapy defined as a topical antibiotic, topical retinoid, and twice‐daily administration of an oral antibiotic   Duration of interventions: not provided	Frequency of CES‐D (The Center for Epidemiologic Studies Depression Scale) scores of 17 or higher, which were suggestive of clinically significant depression. The CES‐D scale is validated in adolescents and considered more sensitive and less specific than other depression instruments   Incidence of suicidal ideation in all participants who scored 17 or higher in CES‐D, assessed by interviewing participants with the mood disorders portion of the Structured Clinical Interview for DSM‐IV Axis I (major mental disorders)   Measurements occurred at baseline and after completing 3 to 4 months of follow‐up	101 subjects completed the study (49 in Group 1; 52 in Group 2). After adjustment for baseline CES‐D and participant gender, there was not a significant association between treatment status (isotretinoin therapy versus conservative therapy) and follow‐up CES‐D score (prevalence at this time point) suggestive of clinically significant depression (odds ratio (OR), 1.1; 95% CI, 0.23 to 5.6). Also intention‐to‐treat analysis did not modify this result (OR, 1.3; 95% CI, 0.31 to 5.7)   CES‐D scores of 17 or higher (> 16) at the 3 or 4 months of follow‐up suggesting new‐onset  of depression (incidence measurement) occurred in 4.1%  (2 cases) of participants  in the oral isotretinoin group, and in 3.8% of participants in  the conservative therapy group (RD 0.3; 95% CI, ‐7.7 to 8.3). The difference between the groups was not statistically significant, even after performing an intention‐to‐treat analysis   No participant had suicidal ideation between participants with CES‐D scores > 16 in the isotretinoin group.  One subject at baseline presented suicidal ideation in the control group (an incidence of 1.4%)	Selection bias: High‐risk Comment: The study analysed the prevalence of CES‐D > 16 at follow‐up using multiple logistic regression, with baseline CES‐D score and gender as covariates in the analysis. However, the incidence of CES‐D > 16 at follow‐up was not presented using a method for control of confounding. Other potential confounders, such as duration of acne and acne severity, were not considered in the study. Also the two different settings of recruitment (one urban/hospital‐affiliated and the other suburban/community‐ affiliated outpatient clinics) could be a confounding factor, and the analysis did not consider this concern Performance bias (all outcomes): Unclear risk Comment: The study did not provide any information regarding blinding of participants and personnel Detection bias (all outcomes): Unclear risk Comment: The study did not provide any information regarding blinding of the outcome assessors Attrition bias (all outcomes): Low risk Comment: An intention‐to‐treat analysis, with adjustments for baseline CES‐D and gender, presented the same results of the adjusted analysis that excluded those unavailable for follow‐up. Despite the high level of loss to follow‐up in the study, there was no potential impact of missing data for all outcomes Selective reporting bias: Low risk Comment: No protocol available; however, there was an adequate report of outcomes listed in methods section Other bias: High risk Comment: The CES‐D is a highly sensitive screening tool, which could have overestimated the degree of mood disorder compared with other psychiatric instruments and have exaggerated the adverse effects estimates of the interventions in this study
Cohen 2007	Design: Prospective cohort   Duration of the study: not clearly provided   Period of recruitment: not provided	Setting: community dermatology practice in Calgary, Canada   Oral isotretinoin: n = 100 (41% male) Oral antibiotics: n = 41 (34.1% male) Topical acne therapy: n = 59 (18.6%  male)   Age – median: Oral isotretinoin: 21.5 years Oral antibiotics: 26.0 years Topical acne therapy: 20 years   Duration of acne: not provided   Acne severity: percentage of participants with moderate or severe Oral isotretinoin (n = 100): 96%, 96 participants Oral antibiotics (n = 41): 58.5%, 24 participants Topical acne therapy (n = 59): 6.8%, 4 participants   Inclusion criteria: being 14 years old, or older, able to provide informed consent, and not currently under pharmacological treatment for depression; not anticipating an alteration in residence during the period of the study; and providing at least two options for follow‐up contact Exclusion criteria: not provided	Intervention group (n = 100): ‐ oral isotretinoin: dosage not specified   Control group (n = 100): Oral antibiotic (n = 41): drug and dosage not specifie Topical acne therapy (n = 59): drug and dosage not specified	Change in scores of Zung Depression Status Inventory from baseline in both, intervention and control groups, within 25 to 35 days of the start of acne treatment   Percentage of participants having a CES‐D (Center for Epidemiologic Studies Depression Scale) score greater than 15 (indicative of clinically significant depression) at baseline and follow‐up Incidence of newly detected