Table S2. Studies included in the systematic review and meta-analysis that assessed

Transcription

1 Appendix 1 to Wu S, Ding Y, Wu F, et al. Serum lipid levels and suicidality: a meta-analysis of 65 epidemiological studies. J Psychiatry Neurosci DOI: /jpn Copyright 2015, Canada Inc. Online appendices are unedited and posted as supplied by the authors. Supplementary data Table S1. MOOSE Checklist Panel S1. Search strategy in Medline and Embase Table S2. Studies included in the systematic review and meta-analysis that assessed the serum lipid levels in different groups Table S3. Studies included in the systematic review and meta-analysis that assessed the risk of suicidality according to the lowest vs highest serum total cholesterol level category Table S4. Quality of the included studies (cohort studies) Table S5. Quality of the included studies (case-control studies) Table S6. Quality of the included studies (cross-s studies) Figure S1. Forest plot displaying the pooled summary estimates of serum TC levels in the suicidal patients versus non-suicidal patients (A), and in the suicidal patients versus normal controls (B), after exclusion of studies not controlled for at least age or sex Figure S2. Forest plot displaying the pooled summary estimate of serum HDL-C levels in the suicidal patients versus normal controls, after exclusion of studies not controlled for at least age or sex Figure S3. Forest plot displaying the pooled summary estimates of serum LDL-C levels in the suicidal patients versus non-suicidal patients (A), and in the suicidal

2 Appendix 1 to Wu S, Ding Y, Wu F, et al. Serum lipid levels and suicidality: a meta-analysis of 65 epidemiological studies. J Psychiatry Neurosci DOI: /jpn Copyright 2015, Canada Inc. Online appendices are unedited and posted as supplied by the authors. patients versus normal controls (B), after exclusion of studies not controlled for at least age or sex Figure S4. Forest plot displaying the pooled summary estimates of serum TG levels in the suicidal patients versus non-suicidal patients (A), and in the suicidal patients versus normal controls (B), after exclusion of studies not controlled for at least age or sex Figure S5. Forest plot displaying the pooled relative risk of suicidality according to the lowest vs highest serum TC level category, after exclusion of studies not controlled for at least age or sex Figure S6. Funnel plots to assess publication bias

3 Table S1: MOOSE Checklist Criteria Reporting of background should include Problem definition Brief description of how the criteria were handled in the meta-analysis There has been considerable controversy about the association between serum lipid levels and suicide in observational and epidemiological studies. Hypothesis statement Suicidality is associated with low serum lipid levels. Description of study outcomes Serum lipid level. Type of exposure or Suicidal behaviour intervention used Type of study designs used Cohort, case-control, or cross-s studies. Study population Population-based or -based psychiatric patients Reporting of search strategy should include Qualifications of searchers (eg, librarians and investigators) Search strategy, including time period included in the synthesis and keywords Effort to include all available studies, including contact with authors Databases and registries searched Search software used, name and version, including special features used (eg, explosion) Use of hand searching (eg, reference lists of obtained articles) List of citations located and The credentials of the two investigators SW and YD are provided in the author list. Medline, Web of Science, Embase, and the Cochrane Database of Systematic Reviews (1980 to Dec 5, 2014) Keywords: lipids, cholesterol, suicide, total cholesterol, high-density lipoprotein, low-density lipoprotein, triglycerides, suicidal, suicidality, in combination with survey and Data Collection. References of all retrieved articles and recent reviews were reviewed. Medline, Web of Science, Embase, and the Cochrane Database of Systematic Reviews We did not employ a special search software. References of all retrieved articles and recent reviews were reviewed. Details of the literature search process are outlined in the 3

4 those excluded, including justification Method of addressing articles published in languages other than English Method of handling abstracts and unpublished studies Description of any contact with authors. Reporting of methods should flow chart. We placed restrictions on English. The search process was not restricted upon full-text articles, but also conference abstracts and unpublished studies. - include Description of relevance or appropriateness of studies The inclusion criteria are presented in the Search strategy and eligibility criteria section. assembled for assessing the hypothesis to be tested Rationale for the selection and coding of data (eg, sound clinical principles or convenience) Study characteristics were extracted independently by two researchers (FW and JH). We extracted means±sd or risk estimates (95% CI) for males and females separately when possible. If a study reported more than one measure of serum lipid level, each serum lipid level was extracted separately. We included the most adjusted estimate when a study reported more than one risk estimate. Documentation of how data were classified and coded (eg, multiple raters, blinding, and Data were independently extracted and analyzed by two investigators (FW and JH) and final decision was reached by consensus. inrerrater reliability) Assessment of confounding (eg, comparability of cases and controls in studies where appropriate) Table S3 presents the adjustment factors for each study. Assessment of study quality, including blinding of quality The quality of each study was assessed by two investigators (FW and PM), using the Newcastle-Ottawa Scale. assessors; stratification or regression on possible predictiors of study results Assessment of heterogeneity The Q-statistic and I-squared (I 2 ) statistic were used to explore the heterogeneity among studies. Description of statistical methods (eg, complete description of fixed or random effects models, justification of whether the chosen models account for predictors of study Description of methods of meta-analyses, subgroup analyses, and of publication bias are detailed in the Statistical analysis section. 4

