Outcome definition | ||||||||
---|---|---|---|---|---|---|---|---|
Study ID and design | Participants | Exposure | Suicide | Attempted suicide | Combined suicide and attempted suicide | Outcome measures (95% CI) | Factors adjusted for in statistical analysis | Critique of study |
Antiepileptic drugs | ||||||||
Arana et al;22 cohort and case–control | THIN, UK, 1/7/1988–31/03/2008; cohort n=5 130 795; first suicide event n=8212 (completed n=464, attempted n=7748); case–control: any suicide event cases n=10 164, controls n=51 005 | AED (carbamazepine/gabapentin/lamotrigine/levetiracetam/oxcarbazepine/pregabalin/tiagabine/topiramate/valproate/zonisamide) | – | – | Read codes (unspecified) for suicide, attempted suicide and intentional self-inflicted injuries plus suicide. Suicide death determined if code for death in month following suicidality code, and final database activity within 6 months | Cohort study: described incidences Case–control study: adjusted OR vs no epilepsy/bipolar/depressive disorder and no AED No epilepsy/bipolar/depressive disorder with AED: 2.57 (1.78 to 3.71) Epilepsy with AED: 2.31 (1.77 to 3.02); Epilepsy and no AED: 3.34 (2.34 to 4.78) OR AED use in epilepsy vs non-use: 0.59 (0.35 to 0.98) | Age, disease duration, history: AED/antidepressant/lithium/antipsychotic/mental illness/alcohol abuse; chronic disease score Excluded: personal or family history of suicide attempt | Grouped outcomes: AEDs grouped and proportions of individual AEDs not presented |
Gibbons et al;23 cohort | PharMetrics Patient Centric Database, USA, 2000–2006. Cohort n=131 178; suicide attempt before initiation n=456, suicide attempt after initiation n=453 | Gabapentin | – | ICD-9 codes E950–959 | – | Adjusted event rate ratio after gabapentin initiation vs before initiation: epilepsy: 0.83 (0.34 to 2.04); pain disorder: 0.99 (0.78 to 1.26) Gabapentin monotherapy: 0.53 (0.16 to 1.73) | Adjustments: age, sex, concomitant diagnosis/treatment. Stratified by conditions. | Few outcomes in epilepsy and gabapentin monotherapy groups |
Nilsson et al;24 case–control | Stockholm County Council In-Patient Care Register, age>15; epilepsy diagnosis and inpatient 1980–1989. Cases: death before 31/12/1997; age <78; controls: alive on 31/12/1992 n=171 suicide and undetermined suicide n=49 (n=26 in analyses) | Controls: phenytoin/carbamazepine/valproate Cases: any AED | ICD-9 E950–959; ICD-10 X60-X84, Undetermined intent: 980–989; Y10–34 | – | – | Adjusted relative risk vs 1 AED:2 AED: 2.0 (0.8 to 5.2); 3 AED: 3.1 (0.6 to 17.5) Relative risk: number of dose changes vs 0 dose changes: 1–5 changes: 1.2 (0.4 to 3.4); unknown number of dose changes: 13.6 (3.8 to 49.2) | Age, sex | No adjustment for psychiatric comorbidities. Cases and controls unmatched but similar distributions seen. Controls subject to immortal time bias because61 AED use required for ≥1 year. Few cases may affect statistical power |
Olesen et al;25 case-crossover analyses and cohort | Danish databases: National Prescription Register, Danish National Patient Register and National Causes of Death Registry 1/1/1997–31/12/2006, age ≥10. Case-crossover: suicide total n=6780, in study period n=898. Cohort: newly prescribed AED n=169 725, Suicide n=670 during treatment n=268 | AED (carbamazepine/clobazam/clonazepam/gabapentin/lamotrigine/levetiracetam/ oxcarbazepine/phenobarbital/ phenytoin/pregabalin/ primidone/ tiagabine/topiramate/valproate/zonisamide) | National Cause of Death Register: ICD-10 X60-X84 | – | – | Case-crossover analyses: OR AED exposure in