depression (CES‐D > 15) at follow‐up in both treatment and control groups Interviews for assessment of outcomes were 'in person' at baseline, but by telephone call at follow‐up, within 25 to 35 days of the start of therapy	There was no statistically significant difference in Zung scores between both time points of assessment (mean difference 0.00 points; 95% CI ‐0.6 to 0.6). Also there were no significant differences between groups regarding to the mean change in Zung score (one way ANOVA; df = 1, F = 1.4, P = 0.24) There was a single participant depressed at baseline (who was not depressed on follow‐up) and two on follow‐up, on the CES‐D score, both in the isotretinoin group. Despite this, incidence of newly detected depression was not statistically significant different between isotretinoin and control groups (Fisher' exact test, P = 0.497). Both new incident depression cases were female and had moderate to severe acne. None of them presented past history of depression at baseline assessment	Selection bias: High risk Comment: The study controlled some potential confounding variables and performed a linear regression analysis to adjust the comparison between groups for acne severity, the main potential confounding factor according to authors. However, duration of acne, another important confounding factor for the analysis of the association between oral isotretinoin use for acne and depression, was not even cited within the report Performance bias (all outcomes): High risk Comment: The study had an open design Detection bias (all outcomes): High risk Comment: The study had an open design Attrition bias: (all outcomes): Unclear risk Comment: No information regarding loss to follow‐up was provided Selective reporting bias: High risk Comment: No protocol available; however, there was an adequate report of outcomes listed in methods section Other bias: High risk Comment: The CES‐D is a highly sensitive screening tool, which could have overestimated the degree of mood disorder compared with other psychiatric instruments and have exaggerated the adverse effects estimates of the interventions in this study. Besides this, the length of the follow‐up period was inadequate, since treatment with oral isotretinoin seldom lasts more than 8 weeks. Also, instruments used to assess outcomes were different between both time points of measurement of the study: a clinical interview at baseline, and a telephone interview at follow‐up
Jick 2000	Design: Retrospective population based‐cohort (record linkage)   Duration of the study: not clearly provided   Periods of recruitment: not provided for United Kingdom cohort; from 1983 to 1997,  Saskatchewan, Canada cohort	Settings: two different centres: province of Saskatchewan, Canada (Canadian Saskatchewan Health Database‐CSHD) and United Kingdom (United Kingdom General Practice Research Database‐UKGPRD)   Saskatchewan cohort: 7195 oral isotretinoin users; 53% male; 79% aged 10 to 29 years 13,700 antibiotic users; 43% male; 75% aged 10 to 29 years   United Kingdom cohort: 340 oral isotretinoin users; 58% male; 78% aged 10 to 29 years 676 oral antibiotic users; 59% male; 78% aged 10 to 29 years   Duration of acne: not provided   Acne severity: not provided   Inclusion criteria: For both groups, had between 6 months and 5 years of computer‐recorded history before, and at least 12 months after, their first isotretinoin or antibiotic prescription for acne, except where death occurred For antibiotics users group, subjects with acne in whom the condition was diagnosed within the same period (± 6 months) and who fell within the same as isotretinoin‐treated participants, but who were treated with only antibiotic therapies, received within 30 days following the acne diagnosis Exclusion criteria: not provided	Current exposure to oral isotretinoin or to antibiotic therapies (tetracycline, erythromycin, clindamycin, minocycline or doxycycline): from the first prescription for acne treatment through 3 months after receiving the last study drug prescription   Recent exposure to oral isotretinoin or to antibiotic therapies (tetracycline, erythromycin, clindamycin, minocycline or doxycycline):  having received the last study drug prescription 4 to 6 months previously   Nonexposed: All other times after finishing treatment with the analised drug (oral isotretinoin or antibiotic)   Detailed information regarding dose and duration of exposures/ interventions were not provided	Prevalence rates of newly diagnosed depression or psychosis (neurotic and psychotic disorders), having person‐time as denominator and the number of cases as numerator, assessed by linkage with medical records of diagnosis codes in each computerised database   Incidence rates of suicide or attempted suicide, having person‐time as denominator and the number of cases of the outcome as numerator, assessed by linkage with medical records of diagnosis codes in each database	Saskatchewan database/cohort: The adjusted relative risk (ARR) estimate for newly diagnosed depression or psychosis was 1.0 (95% CI, 0.7 to 1.3) for current isotretinoin exposure compared with the nonexposed period.  