eAppendix 1. ePRISMA Checklist and MOOSE Checklist
eAppendix 2. Search Strings for PubMed
eTable 1. Articles Excluded After Full-Text Revision, With Reasons
eTable 2. Characteristics of Meta-analyses of Observational Studies Studying the Association Between Antidepressants and Risk of Any adverse Health Outcome
eTable 3. Suggestive Evidence (Class III) for the Association of Antidepressant Use and Risk of Adverse Health Outcomes in Meta-analyses of Observational Studies
eTable 4. Weak Evidence (Class IV) for the Association of Antidepressant Use and Risk of Adverse Health Outcomes in Meta-analyses of Observational Studies
eTable 5. No Evidence (Nonsignificant Associations, P > 0.05) for the Association of Antidepressant Use and Risk of Adverse Health Outcomes in Meta-analyses of Observational Studies
eTable 6. Sensitivity Analysis Results of Nonselected Meta-analyses Due to Overlap
eTable 7. List of Covariates Used for the Sensitivity Analysis Limited to Studies Adjusted for Covariates
eMethods. Supplementary Methods
eResults. Supplementary Results
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Dragioti E, Solmi M, Favaro A, et al. Association of Antidepressant Use With Adverse Health Outcomes: A Systematic Umbrella Review. JAMA Psychiatry. 2019;76(12):1241–1255. doi:10.1001/jamapsychiatry.2019.2859
Is antidepressant use associated with adverse health outcomes, and how credible is the evidence behind this association in published meta-analyses of real-world data?
In this systematic umbrella review of 45 meta-analyses of observational studies, convincing evidence was found for the associations between antidepressant use and suicide attempt or completion among individuals younger than 19 years and between antidepressant use and autism risk among the offspring. However, none of these associations remained at the convincing evidence level after a sensitivity analysis that adjusted for confounding by indication.
This study’s findings suggest that claimed adverse health outcomes associated with antidepressants may not be supported by strong evidence and may be exaggerated by confounding by indication; no absolute contraindication to the use of antidepressants was found to be currently supported by convincing evidence.
Antidepressant use is increasing worldwide. Yet, contrasting evidence on the safety of antidepressants is available from meta-analyses, and the credibility of these findings has not been quantified.
To grade the evidence from published meta-analyses of observational studies that assessed the association between antidepressant use or exposure and adverse health outcomes.
PubMed, Scopus, and PsycINFO were searched from database inception to April 5, 2019.
Only meta-analyses of observational studies with a cohort or case-control study design were eligible. Two independent reviewers recorded the data and assessed the methodological quality of the included meta-analyses. Evidence of association was ranked according to established criteria as follows: convincing, highly suggestive, suggestive, weak, or not significant.
Forty-five meta-analyses (17.9%) from 4471 studies identified and 252 full-text articles scrutinized were selected that described 120 associations, including data from 1012 individual effect size estimates. Seventy-four (61.7%) of the 120 associations were nominally statistically significant at P ≤ .05 using random-effects models. Fifty-two associations (43.4%) had large heterogeneity (I2 > 50%), whereas small-study effects were found for 17 associations (14.2%) and excess significance bias was found for 9 associations (7.5%). Convincing evidence emerged from both main and sensitivity analyses for the association between antidepressant use and risk of suicide attempt or completion among children and adolescents, autism spectrum disorders with antidepressant exposure before and during pregnancy, preterm birth, and low Apgar scores. None of these associations remained supported by convincing evidence after sensitivity analysis, which adjusted for confounding by indication.
Conclusions and Relevance
This study’s findings suggest that most putative adverse health outcomes associated with antidepressant use may not be supported by convincing evidence, and confounding by indication may alter the few associations with convincing evidence. Antidepressant use appears to be safe for the treatment of psychiatric disorders, but more studies matching for underlying disease are needed to clarify the degree of confounding by indication and other biases. No absolute contraindication to antidepressants emerged from this umbrella review.
