Table 1.

Study characteristics and results.

Study	Design	N total: studies, participants (cases; controls where relevant)	Serotonin measure	Medication status	Summary effect (95% CI), p Heterogeneity-p, I2 (95% CI), N for analysis (if different from total sample size)	Quality rating (AMSTAR-2/ *STREGA) % satisfactory (of applicable terms)
Serotonin and 5HIAA: case (depression) vs. healthy control studies comparing concentration of serotonin in plasma and CSF
Huang et al., 2020	Meta-analysis of cohort studies	3 studies, 663 women with depression; 1806 controls	5-HT in plasma	18.3% of patients were taking antidepressants.	β = −0.26 (−0.48 to −0.03), p > 0.05 after correction for multiple testing Heterogeneity not reported Sensitivity analysis: No difference in levels of 5-HT in women with depression not taking antidepressants compared to controls (women without depression not taking antidepressants) (p = 0.528) Lower levels of 5-HT in women with depression taking antidepressants compared to controls (p < 0.0001) and in women taking antidepressants without depression compared to controls (p < 0.0001)	60%
Ogawa et al., 2018	Systematic review and meta-analysis	11 studies, 435 people with depression; 380 controls	5-HIAA in CSF	All participants drug free – washout: 4–57 days (mode 14 days)	g = 0.042 (−0.26 to 0.17), p = 0.70, heterogeneity: I2 = 48.24 (p = 0.026)	48%
Pech et al., 2018	Systematic review and meta-analysis	13 studies, 529 people with depression; 473 controls	5-HIAA in CSF	10 studies: patients were drug free - washout: 4 days-6 weeks (mostly 2–4 weeks). 2 studies: some patients used occasional antidepressants. 1 study: majority of patients taking imipramine and chlorpromazine	Mean difference =  −3.85 (−8.89 to 1.19) (g = 0.06), p = 0.14, heterogeneity – not reported	38%
Receptors: case (depression) vs. healthy control studies comparing 5-HT1 receptor binding
Nikolaus et al., 2016	Systematic review and meta-analysis	14 studies, 245 people with depression; 316 controls	5-HT1A receptor binding	2 studies conducted with drug naïve patients (total N = 19). 11 studies - prior treatment with antidepressants and/or other psychiatric drugs – washout: 1–1825 days (mode 14 days) 1 study medication not specified	Trend for increased 5-HT1 in parahippocampal gyrus (+23%), p = 0.096. Trend for reduced 5-HT1 in midbrain (−17%), p = 0.076. N and heterogeneity – not reported for any brain areas.	19%
Wang et al., 2016	Systematic review and meta-analysis	10 studies, 218 people with depression; 261 controls	5-HT1A receptor binding	9 studies: drug-free – washout: 2–26 weeks (mainly 2–3 weeks). 1 study: medication naïve.	Reduced 5-HT1A binding mesiotemporal cortex SMD = −0.8 (−1.36, −0.24), p = 0.005, heterogeneity: I2 = 87% (P < 0.00001) 218 patients, 261 controls** Reduced 5-HT1A binding hippocampus SMD  = −0.29 (−0.51, −0.07), P = 0.010, heterogeneity: I2 = 33%, p = 0.18 148 patients, 203 controls. Reduced 5-HT1A binding raphe nucleus SMD  =  −0.60 (−1.17, −0.04), p = 0.04 Heterogeneity: I2 = 88%, p < 0.00001 218 patients, 261 controls Reduced 5-HT1A binding insular cortex SMD  =  −0.79 (−1.54, to −0.05), p = 0.04. Heterogeneity: I2 = 91%, p < 0.00001 180 patients, 225 controls. There was no significant difference in the occipital cortex and the anterior cingulate cortex.	52%
Serotonin transporter (SERT): case (depression) vs. healthy control studies comparing SERT binding
Nikolaus et al., 2016	Systematic review and meta-analysis	35 studies, 694 people with depression; 700 controls	SERT binding	26 studies: drug-free - washout: 3–900 days. 5 studies: drug naive 2 studies: patients medicated when scanned. 2 studies: did not report medication.	Reduced SERT in thalamus −12%, p = 0.004 Reduced SERT in amygdala −15%, p = 0.04. Reduced SERT in midbrain/ pons −8%, p = 0.005 Increased SERT in insula +9%, p = 0.037 N, CI and heterogeneity – not reported for any brain areas.	19%
Kambeitz & Howes., 2015	Systematic review and meta-analysis	25 in vivo imaging studies, 25 post-mortem studies, 877 people with depression; 968 controls	SERT binding	In vivo studies: three studies – antidepressant naïve patients. Post Mortem studies: medication status not reported	In vivo Reduced SERT in brainstem g = −0.31 (−0.55 to −0.08), p = 0.01, heterogeneity - I2 = 60.69% (95% CI: 34.39 to 83.35%), N = 880 Reduced SERT in midbrain g = −0.28 (−0.49 to −0.07), p = 0.01, heterogeneity - I2 = 49.68% (95% CI: 14.34% to 78.72%), N = 827 Reduced SERT in amygdala g = −0.37*** (−0.61 to −0.13), p<0.01, heterogeneity - I2 = 0% (95%-CI: 0% to 75.38%), N = 318 Reduced SERT in striatum g = −0.39 (−0.62 to −0.17), P<.001, heterogeneity - I2 = 6.7% (0% to 78.1%), N = 370 No difference: hippocampus, thalamus, cingulate cortex, frontal cortex. Post-mortem: Reduced SERT in hippocampus g = −0.63 (−1.12 to −0.15), p = 0.01 heterogeneity - I2 = 43.97% (0% to 93.34), N = 141 Non-significant after correction for publication bias; g = 0.32, p = 0.32 No difference: brainstem, frontal lobe. Meta-analysis not possible for other areas	48%
Gryglewski et al., 2014	Systematic review and meta-analysis	18 studies, 364 people with depression; 372 controls	SERT binding	149/364 patients - drug naïve; others drug free: 5 days to > 1 year (median 7 weeks)	Reduced SERT in midbrain: g =  −0.49 (−0.84 to −0.14), heterogeneity - I2 =  68.7%, 313 patients, 321 controls. Reduced SERT in amygdala g =  −0.50 (−0.78 to −0.22), heterogeneity - I2 = 0%, 96 patients, 128 controls. No difference: brainstem, thalamus, striatum, frontal cortex, cingulate cortex.	48%
Tryptophan depletion studies: effect of acute tryptophan depletion (ATD) on mood in healthy volunteers; healthy volunteers with family history of depression; drug-free patients with MDD in remission