The ARR estimates for recent isotretinoin use, current antibiotic use, and recent antibiotic use in participants with acne were 0.9, 1.3, and 0.9, respectively. There was no statistically difference between each one of both exposure group and nonexposed group For suicide and attempted suicide, the ARR estimate was 0.9 (95% CI, 0.3 to 2.4) when comparing current isotretinoin exposure with the nonexposed period. Also there was no statistically significant difference between each one of the three exposure groups and the nonexposed group.  History of psychiatric disorder was associated with higher suicide risk during both isotretinoin and antibiotics therapy from CSHD cohort (ARR 8.0, CI 4.1 to 15.5)   GPRD database: Current isotretinoin use yielded an ARR estimate of 1.8 (95% CI, 0.4 to 5.2) for newly diagnosed depression or psychosis compared with the nonexposed period. The ARR estimates were 1.8, 1.5, and 1.7 for recent isotretinoin use, current antibiotic use, and recent antibiotic use, respectively, and there was no statistically significant difference between each one of these exposure groups compared with the nonexposed group There was only one suicide attempt in the GPRD. The subject was nonexposed   In both cohorts (CSHD and UKGPRD): There was no difference between rates of newly diagnosed depression or psychosis in the first 6 months of isotretinoin use and the rates before therapy. This was an analysis of data from isotretinoin users only (a before‐and‐after comparison) Female gender and older age had statistically significant associations with newly diagnosed depression or psychosis	Selection bias: High risk Comment: The study had considered in the analysis some important confounding factors (i.e. age, gender, history of psychiatric disorder) and used multiple logistic regression models. However, there was no report of data related to acne severity, which may be an independent predictor of depression, apart from drug exposure Performance bias (all outcomes): High risk Comment: As the study was a retrospective cohort, probably personnel and participants were not blinded regarding the exposure/drug Detection bias (all outcomes): Unclear risk Comment: The study did not provide any information regarding blinding of the outcome assessors Attrition bias (all outcomes): Low risk Comment: The study had used hazard ratios to measure outcomes (incidence density), which minimise bias due to missing data in both open retrospective cohorts reported Selective reporting outcome: High risk Comment: The study did not provide RR estimates for isotretinoin exposure compared with antibiotic exposure, despite having cited this outcome measurement in methods section Other bias: High risk Comment: The study might be at risk of inappropriate influence of funders, as it was sponsored and promoted by a pharmaceutical company. There was a potential under‐ascertainment of psychiatric outcomes, as researchers only performed record linkage with diagnosis codes, and not psychoactive drug prescriptions or interviews. The study used a potentially insensitive instrument to measure outcomes, record linkage across electronic databases. Also, the study had presented only a cohort cross‐over analysis of data for each one of both cohorts (CSHD and UKGPRD), since the nonexposed group in each cohort included people who developed one of the psychiatric outcomes assessed 6 months after being exposed either to oral isotretinoin or to antibiotic. This design had a potential risk of bias related to the possibility of carry‐over effects in participants of nonexposed groups in both cohorts who had received oral isotretinoin previously
Kaymak 2009	Design: Prospective nonrandomised controlled trial Duration of the study: not clearly provided Period of recruitment: September 2006 to May 2007 Factors influencing the choice of treatment: previous treatment failure, the severity and duration of the acne, participant preference, and participant concerns about adverse drug effects	Setting: One outpatient dermatology clinic of a university health centre Isotretinoin (study group): n = 37 (11 males and 25 females) Topical treatment (control group): n = 41 (9 males and 20 females) Age ‐ mean ± SD (years): Isotretinoin group: 20.61 ± 1.87 Topical treatment group: 20.51 ± 2.01 Duration of acne mean ± SD (years): Isotretinoin group: 4.86 ± 2.81 Topical treatment group: 3.93 ± 2.86 Acne severity: participants with severe, moderate or mild acne in both groups. The report did not provide numbers of participants with each one grade of acne severity for each group Inclusion criteria: not provided Exclusion criteria: not provided	Isotretinoin group: 0.5 – 0.8 mg/kg/day of oral isotretinoin with food in two divided doses for at least 20 weeks; cumulative dose of 100 mg/kg Topical treatment group: either topical antibiotics or topical retinoids. More detailed information regarding this intervention group was not provided	Frequency of psychopathology symptoms assessed by two instruments: The hospital anxiety and depression (HAD), a self‐rating scale used to assess