Accumulating evidence suggests a sharp growth in antidepressant use worldwide. Up to 8% to 10% of adults in the United States take at least 1 antidepressant drug, which is ranked third among prescribed and fourth among sold medications.1,2 Antidepressants are indicated and used for depressive disorders, anxiety disorders, posttraumatic stress disorder, premenstrual dysphoric disorder, obsessive-compulsive disorder, bulimia nervosa, and binge-eating disorder, among others.3-5
The safety profile of antidepressants is controversial. Since the US Food and Drug Administration introduced the black box warnings that associated selective serotonin reuptake inhibitor (SSRI) use with a higher risk of suicidal behavior in children and adolescents,6 the debate about the efficacy, acceptability, and safety profile of antidepressant medications has gradually increased.7-11 Evidence from randomized clinical trials (RCTs) of antidepressants’ efficacy and acceptability has been well documented in both meta-analyses and network meta-analyses,4,8,10,12,13 but safety assessment is inherently biased by certain methodological weaknesses of RCTs. These weaknesses include small and unrepresentative samples, rare and inconsistent reporting of adverse outcomes, and short duration of exposures.14,15
Observational studies complement RCTs by providing evidence with real-world data15 on a number of adverse health outcomes associated with antidepressants, which is not possible in RCTs.16 For example, observational studies can show medication safety because they include representatives of the overall target population, such as patients with comorbid disorders or suicidal thoughts who are often excluded from RCTs. In addition, observational studies typically have a longer follow-up duration compared with RCTs, providing data on the mid- or long-term consequences of antidepressants, such as poor bone status or gastrointestinal bleeding, that may not arise from short-term use.16
Several meta-analyses of observational studies have been published that assess antidepressant safety; however, to our knowledge, no attempt has been made to quantify the credibility of their findings. This quantification is crucial considering the uncertainty surrounding observational research results.17-19 Umbrella reviews make it feasible to summarize the evidence from multiple meta-analyses on the same topic20,21 and enable the ranking of evidence (as convincing, highly suggestive, suggestive, weak, or not significant) according to sample size, strength of the association, and assessment of presence of biases.22-24
In this umbrella review, we graded the evidence from published meta-analyses of observational studies. These studies tested the association between antidepressant use and risk of adverse health outcomes.
The protocol for this study was registered on PROSPERO (CRD42018103462). We followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline25 and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines26 (eAppendix 1 in the Supplement).
We searched PubMed, Scopus, and PsycINFO from database inception to April 5, 2019, to identify systematic reviews with meta-analysis of observational studies of the association between any adverse health outcome and exposure to antidepressants. Our search strategy used a combination of terms related to antidepressants (eg, antidepressants, selective serotonin reuptake inhibitors), to adverse health outcomes (eg, harms, suicide, bleeding, and autism), and to meta-analysis with no age, sex, population, and medical condition restrictions (eAppendix 2 in the Supplement). We also manually searched the cited references of the retrieved articles and reviews.
Two of us (E.D. and M.S.) independently searched titles or abstracts for eligibility and consulted a third reviewer (E.E.) when we could not reach a consensus. The full texts of potentially eligible articles were retrieved, and the same two of us (E.D. and M.S.) independently scrutinized each study for eligibility. Any discrepancies during this process were resolved by our third reviewer (E.E.).
We included only peer-reviewed systematic reviews with meta-analysis of observational studies with a cohort, case-control, or nested case-control study design measuring any association between antidepressant use and any adverse health outcome in any population of any age. Whenever multiple meta-analyses on the same adverse health outcome were performed (ie, overlapping meta-analyses with the same outcome, type of antidepressant used, and clinical or population setting), we assessed only the one that included the largest data set, as previously described.22,24,27 Details of the selection between overlapping meta-analyses are described in the eMethods in the Supplement. For each eligible meta-analysis, we considered the main analysis for all primary and secondary reported outcomes. The concordance between selected and nonselected meta-analyses was examined in a sensitivity analysis.27
We excluded (1) meta-analyses of studies with other study designs (eg, RCTs, cross-sectional) or that included both observational studies and RCTs in the same analysis; (2) meta-analyses published in languages other than English; (3) meta-analyses of individual patient or participant data, pooled analyses of a nonsystematic selection of observational studies, and nonsystematic reviews; (4) meta-analyses of St John’s wort (Hypericum perforatum) or tryptophan; and (5) meta-analyses that provided insufficient or inadequate data for quantitative synthesis.
Two of us (E.D. and M.S.) independently performed data extraction, and disagreements were resolved by a consensus. Adverse health outcomes associated with exposure to antidepressants were extracted as defined by the original authors. For each meta-analysis, we recorded the standard identifier (PMID and DOI), first author, publication year, type of antidepressant, study design, age of participants, adverse health outcomes, exposure and nonexposure, illnesses examined (eg, depression), number of included studies, and total sample size.