5 results, dose-response models, or cumulative meta-analysis) in sufficient detail to be replicated Provision of appropriate tables and graphics Two main tables and six supplemental tables are provided. One flow chart and five forest plots appear in the main text. Funnel plots are provided in supplemental materials. Reporting of results should include Graph summarizing individual study estimates and overall estimate Table giving descriptive information for each study included Results of sensitivity testing (eg, subgroup analysis) Indication of statistical uncertainty of findings Reporting of discussion should include Quantitative of bias (eg, publication bias) Justification for exclusion (eg, exclusion of non-english-language citations) Assessment of quality of included studies Reporting of conclusions should Figures 2-6 Table S2-S3 Results section; Table % confidence intervals were presented with all summary effect estimates. Results section, page 11; Discussion section, page 14. The details of the exclusion of studies are shown in Flow chart. Studies have been subgroup analyzed by the quality. include Consideration of alternative explanations for observed results We discussed that some included studies didn't make adjustment for other factors or only make adjustment for a few important factors, thus, other factors may contribute to the alternations in serum lipid levels in suicidal patients, such as genetic factors, alcohol drinking, and cigarette smoking. Generalization of the conclusions (ie, appropriate for the data presented and within the domain of the literature We discussed that the consistency of results across studies and setting suggests that the findings of lower serum lipid levels in suicidal patients and increased risk of suicide in subjects with lower serum TC levels are robust. review) Guidelines for future research Further studies are urgent needed to explain the association between serum lipid levels and suicide and elucidate the cause-effect relationship. Disclosure of funding source The authors received no specific funding for this work. 5

6 Panel S1: Search strategy in Medline and Embase 1. lipids (MeSH term (Medical Subject Headings)) 2. cholesterol (MeSH term) 3. suicide (MeSH term) 4. total cholesterol (key word) 5. high-density lipoprotein (key word) 6. low-density lipoprotein (key word) 7. triglycerides (key word) 8. suicidal (key word) 9. suicidality (key word) 11. survey (key word) or Data Collection (MeSH term) or and or 5 or 6 or or and and limit 17 to (English language and humans and year= 1980-Current ) 6

7 Table S2: Studies included in the systematic review and meta-analysis that assessed the serum lipid levels in different groups Number and serum Number and serum Number and serum Serum Study Age (years, mean±sd or lipid level of suicidal lipid level of lipid level of normal Study Study site Participants Sex lipid Outcome design range) patients non-suicidal patients controls measure n1 mean±sd n2 mean±sd n3 mean±sd Ainiyet et Poland 69 male, 30.0±10.0 (male), TC SI al, 2014 schizophrenia 79 female 33.0±10.0 (female) a 22 a ±45.00 a 47 a ±47.00 b 35 b ±24.00 b 44 b ±55.00 a 6 a ±22.00 a 63 a ±51.00 b 10 b ±16.00 b 69 b ±54.00 ST a 16 a ±45.00 a 53 a ±52.00 b 22 b ±27.00 b 57 b ±55.00 HDL-C SI a 22 a 45.00±15.00 a 47 a 45.00±10.00 b 35 b 52.00±14.00 b 44 b 55.00±14.00 a 6 a 54.00±16.00 a 63 a 44.00±11.00 b 10 b 49.00±13.00 b 69 b 54.00±14.00 ST a 16 a 47.00±16.00 a 53 a 44.00±11.00 b 22 b 50.00±12.00 b 57 b 55.00±14.00 LDL-C SI a 22 a ±42.00 a 47 a ±42.00 b 35 b ±25.00 b 44 b ±49.00 a 6 a 93.00±21.00 a 63 a ±45.00 b 10 b ±18.00 b 69 b ±47.00 ST a 16 a ±41.00 a 53 a ±45.00 b 22 b ±28.00 b 57 b ±49.00 TG SI a 22 a 91.00±48.00 a 47 a ±84.00 b 35 b 70.00±44.00 b 44 b ±71.00 a 6 a 64.00±20.00 a 63 a ±

8 b 10 b 69.00±51.00 b 69 b ±66.00 ST a 16 a 95.00±51.00 a 53 a ±79.00 b 22 b 65.00±39.00 b 57 b ±69.00 Almeida- Mexico 8 male, 32.0±12.9 (with ), TC ± ±22.20 Montes et depression 25 female 31.9±8.8 (without ) HDL-C ± ±9.30 al, 2000 LDL-C ± ±21.30 TG ± ±59.00 Alvarez et France 40 male, 41.0±13.9 (violent suicide TC Violent ± ±46.00 al, 2000 psychiatric 44 female attempters), disease and 39.3±12.3 (non-violent Non-viole ± ±46.00 normal controls suicide attempters), nt 44.5±14.5 (controls) Alvarez et France 28 male, 44.7±17.1 (violent suicide TC Violent ± ±46.79 al, 1999 psychiatric 6 female attempters), disease and 48.8±16.6 (controls) normal controls Apter et Israel Adolescent 73 male, 15.9±1.5 (with ), TC ST a 15 a ±30.39 a 58 a ±32.10 al, 1999 patients with 79 female 16.2±2.0 (without ) b 32 b 181.1±33.69 b 47 b ±40.17 psychiatric disease Atmaca et Turkey Combined TC ST ± ± ±17.40 al, 2002 borderline personality and normal controls Atmaca et Turkey 23 male, 28.2±5.1 (with ), TC a 7 a ±25.20 a 8 a ±23.50 a 8 a ±26.80 al, 2003 schizophrenia 25 female 30.1±3.6 (without ), b 9 b ±13.50 b 8 b ±19.60 b 8 b ±