the risk and to measure the level of depression and anxiety. It contains 14 questions, 7 related to depression (HAD‐D) and 7 to anxiety (HAD‐A). A validated Turkish version of the full scale (HAD‐T) was also used The Beck depression inventory (BDI), a self‐rating scale composed of 21 items, which evaluates the level of depression. Each question is rated from 0 to 3 points. The sum of the points shows the level of depression. A BDI score over 13 was the index for establishing depression diagnosis Change in mean scores of BDI, HAD‐D, HAD‐A and HAD‐T during treatment compared to baseline in both treatment groups Measurements occurred at baseline, 2, and 4 months of follow‐up	Significantly more participants had BDI scores over 13 (considered as depressive) in the topical treatment group (10/29) compared to the isotretinoin group (4/36) at 4 months time point (P = 0.03) There were also statistically significant differences between the two treatments groups, in favour of isotretinoin, regarding the number of participants with measurements of HAD‐D higher than the range of clinically significant depression (> 8) at 2 (P = 0.01) and 4 (P = 0.02) months	Selection bias: Low risk Comment: The study performed an adjusted analysis data for the most important confounding factors while assessing psychiatric outcomes related to oral isotretinoin for acne Performance bias (all outcomes): Unclear risk Comment: The study did not provide any information regarding blinding of participants and personnel Detection bias (all outcomes): Low risk Comment: The author who assessed participant’s psychological status was blinded Attrition bias (all outcomes): High risk Comment: The level of loss to follow‐up in the study could lead to attrition bias, especially considering that there was a high imbalance of missing data between the two groups. The trial reported a 'per‐protocol' analysis, which did not consider data from loss to follow‐up Selective reporting bias: Low risk Comment: no protocol available; however, there was an adequate report of outcomes listed in methods section Other bias:Low risk Comment: There were no other apparent sources of bias
McGrath 2010	Design: Prospective nonrandomised controlled trial Duration of the study: not clearly provided Period of recruitment: between September 2006 and September 2007	Setting: treatment groups: Royal United Hospital, Bath, UK; control group: matched community sample from: two coeducational schools in South West England; a University campus; and venues at a city centre Gender: Isotretinoin group: n = 65 (45 males and 20 females) Oral antibiotic and a topical retinoid group: n = 31 (16 males and 15 females) Control group: n = 94 (50 males and 44 females) Age ‐ mean ± SD (years): Isotretinoin group: 19.8 ± 3.8 Oral antibiotic and a topical retinoid group: 19.3 ± 3.9 Control group: 19.7 (standard deviation not provided) Duration of acne (years): not provided Acne severity: not clearly specified Inclusion criteria: treatment groups: outpatients commencing treatment for acne at the Royal United Hospital, Bath, UK control group:matched healthy community participants from three sources: two coeducational schools in South West England (children aged 14‐18 years); adults (aged 18‐30 years) from a University campus; and opportunist sampling at venues in a city centre Exclusion criteria: treatment groups: being under 12 years or over 50 years; pregnant or lactating women; control group: being under 14 years or over 30 years old, and currently under hospital treatment for acne	Isotretinoin group: 0.5 mg/kg/daily for the first 2 weeks, followed by 1 mg/kg/daily until reaching the cumulative dose of 120 mg/kg (participants who did not tolerate the 1 mg/kg/day dose received the next highest dose possible) Oral antibiotic and a topical retinoid group: lymecycline 408 mg daily (or minocycline 100 mg daily in the case of intolerance or inefficacy with previous lymecycline use)plus adapalene cream	Occurrence of depression assessed by the Centre for Epidemiological Studies Depression Scale (CES‐D) Measurements occurred at baseline, then subsequently at 3 and 6 months, for all participants in both treatment groups. Participants in the matched community control group completed the WHOQOL‐BREF, Centre for Epidemiological Studies Depression Scale (CES‐D) and visual analogue score only once	Depression scores over time did not present significant changes in treatment groups (F1.64 = 1.06, not significant) and there was no interaction between gender and changes in depression levels (authors included gender as a covariate in the measurements of depression, since scores were higher in women than in men at baseline)	Selection bias: High risk Comment: Despite the analysis of covariance (ANCOVA), controlling for the covariate gender, and the matched healthy control group, the study did not consider other important confounders, such as duration of acne, in the comparison over time between both acne treatment groups Performance bias (all outcomes): High risk Comment: Participants and personnel were not blinded Detection bias (all outcomes): Unclear risk Comment: The study did not provide any information regarding blinding of the outcome assessors Attrition bias (all outcomes): High risk Comment: Authors reported a high level of loss to follow‐up: only 43.7% of participants from treatment groups provided useable data on all three time points of measurements Selective reporting: High risk Comment: Due to the high level of loss to follow‐up, authors carried out data analysis only for the first two time points according to the methods section Other bias:Low risk Comment: There were no other apparent sources of bias