For each primary study, we recorded first author; year of publication; study design (ie, cohort or case-control); number of cases and controls in case-control studies or total population in cohort studies; reported adjusted (or unadjusted) effect size (ie, relative risk, odds ratio, hazard ratio, and standardized mean difference), each with a 95% CI; and study location. We also captured the number and nature of adjustments, the length of follow-up, the study quality score, and whether the studies were controlled for a psychiatric condition (ie, confounding by indication).18,19
The methodological quality of each included meta-analysis was assessed by 2 of us (E.D. and M.S.) using the updated AMSTAR (A Measurement Tool to Assess Systematic Reviews) 2.28 AMSTAR 2 also accounts for the quality of studies included in the meta-analysis beyond a mere technical methodological assessment of the included meta-analysis (eMethods in the Supplement).16
For each association, we extracted effect sizes of individual studies included in each meta-analysis, and we repeated the meta-analyses to calculate the pooled effect sizes and the 95% CIs using random-effects models to compare homogeneously analyzed results.29 We did not transform the initial effect sizes or modify the direction of associations presented by the original authors to compare the results we obtained with the reported results in the meta-analyses. Heterogeneity was assessed with the I2 statistic.30 In addition, we calculated the 95% prediction intervals for the summary random effect sizes, providing the possible range in which the effect sizes of future studies were expected to fall.31
Next, we tested whether smaller studies yielded larger effect sizes compared with larger studies, an indication of small-study effect bias.24,32-34 Small-study effect bias was indicated both by the Egger regression asymmetry test (P ≤ .10) and by the random-effects summary effect size being larger than that of the biggest study in each association.24,32-34
We then assessed the existence of excess significance bias by evaluating whether the observed number of studies with nominally statistically significant results (positive studies as indicated with a 1-sided P ≤ .05) was different from the expected number of studies with statistically significant results.34 The expected number of statistically significant studies per association was calculated by summing the statistical power estimates for each component study. The power estimates of each component study depend on the plausible effect size for the tested association, which we assumed to be the effect size of the largest study (ie, the smallest SE) per association.35 Excess significance bias was set at P ≤ .10. This test was designed to assess whether the published meta-analyses comprised an overrepresentation of false-positive findings.34 All analyses were conducted in Stata/MP, version 10.0 (StataCorp LLC).
We assessed the credibility of the evidence per association provided in meta-analyses by applying several criteria in concordance with previously published umbrella reviews.22,23,32,33,36,37 In brief, associations that presented nominally significant random-effects summary effect sizes (ie, P ≤ .05) were ranked as convincing, highly suggestive, suggestive, or weak evidence according to sample size, strength of the association, and assessment of the presence of biases (Table 1 and eMethods in the Supplement). In addition, to provide an estimate of the epidemiologic implication of findings, we calculated the prevalence of outcomes of interest from cohort studies only (studies with case-control designs should not be considered for prevalence estimates).
We performed sensitivity analyses to assess whether the credibility of the evidence varied within both prospective and retrospective cohort studies, prospective cohort studies, studies adjusted for multiple covariates and for confounding by indication, high-quality primary studies, studies of antidepressant classes (SSRIs, tricyclic antidepressants [TCAs], and other or mixed antidepressants), and locations where studies were conducted (Europe, North America, or other regions). These analyses were performed only for the associations ranked as convincing evidence or highly suggestive evidence (ie, class I or II) in the main analysis.
In total, we identified 4471 studies, scrutinized 252 full-text articles, and ultimately included 45 meta-analyses (17.9%) in this umbrella review38-83 (Figure 1), corresponding to 695 studies, 1012 study estimates, and 13 putative risks (Figure 2). The 207 excluded articles (82.1%) and the reasons for their exclusion are provided in eTable 1 in the Supplement.
Descriptive characteristics of the 45 eligible meta-analyses of observational studies can be found in eTable 2 in the Supplement. All meta-analyses had a control group that was not exposed to antidepressants except for 1 (2.2%), which compared the risk of gastrointestinal bleeding between mirtazapine and SSRIs.47 The median number of adjustments in the analyses was 7 (interquartile range [IQR], 4-11), and the median duration of follow-up was 4 (IQR, 2-5) years.
Thirty-three meta-analyses (73.4%) met the moderate-quality level according to the AMSTAR 2 evaluation, and 8 (17.8%) were of low quality. Two (4.4%) were high quality, whereas 2 others (4.4%) were of critically low quality. The 2 of us (E.D. and M.S.) reached a high level of agreement (91%) on the quality rating.