9 29.4±3.2 (controls) Atmaca et Turkey 21 male, 32.6±6.2 (violent suicide TC Violent ± ±27.90 al, 2008 psychiatric 34 female attempters), disease and 34.1±6.7 (non-violent Non-viole ± ±27.90 normal controls suicide attempters), nt 31.8±5.8 (controls) Atmaca et Turkey 21 male, 28.6±14.2 (with ), TC ± ±32.56 al, 2002 psychiatric 27 female 29.1±13.3 (controls) disease and normal controls Atmaca et Turkey 22 male, 27.9±12.4 (with ), TC ± ±28.60 al, 2006 psychiatric 38 female 28.8±15.3 (controls) disease and normal controls Baek et al, Korea Patients 150 male, 35.5±15.9 (with recent ), TC Recent ± ± with major ±15.4 (with lifetime Lifetime ± ±36.60 depressive female ), disorder 47.5±15.6 (without ) HDL-C Recent ± ±15.40 Lifetime ± ±15.40 LDL-C Recent ± ±33.00 Lifetime ± ±33.00 TG Recent ± ±87.20 Lifetime ± ±

10 Coryell et al, 2007 da Graça Cantarelli et al, 2014 D'Ambros io et al, 2012 Deisenha mmer et al, 2004 Deisenha mmer et al, 2006 Diaz-Sastr e et al, 2007 Favaro et al, 2004 United Cohort Combined Not reported TC CS ± ±43.90 States schizoaffective depression Brazil Subjects with 23 male, ecitonal mood disorder 63 female Italy 83 male, bipolar disorder 128 female Austria a 34 male, major depressive 58 female episode Austria 75 male, alcohol 35 female dependence Spain Case-co Suicide attempters 136 male, ntrol and controls 218 female 27.8±12.2 (with ), TC ± ± ±13.9 (without ) HDL-C ± ±13.13 LDL-C ± ±36.12 TG ± ± ±13.6 (with ), TC ± ± ±16.5 (without ) HDL-C ± ±14.80 TG ± ± ±11.7 TC ± ±36.20 HDL-C ± ±12.40 LDL-C ± ±34.10 TG ± ± ±8.0 (with ), TC ± ± ±8.1 (without ) HDL-C ± ±30.50 LDL-C ± ±41.80 TG ± ± ±9.8 (male attempters), TC ± ± ±12.7 (female attempters), 34.8±9.8 (male controls), 35.5±12.8 (female controls) Italy Female 23.6±7.4 TC ± ±50.27 anorexia nervosa SI ± ±50.27 Fiedorowi United Cohort 26 male, 39.8±14.8 (high TC All ± ±

11 cz et al, States psychiatric 48 female cholesterol), Severe ± ± disease 27.0±8.5 (low cholesterol) Gallerani Italy Parasuicide 186 male, 44.0±21.0 (suicidal TC ± ±50.27 et al, 1995 subjects and 476 patients), TG ± ±23.98 controls female 44.0±20.0 (controls) Guillem et France 83 male, 38.7±12.4 (male), TC a 13 a ±54.14 a 70 a ±45.63 al, 2002 psychiatric 72 female 37.2±11.3 (female) b 21 b ±32.10 b 51 b ±52.59 disease Huang et Taiwan 47 male, 30.6±10.8 TC ± ±36.40 al, 2000 psychiatric 59 female disease Huang, Taiwan 54 male, 34.0±8.0 (male patients), TC a 6 a ±21.80 a 26 a ±38.80 a 22 a ± major depression ±8.6 (female patients), b 15 b ±27.50 b 62 b ±32.40 b 37 b ±22.80 and normal female 31.7±4.6 (male controls), HDL-C a 6 a 52.20±24.90 a 26 a 48.50±13.30 a 22 a 45.60±11.50 controls 28.3±3.7 (female controls) b 15 b 50.80±11.50 b 62 b 51.00±12.00 b 37 b 57.10±12.30 LDL-C a 6 a ±26.20 a 26 a ±38.60 a 22 a ±33.00 b 15 b ±18.40 b 62 b ±27.20 b 37 b 98.40±19.30 TG a 6 a 63.50±15.30 a 26 a ±71.80 a 22 a ±40.40 b 15 b ±62.60 b 62 b 99.10±51.50 b 37 b 67.90±23.20 Jokinen et Sweden Suicide attempters Male (suicide attempters), TC ± ±38.67 al, 2010 and normal (controls) controls Kim et al, Korea Community 300 male, 74.2±6.3 (with SI), TC SI ± ± dwelling elderly ±5.7 (without SI) HDL-C SI ± ±13.60 female LDL-C SI ± ±30.00 TG SI ± ±

12 Kim et al, Korea 327 male, 38.1±17.0 (with ), TC ± ± ± psychiatric ±15.4 (without ), disease and female 38.2±16.3 (controls) normal controls Kim et al, Korea 225 male, 40.5±18.1 (with ), TC ± ± ± major depressive ±16.5 (without ), disorder and female 38.0±15.9 (controls) normal controls Kunugi et Japan Case-co Suicide attempters 69 male, 40.0±16.7 (suicide TC ± ±39.00 al, 1997 ntrol and psychiatric 104 attempters), controls female 39.6±15.5 (psychiatric controls) Suicide attempters 47 male, 38.8±16.3 (suicide TC ± ±45.00 and normal 48 female attempters), controls 33.5±13.4 (controls) Lee et al, Korea 57 male, 36.1±13.8 (with ), TC ± ± ± psychiatric ±14.8 (without ), LDL-C ± ± ±32.64 disease and female 36.5±13.3 (controls) normal controls 22 male, Not reported TC ± ± ±30.95 major depression 42 female TG ± ± male, Not reported TC ± ±38.27 personality 40 female LDL-C ± ±26.10 disorder Maes et Belgium Male Not reported HDL-C ± ±8.10 al, 1997 major depression 12