Ruhé et al., 2007	Systematic review and meta-analysis	Included in meta-analysis: 32 healthy volunteer studies, N = 566; 19 patient studies N = 322	ATD	6 studies involved patients with prior use of antidepressants: 1 study: <3 months; 2 studies 1–3 months 3 studies >6 months 7 studies involved patients with current use of antidepressants. Remainder unspecified.	No effect of ATD in healthy volunteers in parallel group studies (negative for family history of depression): Pooled g = −0.63 (−1.95 to 0.70), N = 151. Sensitivity analysis excluding an outlier study reduced Hedges g to 0.16 (−0.43 to 0.76), N = 125. No effect of ATD in cross-over studies with volunteers with no family history of depression, g = −0.19 (−0.43 to 0.05), p = 0.13, heterogeneity - I2 = 65.6%, p < 0.001, N = 259# ATD lowered mood in cross-over studies with healthy controls with family history of depression: Hedge’s g = −0.56 (−1.00 to −0.13), p = 0.01, heterogeneity - I2 = 50.8%, p = 0.06, N = 75. ATD lowered mood in cross-over studies in drug-free people with remitted MDD: g = −1.90 (−3.02 to −0.78), p = 0.0009, heterogeneity - I2 =  89.4%, P< 0.00001, N = 97. ATD lowered mood in cross-over studies with people with remitted MDD currently using antidepressants : g = −0.49 (−0.89 to −0.10), p = 0.01, heterogeneity - I2 =  53.3%, P = 0.04, N = 83.	71%
Fusar-Poli et al., 2006	Systematic review and narrative synthesis	22 studies (23 cohorts), 64 people with remitted depression; 278 controls	ATD	A portion of the remitted depressed group were taking antidepressants but exact proportion not reported	Results reported narratively. 17/19 studies involving healthy volunteers showed no effect of tryptophan depletion on mood. 4 studies in patients with remitted depression, an unspecified portion of whom were taking antidepressants, found a decrease in mood following tryptophan depletion. Effect sizes or statistical significance were not reported.	22%
SERT gene: association between SERT gene and depression
Border et al., 2019	Genetic association study	Data from two genetic data banks, 48,190–115,257 individuals	Association between 5-HTTLPR polymorphism and depression	N/A	No relationship between 5-HTTLPR polymorphism and estimated lifetime MDD diagnosis OR = 1.000, p = 0.994 (N = 115,257) None of seven other depression outcomes (e.g. estimated lifetime diagnosis, current depression severity) showed an association with the 5-HTTLPR polymorphism. N = 48,190–115,257	88%*
Culverhouse et al., 2018	Collaborative meta-analysis	31 data-sets, 43,165 Individuals	5-HTTLPR polymorphism association with depression	N/A	No significant effect of number of copies of s-allele of 5HTTLPR on lifetime depression. OR 1.00 (0.95 to 1.05), p = 0.95, N = 21,135. Heterogeneity- not reported All other analyses with variations of stress exposure and depression evaluation were non-significant with OR 1.00–1.08, p = 0.36–0.97, N = 13,835–28,252	100%
Oo et al., 2016	Systematic review and meta-analysis	23 studies, 3392 people with depression; 5093 controls	5-HTTLPR polymorphism association with depression	N/A	5-HTTLPR polymorphism associated with depression (per S allele): OR = 1.16 (1.08 to 1.23), p-value. Heterogeneity- I2 =  29.3%, p = 0.09 Homozygote carriers of the S allele of 5HTTLPR polymorphism compared with heterozygote and non-carriers combined (SS vs SL+LL genotype): OR = 1.33 (1.19 to 1.48) for major depressive disorder. Heterogeneity- I2 =  0.1%, p = 0.46	62%
Gatt et al., 2015	Umbrella review	11 meta-analyses included 1014–14,250 participants	7 meta-analyses of case control studies of 5-HTTLPR polymorphism and depression. 4 meta-analyses of GWAS	N/A	Individual gene meta-analyses: 3 studies found the S allele or SS genotype associated with depression in mixed and Caucasian samples (OR 1.11 (1.04–1.19), N  = 9459 OR 1.23 (1.01–1.52), N = 1014 OR 1.40 (1.19–1.65), N = 6884) 3 studies reported no effect of the S allele in mixed and Caucasian samples 2 studies reported no effect of the SS genotype in Caucasian samples 3 studies found that the S allele or SS genotype had no effect in Asian samples No evidence of association in all 4 GWAS meta-analyses (n = 6566 to n = 12,664)	17%
Kiyohara & Yoshimasu, 2010	Systematic review and meta-analysis	22 studies, 2934 people with depression; 4985 controls	5-HTTLPR polymorphism association with depression	N/A	5-HTTLPR polymorphism (SS vs LL) associated with depression in Caucasian samples (but not Asian samples): OR =  1.41 (1.15 to 1.72), Heterogeneity p = 0.23 (N = 5756)	57%
Gene-stress interaction
Border et al., 2019	See Border et al., 2019 (above)	See Border et al., 2019 (above)	5-HTTLPR Polymorphism x environmental interaction effect on depression	N/A	No significant interaction between the 5-HTTLPR polymorphism and childhood trauma. Exp(β) = 0.998, p = 0.919, N = 115,249. Heterogeneity not given as not a meta-analysis. 31 other analyses using different measures of stress or depression found no significant interaction.	See Border et al., 2019 (above)
Bleys et al 2018	Systematic review and meta-analysis	51 studies, N = 51,449	Interaction between stress and 5-HTTLPR polymorphism in depression	N/A	OR = 1.18 (1.09 to 1.28). Heterogeneity- I2 =  52.4%, p < 0.0001	33%
Culverhouse et al., 2018	See Culverhouse et al., 2018 (above)	See Culverhouse et al., 2018 (above)	Interaction between stressful life events and 5-HTTLPR polymorphism for risk of depression	N/A	No significant interaction between number of copies of the s-allele for 5-HTTLPR and stress on risk of depression. OR = 1.05 (0.91 to 1.21), p = 0.49, N = 21,135 I2 = 0.0, p = 0.69 All other analyses with variations of stress and depression evaluation (including continuous measures of stress) were non-significant, with ORs from 0.85 to 1.06, p = 0.17–0.50, N = 13,835-28,252 I2 =  0–16.7, p = 0.26-0.87	See Culverhouse et al., 2018 (above)
Sharpley et al 2014	Systematic review and meta-analysis	81 studies, N = 55,269	Interaction between stress and 5-HTTLPR polymorphism in depression	N/A	OR not reported - Liptak-Stouffer Z-score method was used which combined p-values across studies, p = 9 × 10−7 Heterogeneity- not reported (N = 55,269)	48%
Karg et al 2011	Systematic review and meta-analysis	54 studies, N = 40,749	Whether the 5-HTTLPR polymorphism moderates the relationship between stress and depression	N/A	OR not reported - Liptak-Stouffer Z-score method was used which combined p-values across studies, p = 0.00002. Heterogeneity- not reported (N = 40,749)	29%