case period vs control period: Overall AED: 1.84 (1.36 to 2.49); carbamazepine: 0.48 (0.21 to 1.12); gabapentin: 2.20 (0.83 to 5.83); lamotrigine: 3.15 (1.35 to 7.34); oxcarbazepine: 0.84 (0.30 to 2.32); phenobarbital 1.96 (1.02 to 3.75); valproate: 2.08 (1.04 to 4.16); topiramate: 2.72 (0.23 to 32.78) Cohort Study: adjusted HR AED initiation vs carbamazepine: gabapentin:1.27 (0.66 to 2.44); lamotrigine: 2.09 (1.25 to 3.50); oxcarbazepine 1.69 (0.81 to 3.56); valproate: 2.40 (1.42 to 4.05) | Cohort analyses: age, sex, socioeconomic status, Charlsons’ score, civil status, epilepsy/psychiatric disorder, opiate <180 days prior to index date; concomitant antidepressant/antipsychotic/anxiolytic | Case-crossover design suitability: Exposure may be influenced by indication which may independently increase risk of suicide when exposed to treatment |
Patorno et al;26 cohort | HealthCore Integrated Research Database, USA; 07/2001–12/2006, age ≥15; new AED Cohort n=297 620 new treatment episodes, Suicide n=26, Suicide Attempt n=801 | New exposure to AED (carbamazepine/ethosuxamide/felbamate/gabapentin/lamotrigine/levetiracetam/oxcarbazepine/phenobarbital phenytoin/pregabalin/primidone/tiagabine/topiramate/valproate/zonisamide) | – | – | Suicide attempt: emergency department/hospitalisation ICD-9 E950-E958 Suicide: ICD-10 X60–84 | Adjusted HR for suicide and suicide attempt within 180 days of exposure vs topiramate: carbamazepine: 1.24 (0.77 to 1.99); gabapentin: 1.42 (1.11 to 1.80); lamotrigine: 1.84 (1.43 to 2.37); levetricetam: 1.63 (0.84 to 3.16); oxcarbazepine: 2.07 (1.52 to 2.80); valproate: 1.65 (1.25 to 2.19) Propensity score matched HR for suicide or suicide attempt in epilepsy/seizure disorder stratum vs carbamazepine: gabapentin: 13.92 (1.82 to 106.38); oxcarbazepine: 0.73 (0.16 to 3.28); phenytoin: 3.48 (0.97 to 12.47); topiramate: 0.67 (0.37 to 1.19); valproate: 0.49 (0.09 to 2.70) | 49 covariates including diagnosis of or medication for: depression, mania, psychosis, anxiety, substance/alcohol abuse, personality disorder, other psychiatric disorders, physical disorders Propensity score matched analysis: in sensitivity analysis | Comparator suitability: topiramate selected but low frequency of use in epilepsy, although comparison with carbamazepine was repeated for this stratum. Imprecise estimate for gabapentin when restricted to people with epilepsy/seizure disorder: requires cautious interpretation |
Cardiovascular drugs | ||||||||
Callréus et al;27 Nested case–ontrol | Danish Registry of Cause of Death and The Odense University Pharmacoepidemiological Database, 1991–1998. Suicide Cases n=743, controls n=14 860 | Medications used for lipid lowering, CCB, β-blockers, ACE-I and ARIIB | ICD-8 E950–959, ICD-10 X60-X84, Y87 | – | – | OR for current use vs no use: statin: 1.25 (0.42 to 3.76); any lipid lowering drug: 1.21 (0.45 to 3.28); CCB: 0.96 (0.63 to 1.48); β-blocker: 0.76 (0.47 to 1.25); ACE-I: 1.11 (0.68 to 1.83); ARIIB: 3.52 (1.33 to 9.30) | Adjusted for: ever use of antipsychotics, lithium, antidepressants and drugs for alcohol dependence; use of antidiabetics or AED in past year. Sensitivity analysis: excluded if history of psychotropic drug use | Limited statistical power for some strata |