13 Marcinko Croatia Male 29.0±6.5 (suicidal group), TC ST ± ± ±32.87 et al, 2011 borderline 29.9±7.6 (non-suicidal personality group), disorder and 32.1±7.2 (controls) normal controls Marcinko Croatia Male 28.9±8.8 (suicidal group), TC ST ± ± ±45.63 et al, 2008 schizoaffective 37.4±12.2 (non-suicidal HDL-C ST ± ± ±18.95 disorder and group) LDL-C ST ± ± ±42.15 normal controls 29.9±7.6 (controls) TG ST ± ± ±72.70 Marcinko Croatia Case-co Male 26.5±4.3 (non-violent TC Violent ± ±40.99 et al, 2005 ntrol schizophrenia and suicidal group), normal controls 26.6±4.7 (non-violent Non-viole ± ±40.99 suicidal group), nt 26.3±4.6 (without suicide) Marcinko Croatia a Male 29.7±7.2 (suicidal group), TC ST ± ± ±48.72 et al, 2007 first episode of 29.6±8.7 (non-suicidal psychosis and group), normal controls 30.3±3.1 (controls) Modai et Israel Combined 53.1 (95% CI: ) TC ± ±51.40 al, 1994 psychiatric ( group), SI ± ±51.40 disease 53.7 (95% CI: ) (SI), 52.1 (95% CI: )(without suicide) Olie et al, France ±16.4 (suicidal group, TC a 148 a ±36.00 a 86 a ±43.00 a 1291 a ± psychiatric male, male), b 362 b ±37.00 b 189 b ±42.00 b 1131 b ±

14 disease and ±15.0 (suicidal group, TG a 148 a ±70.00 a 86 a ±88.00 a 1291 a ± normal controls female female), b 362 b ±53.00 b 189 b ±64.00 b 1131 b ± ±13.5 (non-suicidal group, male), 44.1±15.2 (non-suicidal group, female), 54.8±15.5 (controls, male), 53.7±16.2 (controls, female) Ozer et al, Turkey 12 male, 35.1±6.9 (suicidal patients), TC SI ± ± ± panic disorder and 32 female 32.3±10.7 (non-suicidal HDL-C SI ± ± ±9.80 normal controls patients) LDL-C SI ± ± ± ±8.7 (controls) TG SI ± ± ±68.80 Papadopo Greece Suicide attempters 48 male, 45.2±13.2 (suicide TC ± ±46.00 ulou et al, and normal 101 attempters), ± ± controls female 44.6±12.0 (controls) (follow up) Combined 49.1±12.3 (suicide TC ± ±35.00 major affective attempters), ± ±35.00 disorder 48.2±10.7 (controls) (follow up) Papassotir Germany 23 male, 45.0±18.0 (suicidal TC ± ± ±33.00 opoulos et affective disorder 29 female patients), al, 1999 and normal 46.0±12.0 (non-suicidal controls patients), 31.0±9.0 (controls) 14

15 schizophrenia and 17 male, 21 female normal controls 12 male, adjustment 20 female disorder and normal controls Park et al, Korea 16 male, 2014 major depressive 57 female disorder Perez-Rod Spain Case-co 426 male, riguez et ntrol psychiatric 504 al, 2008 disease and female normal controls 29.0±7.0 (suicidal patients), 31.0±9.0 (controls) TC ± ± ±13.0 (suicidal TC ± ±33.00 patients), 31.0±9.0 (controls) 43.0±18.6 (suicidal patients), 48.6±17.5 (non-suicidal patients) 47.4±19.7 (suicidal patients), 47.5±17.2 (non-suicidal patients) 39.7±14.0 (male suicide attempters), 35.4±13.6 (female suicide attempters), 41.3±13.3 (male psychiatric controls), 43.1±12.8 (female psychiatric controls), 34.9±10.8 (male controls), 34.9±11.9 (female controls) TC c SI ± ±38.09 HDL-C c SI ± ±34.56 LDL-C c SI ± ±34.56 TG c SI ± ±83.33 TC d SI ± ±38.06 HDL-C d SI ± ±14.04 LDL-C d SI ± ±35.07 TG d SI ± ±85.26 TC ± ± ±43.10 Plana et Spain Case-co 32 male, 15.4±2.0 (suicidal patients), TC ± ±