Extracted summary effects, confidence intervals and measures of statistical significance are reported. In reviews in which multiple meta-analyses were conducted, only summary effect sizes that were statistically significant are reported, with relevant negative findings that did not reach statistical significance listed. Where no relevant analysis was statistically significant, the results of the principal analysis (e.g., main gene effect) is reported. Data on heterogeneity is reported where relevant. Where sample size for a particular analysis was different from the total sample size, this is reported. For summary effects in the non-genetic studies, preference was given to the extraction and reporting of effect sizes. Mean differences were converted to effect sizes where appropriate data were available [25]. Where these data were not available the most relevant measure, such as the beta co-efficient, or average percentage change in ligand binding was reported. For genetic association studies, and gene-stress interactions studies we odds ratios are reported.

CSF Cerebrospinal fluid, g Hedges g, GWAS Genome Wide Association Studies, MDD Major Depressive Disorder, N/A not applicable, OR Odds Ratio, SMD Standard Mean Difference.

*STREGA rating; **where there were internal discrepancies between reported numbers in text or figures we obtained data from the original authors; ***where there were internal discrepancies between reported numbers in the text, and authors did not respond to queries, details from figures were prioritised, and statistics calculated from other data.

^#sample size has a different meaning in crossover studies since each participant contributes data twice.