Forty-five eligible meta-analyses described 120 associations, including 1012 individual study estimates of adverse health outcomes associated with exposure to antidepressants (Table 2 and eTables 2-5 in the Supplement), with a median (IQR) number of estimates per association of 6 (4-12). Seventy-four (61.7%) of the associations concerned maternal and pregnancy-related adverse health outcomes (Figure 2). Most associations (80 [66.7%]) concerned SSRIs or serotonin-norepinephrine reuptake inhibitors, 9 (7.5%) TCAs, and 31 (25.8%) mixed or other antidepressants.
The median (IQR) number of the total population per association was 1 056 374 (152 180-2 215 969). The median (IQR) number of cases (adverse health outcomes) per association was 12 097 (2585-56 272), and the number of cases was greater than 1000 for 87 associations (72.5%).
A summary of all 120 associations is presented in Table 2 and Table 3 and eTables 3-5 in the Supplement. Seventy-four of the 120 examined associations (61.7%) were nominally statistically significant at P ≤ .05 based on random-effects models, and only 22 (18.3%) reached a P ≤ 1 × 10−6. Almost all statistically significant associations indicated an increased risk for antidepressants and adverse health outcomes except for 2 associations (2.7%) showing the protective property of SSRIs against suicide attempt or completion in adults and in older adults.80
Fifty-two associations (43.3%) had large heterogeneity (I2 > 50%), and the 95% prediction intervals excluded the null value for only 24 associations (20.0%). In 63 associations (52.5%), the effect sizes of the largest study were nominally statistically significant at P ≤ .05. Small-study effects were found for 17 associations (14.2%), and excess significance bias was observed for 9 associations (7.5%).
Among the 120 associations, 3 (2.5%) were supported by convincing evidence, namely, the association between SSRI use and increased risk of suicide attempt or completion in children and adolescents80 as well as the association between exposure to any antidepressant before pregnancy and SSRIs during pregnancy and autism spectrum disorder51,52 (Table 2). The association with suicide risk reached the high-quality level based on AMSTAR 2, whereas the 2 associations with autism spectrum disorder reached moderate quality.
Eleven associations (9.2%) had highly suggestive evidence of the association between any antidepressant use and increased risk of adverse health outcomes (Table 2). The adverse outcomes were attention-deficit/hyperactivity disorder in children, cataract development (associated with TCAs), severe bleeding at any site, upper gastrointestinal tract bleeding, postpartum hemorrhage, preterm birth, lower Apgar score at 5 minutes, osteoporotic fractures (1 associated with TCAs and 1 with SSRIs), and risk of hip fracture. Seven of these associations reached the moderate-quality level based on AMSTAR 2 (Table 2). One association with highly suggestive evidence, however, showed a decreased risk (ie, protective association) of suicide attempt or completion in adults,80 meeting a high-quality level based on AMSTAR 2. The effect sizes of those adverse outcomes supported by convincing and highly suggestive evidence were small and the prevalence was on average low (range, 0.1%-9.7%) as well (Tables 2 and 3).
Suggestive evidence was found for 21 additional associations (17.5%) between antidepressant use and increased risk of adverse health outcomes (eTable 3 in the Supplement). For the remaining associations, either weak evidence (n = 39 [32.5%]) or no evidence (n = 46 [38.3%] was found (ie, all associations with P > .05) (eTables 4 and 5 in the Supplement).
A sensitivity analysis limited to cohort studies, prospective cohort studies, studies controlled for confounding by indication, and North American studies showed that none of the associations within convincing evidence (class I) retained the same rank (Table 3). The most important change was within prospective cohort studies, with 1 association being upgraded to having convincing evidence (preterm birth associated with the use of any antidepressant).
Another association was upgraded to having convincing evidence (lower Apgar scores at 5 minutes) when the sensitivity analysis was limited to SSRIs. The association between antidepressant use and preterm birth was also upgraded to being supported by convincing evidence when the analysis was limited to other or mixed antidepressants (Table 3).
Findings from another sensitivity analysis, limited to excluded meta-analyses owing to overlap, agreed with the results of the main analysis (eResults and eTable 6 in the Supplement). The results of each sensitivity analysis are presented in the eResults in the Supplement, with the full list of covariates in eTable 7 in the Supplement.