16 al, 2010 ntrol psychiatric disease depression Pompili et Italy al, 2010 etional mood disorders Rabe-Jabl Poland Patients in major onska et depression and al, 2000 remission 88 female 15.2±1.7 (non-suicidal patients) Combined Not reported TC ± ± male, 45.8±16.4 (suicidal 62 female patients), 41.8±13.7 (non-suicidal patients) 31 male, Not reported 71 female TC TG a 8 a ±38.40 a 43 a ±39.70 b 18 b ±36.70 b 44 b ±44.50 a 8 a ±60.60 a 43 a ±84.50 b 18 b ±41.20 b 44 b ±62.00 e TC ± ±25.60 ST ± ±25.60 f TC ± ±48.90 ST ± ±48.90 e LDL-C ± ±36.20 ST ± ±36.20 f LDL-C ± ±37.30 ST ± ±37.70 Roy et al, United Cocaine-dependen Male 41.6±8.1 (suicidal patients), TC ± ± States t patients 41.7±7.5 (non-suicidal patients) Roy et al, United 189 male, Not reported TC SI ± ± States diabetes 269 HDL-C SI ± ±18.00 female LDL-C SI ± ±40.00 Ruljancic Croatia Case-co 104 male, (suicidal patients), TC ± ± ±51.43 et al, 2011 ntrol depression and (non-suicidal normal controls female patients), (controls) 16

17 Sarchiapo Italy 6 male, 44.6±15.4(patients), TC ± ± ±25.00 ne et al, depression and 12 female 39.1±9.5 (controls) 2001 normal controls Sarchiapo Italy 62 male, 35.4±16.6(suicidal TC ± ±30.00 ne et al, psychiatric 178 patients), 2000 disease and female 35.2±15.5 (controls) normal controls Sullivan et New 39 male, 28.7±8.9 (patients with TC SI ± ±41.76 al, 1994 Zealand major depression 51 female suicidal ideation), TC ± ± ±8.9 (patients with TG SI ± ±17.01 suicidal attmpt), TG ± ± ±10.8 (non-suicidal patients) Tripodian Greece 66 male, 35.8±11.1 (patients with TC Violent ± ±35.00 akis et al, and normal 107 violent ), 2002 controls female 26.8±10.5 (patients with Non-viole ± ±35.00 non-violent ), nt 31.5±8.1 (controls) Verma et India 44 male, 20.3±2.5 (suicide TC ± ±23.77 al, 1999 and normal 36 female attempters), HDL-C ± ±9.17 controls 20.0±2.3 (controls) LDL-C ± ±22.05 TG ± ±32.01 Vevera et Czech Case-co Female 42.4±19.7 (patients with TC Violent ± ±52.98 al, 2003 Republic ntrol and normal violent ), controls 42.3±17.5 (patients with Non-viole ± ±52.98 non-violent ), nt 17

18 44.3±18.1 (controls) Vuksan-C Croatia Male 36.4±10.0 (suicidal TC ± ±61.10 usa et al, bipolar disorder patients), HDL-C ± ± ±14.4 (non-suicidal LDL-C ± ±51.04 patients) TG ± ±67.29 Zhang et United ±12.3 (male), TC SI a 151 a ±3.40 a 1269 a ±1.84 al, 2005 States and normal male, 29.0±9.6 (female) b 224 b ±3.92 b 1419 b ±2.14 controls 1770 a 47 a ±4.88 a 1269 a ±1.84 female b 127 b ±5.20 b 1419 b ±2.14 HDL-C SI a 151 a 43.85±1.32 a 1269 a 46.54±0.75 b 224 b 53.63±1.27 b 1419 b 55.73±0.63 a 47 a 49.91±3.46 a 1269 a 46.54±0.75 b 127 b 50.61±2.68 b 1419 b 55.73±0.63 LDL-C SI a 151 a ±3.28 a 1269 a ±1.75 b 224 b ±4.15 b 1419 b ±1.88 a 47 a ±5.63 a 1269 a ±1.75 b 127 b ±4.10 b 1419 b ±1.88 SD=standard deviation. TC=total cholesterol. HDL-C=high-density lipoprotein-cholesterol. LDL-C=low-density lipoprotein cholesterol. TG=triglycerides. SI=suicidal ideation. =suicide attempt. SC=suicidal tendencies. CS=committed suicide. a Number and serum lipid level for males. b Number and serum lipid level for females c Defined according to the score for the Hamilton Depression Rating Scale item 3 d Defined according to the score for the Beck Depression Inventory item 9 e Subjects in major depression f Subjects in remission 18

19 Table S3: Studies included in the systematic review and meta-analysis that assessed the risk of suicidality according to the lowest vs highest serum total cholesterol level category Study Bocchetta et al, 2001 Study site Study design Participants n Sex Outcome OR (95% CI) Factors adjusted for Italy Lithium patients male, Lifetime of a 1.82( ) None 496 female any b 0.95( ) Lifetime of violent a 2.8( ) b 1.32( ) Brunner et al, 2006 Germany Subjects with depressive symptoms de Leon et al, 2011 Spain Case-control psychiatric disease Ellison et al, 2001 Canada Cohort Subjects from general population Golier et al, United 1995 States psychiatric disease 1205 Combined 0.22( ) Age and gender male, 0.6( ) Sex, young age, low 515 female BMI, and clinical diagnoses male, CS 6.39( ) Sex and age 6238 female male, a 2.38( ) Age, weight, 343 female race/ethnicity, 0.91( ) socioeconomic status, diagnosis of major depression, and recent alcohol use 19