Gasse et al;28 Nested case–control | GPRD, UK, 01/1991–08/1998, cases n=38 controls n=140 nested in cohort of individuals with hypertension diagnosis and prescription for antihypertensive medication | Antihypertensive (including CCB, β-blocker, ACE-I, diuretic) | OXMIS 3009D | – | – | Adjusted relative risk CCB vs other antihypertensive: 0.98 (0.30 to 3.18) | Body mass index, smoking and history of mental health illness | Potential under ascertainment of suicide outcomes by use of OXMIS codes but this is likely to be non-differential (N.B. at time of publication outcome definition suitable) |
Haukka et al;29 cohort | Finish databases: Social Insurance Institution, National Hospital Discharge Register, Causes of Death Register; 01/01/1997–31/12/2005. Exposed n=336 618; Unexposed n=336 618; Suicide n=350 | Statin | Cause of Death Register: ICD-10 X60-X84 | – | – | Adjusted HR statin exposed vs unexposed: 0.53 (0.43 to 0.65) Poisson regression analysis of suicide per years of follow-up in statin use vs non-use: 0.5 ≤1 year: 0.49 (0.20 to 1.16);1 ≤2 years: 0.59 (0.28 to 1.25); 2 ≤3 years: 0.26 (0.11 to 0.60); 3 ≤4 years: 0.50 (0.21 to 1.19) | Adjusted: Sex, age, baseline diseases Additional adjustments in Poisson model: follow-up time, statin group Excluded: history of antidepressant medication | Limited adjustment for confounding factors: particularly psychiatric comorbidity (although individuals with depression excluded). Not generalisable to individuals with depression |
Lindberg et al;30 cohort | Swedish pharmacy data,1988–1989, cohort n=3397; CCB n=617, other cardiovascular medication n=2780 | Cardiovascular medication including CCB, β-blocker, ACE I, diuretic | Swedish Mortality Register: ICD-9 E950–959, E980–989 | – | – | 7-year suicide risk difference for CCB use vs non-use: 7.5/1000 person-years (p=0.002) Adjusted relative risk: 5.4 (1.4 to 20.5) | RR adjusted for age and sex. | Limited statistical power: 9 suicides in total. No adjustment for many confounders: including history of mental illness. Indication unknown |
Sørensen et al; 31 cohort | Pharmacoepidemiological Prescription database of North Jutland County, Denmark, 01/01/1989–31/12/1995; cohort n=58 529 | CCB, β-blocker or ACE inhibitor | National Death Certificate Files: ICD 8/10 codes | – | – | SMR ever-use: β-blocker: 1.6 (1.2 to 2.1); CCB: 1.2 (0.8 to 1.7); ACE-I 1.2 (0.7 to 1.8). SMR present use of single study drug: β-blocker: 1.4 (0.9 to 2.1); CCB: 1.2 (0.7 to 1.9); 1.2 (0.5 to 2.4). SMR for β-blocker use only (present and former): low lipid solubility: 0.9 (0.4 to 1.9); high lipid solubility: 2.7 (1.7 to 4.1) | – | Comparison to general population: differences in baseline characteristics and adjustment for confounders not possible (although some potential confounding factors compared between treatment groups). Potential overestimation of exposure: present use considered for 180 days after prescription received |
Leukotriene receptor antagonists | ||||||||
Jick et al;32 cohort | GPRD, UK, 02/1998–03/2007; cohort n=23 500 | Montelukast | Computer recorded diagnosis | – | – | Suicides in 21 050 person years n=0 | No comparison group-would be necessary to compare incidence if cases were identified | |
Schumock et al;33 Nested case–control | LifeLink Health Plan Claims Database, USA, 1/1/1997–31/12/2006, asthma and new use of an asthma treatment; age 5–24. Cases n=344 Controls n=3438 | Montelukast, zafirlukast, zileuton | – | ICD-9 code E950-E959 | – | Adjusted OR vs never use: current use: 0.70 (0.36 to 1.39), immediate past use: 0.95 (0.36 to 2.50), past use: 0.69 (0.32 to 1.50). Ever use: 0.74 (0.46 to 1.20) | History: bipolar disorder, depression, other mental disorder, substance abuse, suicide attempt; psychological counselling; asthma severity (by proxy) | All cases were exposed to montelukast rather than other drugs in this class (reflects prescribing trends), but results attributed to entire class |