Ng 2002	Design: Prospective nonrandomised controlled trial Duration of the study: not clearly provided Period of recruitment: between December 1998 and March 2000	Setting: two private dermatology clinics and a public hospital outpatient dermatology clinic in Melbourne, Australia Isotretinoin group: n = 174 (58.6% male) Antibiotic plus topical treatment group: n = 41 (41.5% male) Age – mean ± SD (years)/range (years): Isotretinoin group: 20.2 ± 5.8/15‐46 Antibiotic plus topical treatment group: 20.7 ± 5.8/15 – 38 Duration of acne: not provided Acne severity: participants with severe or moderate acne in both groups. The report did not provide numbers of participants with each one of the these two grades of acne severity for each treatment group Inclusion criteria: participants between 15 and 50 years having moderate to severe acne, and able to comprehend the rating instructions and comply with the study protocol Exclusion criteria: current diagnosis of depression; concomitant use of antidepressants, corticosteroids, anabolic steroids or other depression‐inducing medications; pregnancy or breastfeeding	Isotretinoin group (n = 174): oral isotretinoin starting at 40 mg/day, increased to a dose of 1.0 mg/kg/day over 1 month according to tolerability, and continued for a total cumulative dose of 120 mg/kg (over 5–6 months) Antibiotic plus topical treatment group: a standard course of minocycline 100–200 mg/day, titrated according to weight, response and tolerance, and topical treatment consisted of adapalene 0.1% gel, tretinoin 0.05% cream or isotretinoin 0.05% gel	Changes in mean Beck Depression Inventory (BDI) scores during treatment compared to baseline in both groups Percentage of participants having a BDI score of 10 or greater (indicative of at least a moderate level of depressive symptoms) Rate of withdrawal from the study because of worsening of mood Measurements of outcomes occurred at baseline, 1 month, 3 months and at the end of treatment course or 6 months	Changes in mean BDI scores over the therapy course were not significantly different between both treatment groups (P = 0.62) 18 participants from the whole sample had a BDI score of 10 or greater at baseline; the incidence of moderate depressive symptoms in the isotretinoin group remained relatively unchanged during the study and at lower levels than in the antibiotic/topical group 5 participants in the isotretinoin group (n = 174) had dropped out from the study due to depressed mood, 3 males and 2 females. Only in 2 (both male) of these participants, the relationship to isotretinoin use seemed possible. No participant was withdrawn from the antibiotic plus topical group (n = 41) because of mood changes	Selection bias: High risk Comment: Despite having controlled potential confounding variables and made a linear regression analysis to adjust the comparison between groups, not all important confounding factors were explored by the study analysis. Duration of acne and previous duration of acne treatment were not reported. Performance bias (all outcomes): High risk Comment: The study had an open design Detection bias (all outcomes): High risk Comment: The study had an open design Attrition bias: (all outcomes): High risk Comment: Despite the intention‐to‐treat analysis of data, there was an imbalance in dropout rates between the two groups Selective reporting bias: Low risk Comment: No protocol available; however, there was an adequate report of outcomes listed in methods section Other bias: Low risk Comment: There were no other apparent sources of bias
Sundstrom 2010	Design: Retrospective population‐based cohort Duration of the study: not clearly provided Period of recruitment: between 1980 and 1990	Setting: Sweden 5756 participants (3613 males and 2143 females) who were aged 15 to 49 years Duration of acne: not provided Acne severity: severe acne Inclusion criteria: having had at least one course of oral isotretinoin granted by the Medical Products Agency in Sweden during the period of recruitment Exclusion criteria: not provided	Oral isotretinoin administered in a mean dose of 44.5 mg (SD 15.7) for males and 39.2 mg (SD 13.1) for females Duration of intervention: a mean period of 6.0 months (SD 4.0) for males and 6.1 (SD 3.9) for females	Rates of attempted and completed suicides in the different time windows of the cohort compared between themselves (before, during, and after treatment with oral isotretinoin) and with those of the general population, assessed by searching for related events between 1980 