We reviewed 45 meta-analyses of observational studies and found that only a few of the 74 statistically significant associations between antidepressants and adverse health outcomes were supported by convincing evidence in the main and sensitivity analyses, namely, the association between antidepressant use and increased suicide attempt or completion in individuals younger than 19 years (SSRI studies),80 autism risk in the offspring,51,52 preterm birth,66 and neonatal adaptation.71 However, the few with convincing evidence associations did not reflect causality, and none of them remained at the convincing evidence level after accounting for confounding by indication. Overall, the results showed that the association between antidepressant use and adverse health outcomes was not supported by robust evidence and that the underlying disease likely inflated the findings in a relevant way.39,44
To our knowledge, this study is the first umbrella review that systematically assessed the potential risk of adverse health outcomes associated with antidepressant use across a large spectrum of published meta-analyses of observational studies, grading the evidence by using well-recognized criteria of credibility.22,23,32,33,36,37 The umbrella review approach has been applied to assess the associations between adverse health outcomes and other medical variables, such as dietary fiber consumption,37 serum uric acid level,23 and vitamin D concentration.22 This approach fits in a research field that is undeniably complex and uncertain, as conveyed here.22,23,32,33,36,37 The large median number of participants and cases per association allowed for robust classifications; the number of cases was greater than 1000 for 87 of the 120 associations. Quality ratings of the included meta-analyses with AMSTAR 2 also allowed for the confident interpretation of the results. Sensitivity analyses provided additional evidence from the cohort studies, high-quality studies, and studies controlled for a psychiatric condition, thus further increasing the reliability of the results.
These results need to be considered when contemplating the use of antidepressants in children and adolescents or integrated with efficacy data from RCTs. A network meta-analysis of RCTs in children and adolescents showed that no antidepressant medication was superior to placebo apart from fluoxetine, that several antidepressants had higher discontinuation rates compared with placebo, and that venlafaxine increased the risk of suicidality even in the short-term duration of an RCT.13 However, although 1 single antidepressant, venlafaxin (odds ratio, 7.7), was associated with an increased risk of suicidality compared with placebo, none of the other SSRIs or antidepressants had an association. Not only did placebo have a substantially reduced risk of suicidality (87% lower) compared with venlafaxine, but the same was true (and with a similar degree) for 5 antidepressants (duloxetine, escitalopram, fluoxetine, imipramine, and paroxetine), with an 81% to 86% reduced risk compared with venlafaxine; according to the network meta-analysis, these antidepressants were safe with regard to suicidality as an adverse effect.13 Moreover, antidepressants’ lack of superiority over placebo,12 especially in children, was associated with a high placebo response, which has been an increasing problem in RCTs in psychiatry. In addition, the increased suicidality in children and adolescents who use antidepressants may be associated with the unsuccessful reduction of depressive symptoms in suicidal individuals rather than a direct result of antidepressant use. Furthermore, the results showed that confounding by indication probably contributes to the safety concerns of using these drugs in children and adolescents. Besides, the risk-benefit evaluation in children and adolescents is different for antidepressants (predominantly SSRIs) when used for psychiatric conditions, such as anxiety disorders and obsessive-compulsive disorder.3-5,12
Conversely, we found highly suggestive evidence supporting the protective role of antidepressants against suicidality in adults,80 which is consistent with results of a network meta-analysis of RCTs in adults that showed all antidepressants were superior to placebo in reducing depressive symptoms.10 Similarly, meta-analyses support the efficacy of antidepressant use for anxiety disorders5 and obsessive-compulsive disorder3 in adults. In adults, the risk-benefit ratio must account for clear efficacy of antidepressants and protection against suicide, which should be balanced with other safety concerns that emerged from the present umbrella review. Overall, several adverse outcomes associated with antidepressant use supported by highly suggestive evidence (ie, poor bone status, gastrointestinal tract bleeding) can be prevented medically, as previously reported.82 Hence, the advantages of antidepressant use in adults and older adults may well trump preventable safety issues given their efficacy in treating various psychiatric disorders.3-5,12 Moreover, the association between antidepressant use and certain adverse health outcomes varied within specific age groups. For instance, increased risk of fractures applied predominantly to an older population (>65 years) already prone to poor bone status and multimorbidity84 and not to people aged 20 to 40 years.