20 Lindberg et al, 1992 Sweden Cohort Subjects from general population Neaton et al, 1992 United Cohort Subjects from States general population Olie et al, 2011 France psychiatric disease and normal controls Papassotiropoulos et Germany al, 1999 psychiatric disease Partonen et al, 1999 Finland Cohort Subjects from general population Tanskanen et al, Finland Cohort Subjects from 2000 general population male, female CS 4.1( ) Age and follow up duration Male CS 1.61( ) Age, cigarettes smoked per day, diastolic blood pressure, race, season of year, and income male, c a 7.33(3-17.4) Age 1682 female b 15.6( ) d a 6.45( ) b 3.21( ) male, 6.67( ) Age, BMI, and gender 44 female Male CS 2.38( ) Age, BMI, carbohydrate intake, consumption of alcohol, education, marriage, self-reported amdety, self-reported depression and smoking male, female CS (violent suicides) CS (non-violent 0.42( ) 1.04( ) Sex, age, marital status, education, smoking, alcohol consumption, coffee drinking, physical activity, BMI, minor 20

21 Zureik et al, 1996 France Cohort Subjects from general population suicides) psychiatric symptoms, psychotropic medicaiton, coronary heart disease, and subjective general health 6393 Male CS 3.16( ) Age, smoking habits, and mean corpuscular volume at first examination OR=odds ratio. CI=confidence interval. =suicide attempt. CS=committed suicide a ORs and 95% CIs for males. b ORs and 95% CIs for females. c Compared with psychiatric controls. d Compared with normal controls 21

22 Table S4: Quality of the included studies (cohort studies) Study Representative ness of the exposed cohort Selection Comparability Outcome Demonstratio Comparability Selection n that of cohorts on of the non Ascertainment of outcome of the basis of the exposed exposure interest was design or cohort not present at analysis start of study Was follow-up Assessment long enough of outcome for outcomes to occur Adequacy of follow up of cohorts Overall quality score (of a maximum of 9) Coryell * Somewhat * Drawn * Structured * The study * Subjects * * The study * Subjects 8 et al, representative from the interview demonstrated don t show Independent select an lost to follow 2007 of the average same that suicide significant blind adequate up unlikely to population in communit was not differences in follow up introduce bias the y as the present at age in different period for (95% follow exposed start of study groups outcome of up) cohort interest Ellison * Truly * Drawn * Structured * The study * Study * * The study * Subjects 8 et al, representative from the interview demonstrated controls for age Independent select an lost to follow 2001 of the average same that suicide and sex blind adequate up unlikely to population in communit was not follow up introduce bias the y as the present at period for (99% follow exposed start of study outcome of up) cohort interest Fiedoro * Somewhat * Drawn * Structured * The study Study does not * * The study * Subjects 7 wicz et representative from the interview demonstrated control for Independent select an lost to follow al, 2007 of the average same that suicide other factors blind adequate up unlikely to 22

23 population in communit was not follow up introduce bias the y as the present at period for (100% follow exposed start of study outcome of up) cohort interest Lindberg * Turly * Drawn * Structured * The study * Study * * The study * Subjects 8 et al, representative from the interview demonstrated controls for age Independent select an lost to follow 1992 of the average same that suicide and follow up blind adequate up unlikely to population in communit was not duration follow up introduce bias the y as the present at period for (99% follow exposed start of study outcome of up) cohort interest Neaton * Turly * Drawn * Structured * The study ** Study * * The study * Subjects 9 et al, representative from the interview demonstrated controls for Independent select an lost to follow 1992 of the average same that suicide age, cigarettes blind adequate up unlikely to population in communit was not smoked per follow up introduce bias the y as the present at day, diastolic period for (97% follow exposed start of study blood pressure, outcome of up) cohort race, season of interest year, and income Partonen * Turly * Drawn * Structured * The study ** Study * * The study * Subjects 9 et al, representative from the interview demonstrated controls for Independent select an lost to follow 1999 of the average same that suicide age, BMI, blind adequate up unlikely to population in communit was not carbohydrate follow up introduce bias the y as the present at intake, period for (97% follow exposed start of study consumption of outcome of up) 23

24 cohort alcohol, interest education, marriage, self-reported amdety, self-reported depression and smoking Tanskan * Turly * Drawn * Structured * The study ** Study * * The study * Subjects 9 en et al, representative from the interview demonstrated controls for Independent select an lost to follow 2000 of the average same that suicide sex, age, blind adequate up unlikely to population in communit was not marital status, follow up introduce bias the y as the present at education, period for (100% follow exposed start of study smoking, outcome of up) cohort alcohol interest consumption, coffee drinking, physical activity, BMI, minor psychiatric symptoms, psychotropic medicaiton, coronary heart disease, and 24

25 subjective general health Zureik et * Turly * Drawn * Structured * The study * Study * * The study * Subjects 8 al, 1996 representative from the interview demonstrated controls for Independent select an lost to follow of the average same that suicide age, smoking blind adequate up unlikely to population in communit was not habits, and follow up introduce bias the y as the present at mean period for (83% follow exposed start of study corpuscular outcome of up) cohort volume at first interest examination 25

26 Table S5: Quality of the included studies (case-control studies) Selection Comparability Exposure Comparability Same method Overall quality of cases and Is the case Representative of Study Selection of Definition of controls on the Ascertainment Non-Respo definition ness of the ascertainment score (of a controls Controls basis of the of exposure nse rate adequate? cases for cases and maximum of 9) design or controls analysis de Leon * Yes, with * Somewhat * Drawn from * No history ** The study * Secure record * Yes * Same rate 9 et al, 2011 independent validation representative of the average population in the the same as the cases of disease controls for sex, young age, low BMI, and clinical diagnoses for both groups Diaz-Sas * Yes, with * Somewhat * Drawn from * No history * Subjects are * Secure record * Yes * Same rate 8 tre et al, 2007 independent validation representative of the average population in the the same as the cases of disease matched by gender, age, and BMI for both groups Kunugi * Yes, with * Somewhat * Drawn from * No history Study does not * Secure record * Yes * Same rate 7 et al, 1997 independent validation representative of the average population in the the same as the cases of disease control for other factors for both groups Marcink * Yes, with No description * Drawn from * No history ** Subjects * Secure record * Yes * Same rate 8 26