Isotretinoin | ||||||||
Jick et al;34 cohort | Canadian Saskatchewan Health Database,1983–1997, acne diagnosis and drug exposure; isotretinoin n=7195, antibiotic n=13 700; suicide or SA n=38 GPRD: data not analysed | Isotretinoin or antibiotic (erythromycin/ doxycycline/ minocycline/tetracycline) | – | – | ICD-9 E codes | Adjusted relative risk vs non-exposure: isotretinoin: current use: 0.9 (0.3 to 2.4); recent use: 1.1 (0.2 to 3.7) Antibiotic: current use: 0.8 (0.4 to 1.7); recent use: 0.5 (0.1 to 1.4) Adjusted relative risk vs non-exposure in individuals with no psychiatric history: Isotretinoin: current use: 1.3 (0.3 to 4.6); recent use: 1.0 (0.1 to 5.7) Antibiotic: current use: 0.5 (0.1 to 1.6); recent: 0.7 (0.1 to 2.7) | Adjusted for sex, history of psychiatric disorder (depression, psychosis, attempted suicide) Stratified: by history of psychiatric disorder | Stratified analysis of individuals with no psychiatric history: small number may limit power. |
Sundström et al;35 cohort and crossover analysis | Patient Register (mandatory), Sweden; 1980–1989 (outcomes identified until 2001), aged 15–49, cohort: n=5756; suicide: n=24; suicide attempt: n=128 | Isotretinoin | National cause of death registry (including unclear intent), ICD codes | Suicide attempt hospitalisation ICD-8 and 9 E950-E958, E980-E988: ICD-10 X60–64; Y10-Y34 | – | Cohort: standardised incidence ratio for all suicide attempts: isotretinoin users vs general population: 3 years pre-treatment: 0.99 (0.65 to 1.44); 1 year pretreatment: 1.57 (0.86 to 2.63); <6 months post-treatment: 1.78 (1.04 to 2.85); 3 years post-treatment: 1.04 (0.74 to 1.43) Case-crossover: rate difference 1 year pretreatment vs 6 months post-treatment: 1st attempts: 0.86 (−0.78 to 2.50) cases/1000 person-years; all attempts: 0.40 (−1.46 to 2.26) cases/1000 person-years SMR for suicide: incompletely recorded | – | Outcome misclassification possible: attribution of suicide attempt to exposure for up to15 years following exposure, may overestimate attempts. Comparison with general population: no control for confounding factors, including confounding by indication. This was explored in the case-crossover design |
Corticosteroids | ||||||||
Fardet et al;36 cohort | THIN, UK, 01/01/1990–31/12/2008; age ≥18; new glucocorticoid exposure n=372 696; exposed with indication n=261 272; unexposed n=1 224 984; unexposed matched by indication n=660 776; exposed groups suicide n=19; suicide attempt n=90 | Oral glucocorticoid (betamethasone/deflazacort/dexamethasone/hydrocortisone/methylprednisolone/prednisone/prednisolone/triamcinolone/) | – | – | Read codes and cross-searched death certificates | Adjusted HR: exposed vs unexposed: 5.27 (3.82 to 7.29) Exposed vs unexposed, matched by indication: 6.89 (4.52 to 10.50) | Adjusted: age, sex, history neuropsychiatric disorder Separate analysis for cohorts matched by indication | Handling of repeated courses: random selection of course could alter baseline risk dependent on course selected. If later courses chosen, individual subject to immortal time61 bias until this time |
Fardet et al;37 cohort and self-controlled case series | THIN, UK 01/01/1990–31/12/2008; age≥18; glucocorticoid use for 1–3 years. Cohort: n=21 995. Eligible for self-controlled case series analysis: n=991; suicide or suicide attempt n=6 | Oral glucocorticoid (betamethasone/deflazacort/dexamethasone/hydrocortisone/methylprednisolone/prednisone/prednisolone/triamcinolone/) | – | – | Read codes | Cohort: incidence rate for suicide or attempted suicide during withdrawal period: 0.03 (0.01 to 0.2) Self-controlled case series: incident rate ratio in withdrawal period vs ref. period: 0.62 (0.06 to 6.92) (ref. period:5–3 months prior to discontinuation) | Inadequate statistical power: only 6 cases of suicide or attempted suicide. Potential immortal time bias:61 for entry into the cohort, must not have died in first year of glucocorticoid use. To be eligible for the self-controlled case series analysis, any suicides must occur in the withdrawal period | |
Antibiotics (quinolones) | ||||||||
Jick et al;38 Nested ase Control | GPRD, UK, ever exposed to quinolone, 01/01/1991–30/4/1995, age 15–84. Cases n=348 (suicide n=13, suicide attempt n=206 suicidal ideation n=129) Control n=808 (NB. Outcomes analysed separately) | Quinolone or other antibiotic in 1–30 or 31–180 days prior to index date | OXMIS code 3009D | OXMIS code L3009P, 9779 L, 3009C | – | Adjusted relative risk estimate for suicide attempt vs non-exposure: quinolone 1–30 days: N/A; quinolone 31–180 days: 0.6 (0.2 to 1.5); other antibiotic 1–30 days: 1.2 (0.5 to 2.6); other antibiotic 31–180 days: 0.9 (0.5 to 1.5) | Age, sex, history of: depression, suicidal behaviour, insomnia, psychosis, anxiety, alcoholism and epilepsy | Inadequate statistical power: precluded calculation of risk within first month of quinolone exposure or risk of suicide death. Possible underestimation of suicide and attempted suicide: because 1st recorded event used in multiple outcomes. However, all suicidal ideation comparisons were non-significant |
Varenicline | ||||||||
Gibbons and Mann;39 cohort | Military Healthcare System, USA, 01/08/2006–31/08/2007. Cohort treated with varenicline (n=19 933); NRT patches (n=15 867). After matching by propensity score: patients n=26 430; included suicide attempts n=5 | Varenicline or NRT patches | – | ICD-9 E950–959 | – | OR varenicline vs NRT in patients for whom propensity score matching was possible: 0.67 (0.11 to 3.99) | Propensity score matching: age, marital status, race, sex, Charlsons’ score, inpatient admissions, psychiatric comorbidity, psychotropic medication | Limited power: few suicide attempt events (n=5) |
Other medication | ||||||||
Voaklander et al;40 case–control | British Columbia Vital Statistics, Health Insurance Registration File, Pharmacare and Physician Claim File, Canada, 1993–2002, age ≥66, suicide cases n=602 controls n=2999 | anti-diabetic agents, anticoagulants, cardiovascular drugs, NSAID, ulcer medication, steroids | ICD-9 E950-E959; ICD-10: X60—X84 | – | – | Unadjusted ORs vs non-use: antihypertensive medication: 0.94 (0.67 to 1.31); lipid lowering medication: 0.60 (0.28 to 1.26); anticoagulants: 1.07 (0.52 to 2.22); diuretics: 0.94 (0.66 to 1.36); ulcer medication: 1.88 (1.35 to 2.62); steroids: 1.33 (0.88 to 2.00). Fully adjusted OR: diuretics: 0.49 (0.31 to 0.76) | Fully adjusted analyses: demographics, co-morbidity, medication use | Fully adjusted analysis not done for all medications which suggested significance at the unadjusted level (eg, ulcer medications) |
AED, antiepileptic drugs; ARIIB, angiotensin receptor II blocker; CCB, calcium blocker; CNS, central nervous system; GPRD, general practice research database; ICD, International Classification of Diseases; NRT, nicotine replacement therapy; NSAID, non-steroidal anti-inflammatory drug; OXMIS, Oxford Medical Information System; SMR, standardised mortality ratio; THIN, The Health Improvement Network.