Convincing evidence, before accounting for confounding by indication, that supported the association between antidepressant use and autism, as well as other offspring adverse health outcomes, may call for the restriction of antidepressant use during pregnancy among women with a high risk of relapse and severe clinical presentations. Warnings to avoid prescribing medications in early pregnancy have been issued.85 However, autism remains a rare event, with a prevalence from cohort studies of less than 1% according to data pooled in this study. The convincing evidence level was not confirmed when confounding by indication was considered, suggesting that the association between antidepressant use and autism as well as suicidality in youth and other outcomes may be due to the underlying disease rather than to the use of antidepressants,39,43,44,50 as shown in a recent umbrella review on risk factors for autism.86
Comparing 2 depression-matched groups with or without antidepressant exposure may be more methodologically accurate than adjusting analyses statistically. Several adverse outcomes had small effect sizes in addition to low prevalence and no proof of a causal relationship between antidepressants and adverse health outcomes.
Hence, given that a depressive episode itself can impair adolescents and both maternal and fetal health, individualized and shared clinical decisions about the risk-benefit ratio of antidepressant use during adolescence and pregnancy should be implemented, but adolescence and pregnancy should not be considered absolute contraindications to the use of antidepressants.
Further research in RCTs and with real-world samples matched for underlying disease is needed to confirm a possible causal association between antidepressants and adverse outcomes. Such research should consider dose-effect response; mechanistic processes; and patient-specific data such as age, clinical diagnoses, and severity of clinical condition. No absolute contraindication against the use of antidepressants is currently supported by convincing evidence.
This study had several limitations. First, we did not grade the evidence from meta-analyses of RCTs, instead focusing on a portion of available evidence. However, evidence from RCTs was limited by the selection of healthier patients and frequent short-term follow-up, among other factors.16 Many severe adverse outcomes cannot be addressed in RCTs, and observational research is the most feasible method for low-frequency and long-term health risks.17 Nevertheless, observational studies are not free from bias, either.18,87 Their results yield associations, which do not imply causality. Second, results from main analyses were affected by various confounders owing to lack of randomization, potential channeling bias, and confounding by indication.17,18,80 Specifically, the nature of the control groups was only insufficiently characterized; according to the evidence, risk differences, when matched (and not adjusted) for the underlying psychiatric disorder, become smaller or nonsignificant.43,44,51 Thus, the association with suicidality may be contributed to by the antidepressants’ limited efficacy in suicidal children and adolescents, according to results from RCTs,19 rather than by antidepressant use increasing suicidality. The association between autism spectrum disorder and SSRI use during pregnancy52 included studies that were not adjusted for confounders, in contrast with weak evidence of an association between any antidepressant use during pregnancy and autism spectrum disorder when adjusted for confounders50 (eTable 4 in the Supplement). Third, no inference can be made about newer antidepressants (eg, vortioxetine hydrobromide) that have not been assessed in any of the included meta-analyses. Fourth, the data on cardiometabolic outcomes were insufficient, which is an emerging concern regarding the increased prescribing rates of antidepressants and is a crucial area for future research.88 Fifth, we used a grading system that can provide only warnings of the potential presence of systematic biases but cannot provide evidence of the nature and extent of these biases,16,32,33 just as umbrella reviews cannot supply any comparative ranking as in network meta-analyses.
The findings of this umbrella review are important in the context of increased antidepressant use worlwide.1,2 Convincing evidence was found for the association between antidepressant use and a few adverse health outcomes, yet the prevalence of those outcomes was low in general, and no association was supported by convincing evidence after confounding by indication. Future research is needed to identify whether a causal association exists between antidepressant use and adverse outcomes.
Accepted for Publication: June 21, 2019.
Corresponding Author: Elena Dragioti, PhD, Pain and Rehabilitation Centre and Department of Medicine and Health Sciences, Linköping University, SE-581 85 Linköping, Sweden (firstname.lastname@example.org).
Published Online: October 2, 2019. doi:10.1001/jamapsychiatry.2019.2859
Author Contributions: Drs Dragioti and Solmi contributed equally to the manuscript as joint first authors. Drs Dragioti and Solmi had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Dragioti, Solmi, Favaro, Fusar-Poli, Stubbs, Firth, Tsartsalis, Vieta, McGuire, Young, Correll, Evangelou.
Acquisition, analysis, or interpretation of data: Dragioti, Solmi, Fusar-Poli, Dazzan, Thompson, Stubbs, Fornaro, Carvalho, Vieta, Shin, Correll, Evangelou.
Drafting of the manuscript: Dragioti, Solmi, Dazzan, Thompson, Stubbs, Firth, Fornaro, Tsartsalis, Vieta.