27 o et al, independent the same of disease don t show for both 2005 validation as significant groups the cases differences in age, BMI, marital status, family status, and employment status in different groups Perez-R * Yes, with * Truly * Drawn from * No history Study does not * Secure record * Yes * Same rate 7 odriguez independent representative the same of disease control for for both et al, validation of the average as other factors groups 2008 population in the cases the Plana et * Yes, with * Somewhat * Drawn from * No history * Subjects are * Secure record * Yes * Same rate 8 al, 2010 independent representative the same of disease matched by for both validation of the average as psychiatric groups population in the cases diagnosis, age, the and gender Ruljanci * Yes, with * Somewhat * Drawn from * No history * Subjects are * Secure record * Yes * Same rate 8 c et al, independent representative the same of disease matched by for both 2011 validation of the average as age and sex groups population in the cases the 27

28 Vevera * Yes, with * Somewhat * Drawn from * No history * Subjects are * Secure record * Yes * Same rate 8 et al, independent representative the same of disease matched by for both 2003 validation of the average as age and groups population in the cases diagnosis with the no history of suicide 28

29 Table S6: Quality of the included studies (cross-s studies) Study Selection Comparability Exposure Overall quality Representativeness of the sample Ascertainment of Comparability of groups on the Assessment of outcome exposure basis of the design or analysis score (of a maximum of 5) Ainiyet et al, 2014 * Somewhat representative of the * Structured interview Study does not control for other 3 factors Almeida-Montes No description * Structured interview * Subjects don t show 3 et al, 2000 significant differences in age and BMI in different groups Alvarez et al, 2000 * Somewhat representative of the * Structured interview * Study controls for sex ratio 4 and age Alvarez et al, 1999 * Somewhat representative of the * Structured interview *Subjects are matched by sex, 4 age, and month of inclusion Apter et al, 1999 * Somewhat representative of the * Structured interview * Subjects don t show 4 significant differences in age, BMI, and education in different groups Atmaca et al, 2002 No description * Structured interview * Study controls for age or BMI 3 Atmaca et al, 2003 No description * Structured interview *Subjects are matched by sex and age 3 29

30 Atmaca et al, 2008 No description * Structured interview * Subjects are matched by sex and age Atmaca et al, 2002 No description * Structured interview * Subjects don t show significant differences in age, sex, and BMI in different groups Atmaca et al, 2006 No description * Structured interview * Subjects don t show significant differences in age, sex, and BMI in different groups Baek et al, 2014 * Somewhat representative of the * Structured interview ** Subjects don t show significant differences in education, sex, employment status, and married status in different groups Bocchetta et al, 2001 Brunner et al, 2006 da Graça Cantarelli et al, 2014 D'Ambrosio et al, 2012 * Somewhat representative of the * Structured interview Study does not control for other factors * Truely representative of the * Structured interview * Study controls for age and gender No description * Structured interview * Subjects don t show significant differences in age and education * Somewhat representative of the * Structured interview ** Subjects don t show significant differences in age, education, sex, age of disease onset, duration of illness, family

31 history of mood disorders, family history of completed suicide, and BMI in different groups Deisenhammer et * Somewhat representative of the No description * Subjects don t show 3 al, 2004 significant differences in weight and BMI in different groups Deisenhammer et * Somewhat representative of the * Structured interview ** Subjects don t show 5 al, 2006 significant differences in sex, BMI, history of a major depressive episode, age at onset of alcohol dependence, and duration of problematic alcohol consumption per day in different groups Favaro et al, 2004 * Somewhat representative of the * Structured interview Study does not control for other 3 factors Gallerani et al, * Somewhat representative of the * Structured interview ** Subjects don t show significant differences in age, sex, BMI, drug misuse, diabetes, hypertension, alcoholism, and heterozygote thalassaemia in different groups Golier et al, 1995 * Somewhat representative of the * Structured interview ** Study controls for age, 5 31

32 weight, race/ethnicity, socioeconomic status, diagnosis of major depression, and recent alcohol use Guillem et al, 2002 * Somewhat representative of the * Structured interview Study does not control for other 3 factors Huang et al, 2000 * Somewhat representative of the * Structured interview Study does not control for other 3 factors Huang, 2005 * Somewhat representative of the * Structured interview Study does not control for other 3 factors Jokinen et al, 2010 No description * Structured interview * Subjects don t show 3 significant differences in BMI in different groups Kim et al, 2014 * Somewhat representative of the * Structured interview ** Subjects don t show 5 significant differences in gender, education, past manual occupation, current problem drinker, BMI, hypertension, and diabetes in different groups Kim et al, 2002 * Somewhat representative of the * Structured interview * Subjects don t show 4 significant differences in sex and age in different groups Kim et al, 2004 * Somewhat representative of the * Structured interview * Subjects are matched by sex, 4 32

33 age, and BMI Lee et al, 2003 * Somewhat representative of the * Structured interview * Subjects are matched by diagnosis, sex, age, and BMI Maes et al, 1997 No description * Structured interview Study does not control for other factors Marcinko et al, * Somewhat representative of the * Structured interview * Subjects don t show 2011 significant differences in age and BMI in different groups Marcinko et al, * Somewhat representative of the * Structured interview * Subjects don t show 2008 significant differences in age, BMI, previous suicide, and previous hospitalization in different groups Marcinko et al, * Somewhat representative of the * Structured interview ** Subjects don t show 2007 significant differences in age, BMI, family status, church attending, and family history of suicide or suicide attempt in different groups Modai et al, 1994 * Somewhat representative of the * Structured interview * Subjects are matched by age and sex Olie et al, 2011 * Somewhat representative of the * Structured interview * Study controls for age

34 Ozer et al, 2004 No description * Structured interview * Subjects don t show significant differences in age and weight in different groups Papadopoulou et * Somewhat representative of the * Structured interview * Subjects are matched by age al, 2013 and sex Papassotiropoulos No description * Structured interview * Study controls for age, BMI, et al, 1999 and gender Park et al, 2014 * Somewhat representative of the * Structured interview * Subjects don t show significant differences in age, sex, and BMI in different groups Pompili et al, 2010 * Somewhat representative of the * Structured interview * Subjects don t show significant differences in age and sex in different groups Rabe-Jablonska et * Somewhat representative of the * Structured interview Study does not control for other al, 2000 factors Roy et al, 2001 * Somewhat representative of the * Structured interview * Subjects don t show significant differences in age in different groups Roy et al, 2006 * Somewhat representative of the * Structured interview Study does not control for other factors Sarchiapone et al, No description * Structured interview Study does not control for other 2001 factors

35 Sarchiapone et al, * Somewhat representative of the * Structured interview * Subjects are matched by age and sex Sullivan et al, * Somewhat representative of the * Structured interview * Subjects don t show significant differences in BMI in different groups Tripodianakis et al, * Somewhat representative of the * Structured interview * Subjects don t show significant differences in age in different groups Verma et al, 1999 * Somewhat representative of the * Structured interview * Subjects are matched by sex 4 and age Vuksan-Cusa et al, * Somewhat representative of the * Structured interview ** Subjects don t show significant differences in age, duration of illness, BMI, waist circumference, GUK, CGI, BPRS score, HDRS-17 score, and YMRS score in different groups Zhang et al, 2005 * Truly representative of the * Structured interview Subjects does not control for 3 other factors 35

36 Figure S1: Forest plot displaying the pooled summary estimates of serum TC levels in the suicidal patients versus non-suicidal patients (A), and in the suicidal patients versus normal controls (B), after exclusion of studies not controlled for at least age or sex. A 36

37 B 37

38 Figure S2: Forest plot displaying the pooled summary estimate of serum HDL-C levels in the suicidal patients versus normal controls, after exclusion of studies not controlled for at least age or sex. 38

39 Figure S3: Forest plot displaying the pooled summary estimates of serum LDL-C levels in the suicidal patients versus non-suicidal patients (A), and in the suicidal patients versus normal controls (B), after exclusion of studies not controlled for at least age or sex. A 39

40 B 40

41 Figure S4: Forest plot displaying the pooled summary estimates of serum TG levels in the suicidal patients versus non-suicidal patients (A), and in the suicidal patients versus normal controls (B), after exclusion of studies not controlled for at least age or sex. A 41

42 B 42

43 Figure S5: Forest plot displaying the pooled relative risk of suicidality according to the lowest vs highest serum TC level category, after exclusion of studies not controlled for at least age or sex. 43

44 Figure S6: Funnel plots to assess publication bias Plots show study size as a function of effect size for studies included in the meta-analysis. A. For studies assessing serum total cholesterol level between suicidal patients and non-suicidal patients. Egger s regression test (p=0.746), and Begg-Mazumdar test (p=0.093) B. For studies assessing serum total cholesterol level between suicidal patients and normal controls. Egger s regression test (p<0.001), and Begg-Mazumdar test (p=0.129) 44

45 C. For studies assessing serum high-density lipoprotein cholesterol level between suicidal patients and non-suicidal patients. Egger s regression test (p=0.997), and Begg-Mazumdar test (p=0.714) D. For studies assessing serum high-density lipoprotein cholesterol level between suicidal patients and normal controls. Egger s regression test (p=0.834), and Begg-Mazumdar test (p=0.929) 45

46 E. For studies assessing serum low-density lipoprotein cholesterol level between suicidal patients and non-suicidal patients. Egger s regression test (p=0.003), and Begg-Mazumdar test (p=0.045) F. For studies assessing serum low-density lipoprotein cholesterol level between suicidal patients and normal controls. Egger s regression test (p=0.352), and Begg-Mazumdar test (p=0.586) 46

47 G. For studies assessing serum triglycerides level between suicidal patients and non-suicidal patients. Egger s regression test (p=0.033), and Begg-Mazumdar test (p=0.632) H. For studies assessing serum triglycerides level between suicidal patients and normal controls. Egger s regression test (p=0.215), and Begg-Mazumdar test (p=0.677) 47

48 I. For studies assessing the risk of suicide according to the lowest vs highest serum total cholesterol category. Egger s regression test (p=0.399), and Begg-Mazumdar test (p=0.553) 48