and 2001 in the national patient register of in‐hospital care and in the cause of death register	Standardised incidence ratios (comparing the study cohort with the general population) for first suicide attempts and for all attempts rose, respectively, from 0.89 (95% CI, 0.54 to 1.37) and 0.99 (95% CI, 0.65 to 1.44) three years before treatment to 1.36 (95% CI, 0.65 to 2.50) and 1.57 (95% CI, 0.86 to 2.63) in the year preceding treatment. Both ratios were highest within six months after beginning treatment: 1.93 (95% CI, 1.08 to 3.18) for first attempts and 1.78 (95% CI, 1.04 to 2.85) for all attempts. Within three years after treatment, the number of suicide attempts in the cohort was almost the same as the number observed in general population, and standardised incidence ratios were 0.97 (95% CI, 0.64 to 1.40) for first attempts and 1.04 (95% CI, 0.74 to 1.43) for all attempts. However, there was a significant increase in the standardised incidence ratio of repeated events (but not of first attempts) 11 years after treatment among female participants of the cohort: 1.36 (95% CI, 1.06 to 1.70) Females who attempted suicide after completing treatment received significantly more than one course in comparison to female participants of the cohort who never attempted treatment There was a statistically significant difference in the chance of committing another suicide attempt between participants who had their first attempt before treatment and those who made a first suicide attempt during treatment or within six months after the finish of the course Treatment with oral isotretinoin reinforced more significantly the suicidal behaviour for participants who committed the first suicide attempt during treatment, or within six months after the end, in comparison with those who first attempted before using oral isotretinoin There was an increase in rates of attempted suicide per person‐years of follow‐up within the isotretinoin Swedish cohort (cohort cross‐over analysis) from time points of measurement before and after the treatment. For first attempts, the rate difference between the year before treatment and six months following the end of treatment was the highest: 0.86 cases per 1000 person‐years (95% CI, ‐0.78 to 2.50); for all attempts; this rate difference was 0.40 (95% CI, ‐1.40 to 2.26) per 1000 person‐years. The number needed to treat for an additional harmful outcome (first suicide attempt and one additional repeated attempt) was 2300 and 5000 per year, respectively 24 participants had death by suicide (17 males and 7 females) by the end of 2001. The standardised mortality ratio for males who committed suicide within one year after treatment was 1.9 (95% CI, 0.4 to 5.4), and decreased to around one within two years after treatment. For female participants, the highest standardised mortality ratio was 1.8 (95% CI, 0.7 to 3.9), which occurred within 11 years after the treatment	Selection bias: High risk Comment: Authors considered only age, gender, and calendar year as potential confounding factors while calculating specific rates in the general population control group. The study did not take in account the fact that presence of acne and its severity in the control group (general population) could be a potential confounder Performance bias (all outcomes): High risk Comment: As the study was a retrospective cohort, probably personnel and participants were not blinded for the exposure/drug Detection bias (all outcomes): Unclear risk Comment: The study did not provide any information regarding blinding of the outcome assessors Attrition bias (all outcomes): Unclear risk Comment: There was no information about missing data or the potential for data to be missing to permit judgment. Selective reporting bias: Low risk Comment: no protocol available; however, there was an adequate report of outcomes listed in methods section Other bias: High risk Comment: The study used a potentially insensitive instrument to measure outcomes, record linkage across electronic databases. Besides this, the exclusion of the outpatients database from the estimate of rates of attempted suicide might had underestimated the primary outcome measurement in all the cohort. Also, authors reported only a single intervention group (there was no comparison within two or more interventions in the cohort), and analysis of data had an internal cross‐over pattern, a poorer design if compared with the classical cohort

^{ANCOVA: analysis of covariance
 ARR: adjusted relative risk
 BDI: Beck depression inventory
 CES‐D: The Center for Epidemiologic Studies Depression Scale
 CSHD: Canadian Saskatchewan Health Database
 df: degrees of freedom
 DSM‐IV: Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition
 HAD‐A: hospital anxiety and depression ‐ anxiety
 HAD‐D: hospital anxiety and depression ‐ depression
 HAD‐T: hospital anxiety and depression ‐ Turkish version
 RAMQ: Régie de l' Assurance Maladie du Québec
 SD: standard deviation
 UKGPRD: United Kingdom (United Kingdom General Practice Research Database)
 WHOQOL‐BREF: World Health Organization Quality of Life‐Brief version}