Critical revision of the manuscript for important intellectual content: Dragioti, Solmi, Favaro, Fusar-Poli, Dazzan, Thompson, Stubbs, Fornaro, Tsartsalis, Carvalho, Vieta, McGuire, Young, Shin, Correll, Evangelou.
Statistical analysis: Dragioti, Solmi, Fusar-Poli, Fornaro, Carvalho, Young, Shin, Evangelou.
Obtained funding: McGuire.
Administrative, technical, or material support: Dragioti, Stubbs, Firth, Vieta, McGuire.
Supervision: Dragioti, Favaro, Fusar-Poli, Dazzan, Stubbs, Tsartsalis, Vieta, McGuire, Young, Shin, Correll, Evangelou.
Approval of protocol: Thompson.
Conflict of Interest Disclosures: Dr Fusar-Poli reported receiving grants and personal fees from Lundbeck outside the submitted work. Dr Vieta reported receiving grants and serving as a consultant, advisor, or continuing medical education speaker for the following entities: AB-Biotics, Abbott, Allergan, Angelini, AstraZeneca, Bristol-Myers Squibb, Dainippon Sumitomo Pharma, Farmindustria, Ferrer, Forest Research Institute, Galenica, Gedeon Richter, GlaxoSmithKline, Janssen, Lundbeck, Otsuka, Pfizer, Roche, SAGE, Sanofi, Servier, Shire, Sunovion, Takeda, the Brain and Behaviour Foundation, CIBERSAM, the Seventh European Framework Programme and Horizon 2020, and the Stanley Medical Research Institute. Dr Young reported receiving grants and personal fees from Janssen; receiving personal fees from Lundbeck, Allegan, Sunovion, Livanova, Johnson & Johnson, and Bionomics outside the submitted work; being employed by King's College London and an honorary consultant for SLaM (NHS United Kingdom); receiving fees for lectures and service on advisory boards for the following companies with drugs for affective and related disorders: AstraZeneca, Eli Lilly, Lundbeck, Sunovion, Servier, Livanova, Janssen, Allegan, and Bionomics; being a consultant to Johnson & Johnson and Livanova; not having share holdings in pharmaceutical companies; being lead investigator for the Embolden Study (AstraZeneca), BCI Neuroplasticity Study, and Aripiprazole Mania Study and participating in investigator-initiated studies from AstraZeneca, Eli Lilly, Lundbeck, Wyeth, and Janssen; and receiving past and present grant funding from the National Institute of Mental Health, Canadian Institutes of Health Research, National Alliance for Research on Schizophrenia and Depression, Stanley Medical Research Institute, Medical Research Council, Wellcome Trust, Royal College of Physicians, British Medical Association, Vancouver General Hospital and University of Bristish Columbia Hospital Foundation, Western Economic Diversification Canada, Canadian Cancer Society Depression Research Fund, Michael Smith Foundation for Health Research, National Institute for Health Research (NIHR), and Janssen. Dr Correll reported receiving personal fees from Alkermes, Allergan, Angelini, Boehringer-Ingelheim, Gedeon Richter, Indivior, LB Pharma, Lundbeck, MedAvante-ProPhase, Merck, Neurocrine, Noven, Otsuka, Pfizer, Recordati, Rovi, Servier, Sumitomo Dainippon, Sunovion, and Supernus; receiving grants and personal fees from Janssen/Johnson & Johnson and Teva outside the submitted work; serving on a data safety monitoring board for Boehringer-Ingelheim, Lundbeck, Rovi, Supernus, and Teva; providing expert testimony for Bristol-Myers Squibb, Janssen, and Otsuka; receiving royalties from UpToDate and grant support from Janssen and Takeda; and being shareholder of LB Pharma. No other disclosures were reported.
Funding/Support: This independent research was funded by the NIHR Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King’s College London. Dr Firth was supported by a Blackmores Institute Fellowship. Dr Stubbs holds a clinical lectureship supported by Health Education England and the NIHR Integrated Clinical Academic (ICA) Programme (ICA-CL-2017-03-001) and was supported in part by the Maudsley Charity and the NIHR Collaboration for Leadership in Applied Health Research and Care South London at King’s College Hospital NHS Foundation Trust.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The views expressed here are those of the authors and do not necessarily reflect those of the NHS, the NIHR, or the Department of Health and Social Care.
Create a personal account or sign in to: