Microbial genome-wide association studies: lessons from human GWAS.

The reduced costs of sequencing have led to whole-genome sequences for a large number of microorganisms, enabling the application of microbial genome-wide association studies (GWAS). Given the successes of human GWAS in understanding disease aetiology and identifying potential drug targets, microbial GWAS are likely to further advance our understanding of infectious diseases. These advances include insights into pressing global health problems, such as antibiotic resistance and disease transmission. In this Review, we outline the methodologies of GWAS, the current state of the field of microbial GWAS, and how lessons from human GWAS can direct the future of the field.

RNASeq_similarity_matrix: visually identify sample mix-ups in RNASeq data using a 'genomic' sequence similarity matrix.

MOTIVATION: Mistakes in linking a patient's biological samples with their phenotype data can confound RNA-Seq studies. The current method for avoiding such sample mixups is to test for inconsistencies between biological data and known phenotype data such as sex. However, in DNA studies a common QC step is to check for unexpected relatedness between samples. Here, we extend this method to RNA-Seq, which allows the detection of duplicated samples without relying on identifying inconsistencies with phenotype data. SUMMARY: We present RNASeq_similarity_matrix: an automated tool to generate a sequence similarity matrix from RNA-Seq data, which can be used to visually identify sample mix-ups. This is particularly useful when a study contains multiple samples from the same individual, but can also detect contamination in studies with only one sample per individual. AVAILABILITY: RNASeq_similarity_matrix has been made available as a documented GPL licensed Docker image on www.github.com/nicokist/RNASeq_similarity_matrix. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Heritability of Individual Psychotic Experiences Captured by Common Genetic Variants in a Community Sample of Adolescents.

Occurrence of psychotic experiences is common amongst adolescents in the general population. Twin studies suggest that a third to a half of variance in adolescent psychotic experiences is explained by genetic influences. Here we test the extent to which common genetic variants account for some of the twin-based heritability. Psychotic experiences were assessed with the Specific Psychotic Experiences Questionnaire in a community sample of 2152 16-year-olds. Self-reported measures of Paranoia, Hallucinations, Cognitive Disorganization, Grandiosity, Anhedonia, and Parent-rated Negative Symptoms were obtained. Estimates of SNP heritability were derived and compared to the twin heritability estimates from the same sample. Three approaches to genome-wide restricted maximum likelihood (GREML) analyses were compared: (1) standard GREML performed on full genome-wide data; (2) GREML stratified by minor allele frequency (MAF); and (3) GREML performed on pruned data. The standard GREML revealed a significant SNP heritability of 20 % for Anhedonia (SE = 0.12; p < 0.046) and an estimate of 19 % for Cognitive Disorganization, which was close to significant (SE = 0.13; p < 0.059). Grandiosity and Paranoia showed modest SNP heritability estimates (17 %; SE = 0.13 and 14 %; SE = 0.13, respectively, both n.s.), and zero estimates were found for Hallucinations and Negative Symptoms. The estimates for Anhedonia, Cognitive Disorganization and Grandiosity accounted for approximately half the previously reported twin heritability. SNP heritability estimates from the MAF-stratified approach were mostly consistent with the standard estimates and offered additional information about the distribution of heritability across the MAF range of the SNPs. In contrast, the estimates derived from the pruned data were for the most part not consistent with the other two approaches. It is likely that the difference seen in the pruned estimates was driven by the loss of tagged causal variants, an issue fundamental to this approach. The current results suggest that common genetic variants play a role in the etiology of some adolescent psychotic experiences, however further research on larger samples is desired and the use of MAF-stratified approach recommended.

A twin study of breastfeeding with a preliminary genome-wide association scan.

Breastfeeding has been an important survival trait during human history, though it has long been recognized that individuals differ in their exact breastfeeding behavior. Here our aims were, first, to explore to what extent genetic and environmental influences contributed to the individual differences in breastfeeding behavior; second, to detect possible genetic variants related to breastfeeding; and lastly, to test if the genetic variants associated with breastfeeding have been previously found to be related with breast size. Data were collected from a large community-based cohort of Australian twins, with 3,364 women participating in the twin modelling analyses and 1,521 of them included in the genome-wide association study (GWAS). Monozygotic (MZ) twin correlations (r MZ = 0.52, 95% CI 0.46-0.57) were larger than dizygotic (DZ) twin correlations (r DZ = 0.35, 95% CI 0.25-0.43) and the best-fitting model was the one composed by additive genetics and unique environmental factors, explaining 53% and 47% of the variance in breastfeeding behavior, respectively. No breastfeeding-related genetic variants reached genome-wide significance. The polygenic risk score analyses showed no significant results, suggesting breast size does not influence breastfeeding. This study confers a replication of a previous one exploring the sources of variance of breastfeeding and, to our knowledge, is the first one to conduct a GWAS on breastfeeding and look at the overlap with variants for breast size.

A recessive genetic model and runs of homozygosity in major depressive disorder.

Genome-wide association studies (GWASs) of major depressive disorder (MDD) have yet to identify variants that surpass the threshold for genome-wide significance. A recent study reported that runs of homozygosity (ROH) are associated with schizophrenia, reflecting a novel genetic risk factor resulting from increased parental relatedness and recessive genetic effects. Here, we explore the possibility of such a recessive model in MDD. In a sample of 9,238 cases and 9,521 controls reported in a recent mega-analysis of 9 GWAS we perform an analysis of ROH and common variants under a recessive model. Since evidence for association with ROH could reflect a recessive mode of action at loci, we also conducted a genome-wide association analyses under a recessive model. The genome-wide association analysis using a recessive model found no significant associations. Our analysis of ROH suggested that there was significant heterogeneity of effect across studies in effect (P = 0.001), and it was associated with genotyping platform and country of origin. The results of the ROH analysis show that differences across studies can lead to conflicting systematic genome-wide differences between cases and controls that are unaccounted for by traditional covariates. They highlight the sensitivity of the ROH method to spurious associations, and the need to carefully control for potential confounds in such analyses. We found no strong evidence for a recessive model underlying MDD.

Genetic relationships between suicide attempts, suicidal ideation and major psychiatric disorders: a genome-wide association and polygenic scoring study.

Epidemiological studies have recognized a genetic diathesis for suicidal behavior, which is independent of other psychiatric disorders. Genome-wide association studies (GWAS) on suicide attempt (SA) and ideation have failed to identify specific genetic variants. Here, we conduct further GWAS and for the first time, use polygenic score analysis in cohorts of patients with mood disorders, to test for common genetic variants for mood disorders and suicide phenotypes. Genome-wide studies for SA were conducted in the RADIANT and GSK-Munich recurrent depression samples and London Bipolar Affective Disorder Case-Control Study (BACCs) then meta-analysis was performed. A GWAS on suicidal ideation during antidepressant treatment had previously been conducted in the Genome Based Therapeutic Drugs for Depression (GENDEP) study. We derived polygenic scores from each sample and tested their ability to predict SA in the mood disorder cohorts or ideation status in the GENDEP study. Polygenic scores for major depressive disorder, bipolar disorder and schizophrenia from the Psychiatric Genomics Consortium were used to investigate pleiotropy between psychiatric disorders and suicide phenotypes. No significant evidence for association was detected at any SNP in GWAS or meta-analysis. Polygenic scores for major depressive disorder significantly predicted suicidal ideation in the GENDEP pharmacogenetics study and also predicted SA in a combined validation dataset. Polygenic scores for SA showed no predictive ability for suicidal ideation. Polygenic score analysis suggests pleiotropy between psychiatric disorders and suicidal ideation whereas the tendency to act on such thoughts may have a partially independent genetic diathesis.

Genome-wide association analysis accounting for environmental factors through propensity-score matching: application to stressful live events in major depressive disorder.

Stressful life events are an established trigger for depression and may contribute to the heterogeneity within genome-wide association analyses. With depression cases showing an excess of exposure to stressful events compared to controls, there is difficulty in distinguishing between "true" cases and a "normal" response to a stressful environment. This potential contamination of cases, and that from genetically at risk controls that have not yet experienced environmental triggers for onset, may reduce the power of studies to detect causal variants. In the RADIANT sample of 3,690 European individuals, we used propensity score matching to pair cases and controls on exposure to stressful life events. In 805 case-control pairs matched on stressful life event, we tested the influence of 457,670 common genetic variants on the propensity to depression under comparable level of adversity with a sign test. While this analysis produced no significant findings after genome-wide correction for multiple testing, we outline a novel methodology and perspective for providing environmental context in genetic studies. We recommend contextualizing depression by incorporating environmental exposure into genome-wide analyses as a complementary approach to testing gene-environment interactions. Possible explanations for negative findings include a lack of statistical power due to small sample size and conditional effects, resulting from the low rate of adequate matching. Our findings underscore the importance of collecting information on environmental risk factors in studies of depression and other complex phenotypes, so that sufficient sample sizes are available to investigate their effect in genome-wide association analysis.

Dissecting the genetic heterogeneity of depression through age at onset.

Genome-wide studies in major depression have identified few replicated associations, potentially due to heterogeneity within the disorder. Several studies have suggested that age at onset (AAO) can distinguish sub-types of depression with specific heritable components. This paper investigates the role of AAO in the genetic susceptibility for depression using genome-wide association data on 2,746 cases and 1,594 screened controls from the RADIANT studies, with replication performed in 1,471 cases and 1,403 controls from two Munich studies. Three methods were used to analyze AAO: First a time-to-event analysis with controls censored, secondly comparing controls to case-subsets defined using AAO cut-offs, and lastly analyzing AAO as a quantitative trait. In the time-to-event analysis three SNPs reached suggestive significance (P < 5E-06), overlapping with the original case-control analysis of this study. In a case-control analysis using AAO thresholds, SNPs in 10 genomic regions showed suggestive association though again none reached genome-wide significance. Lastly, case-only analysis of AAO as a quantitative trait resulted in 5 SNPs reaching suggestive significance. Sex specific analysis was performed as a secondary analysis, yielding one SNP reaching genome-wide significance in early-onset males. No SNPs achieved significance in the replication study after correction for multiple testing. Analysis of AAO as a quantitative trait did suggest that, across all SNPs, common genetic variants explained a large proportion of the variance (51%, P = 0.04). This study provides the first focussed analysis of the genetic contribution to AAO in depression, and establishes a statistical framework that can be applied to a quantitative trait underlying any disorder.

GWAS of Suicide Attempt in Psychiatric Disorders and Association With Major Depression Polygenic Risk Scores.

OBJECTIVE: More than 90% of people who attempt suicide have a psychiatric diagnosis; however, twin and family studies suggest that the genetic etiology of suicide attempt is partially distinct from that of the psychiatric disorders themselves. The authors present the largest genome-wide association study (GWAS) on suicide attempt, using cohorts of individuals with major depressive disorder, bipolar disorder, and schizophrenia from the Psychiatric Genomics Consortium. METHODS: The samples comprised 1,622 suicide attempters and 8,786 nonattempters with major depressive disorder; 3,264 attempters and 5,500 nonattempters with bipolar disorder; and 1,683 attempters and 2,946 nonattempters with schizophrenia. A GWAS on suicide attempt was performed by comparing attempters to nonattempters with each disorder, followed by a meta-analysis across disorders. Polygenic risk scoring was used to investigate the genetic relationship between suicide attempt and the psychiatric disorders. RESULTS: Three genome-wide significant loci for suicide attempt were found: one associated with suicide attempt in major depressive disorder, one associated with suicide attempt in bipolar disorder, and one in the meta-analysis of suicide attempt in mood disorders. These associations were not replicated in independent mood disorder cohorts from the UK Biobank and iPSYCH. No significant associations were found in the meta-analysis of all three disorders. Polygenic risk scores for major depression were significantly associated with suicide attempt in major depressive disorder (R2=0.25%), bipolar disorder (R2=0.24%), and schizophrenia (R2=0.40%). CONCLUSIONS: This study provides new information on genetic associations and demonstrates that genetic liability for major depression increases risk for suicide attempt across psychiatric disorders. Further collaborative efforts to increase sample size may help to robustly identify genetic associations and provide biological insights into the etiology of suicide attempt.

Genome-wide Association for Major Depression Through Age at Onset Stratification: Major Depressive Disorder Working Group of the Psychiatric Genomics Consortium.

BACKGROUND: Major depressive disorder (MDD) is a disabling mood disorder, and despite a known heritable component, a large meta-analysis of genome-wide association studies revealed no replicable genetic risk variants. Given prior evidence of heterogeneity by age at onset in MDD, we tested whether genome-wide significant risk variants for MDD could be identified in cases subdivided by age at onset. METHODS: Discovery case-control genome-wide association studies were performed where cases were stratified using increasing/decreasing age-at-onset cutoffs; significant single nucleotide polymorphisms were tested in nine independent replication samples, giving a total sample of 22,158 cases and 133,749 control subjects for subsetting. Polygenic score analysis was used to examine whether differences in shared genetic risk exists between earlier and adult-onset MDD with commonly comorbid disorders of schizophrenia, bipolar disorder, Alzheimer's disease, and coronary artery disease. RESULTS: We identified one replicated genome-wide significant locus associated with adult-onset (>27 years) MDD (rs7647854, odds ratio: 1.16, 95% confidence interval: 1.11-1.21, p = 5.2 × 10-11). Using polygenic score analyses, we show that earlier-onset MDD is genetically more similar to schizophrenia and bipolar disorder than adult-onset MDD. CONCLUSIONS: We demonstrate that using additional phenotype data previously collected by genetic studies to tackle phenotypic heterogeneity in MDD can successfully lead to the discovery of genetic risk factor despite reduced sample size. Furthermore, our results suggest that the genetic susceptibility to MDD differs between adult- and earlier-onset MDD, with earlier-onset cases having a greater genetic overlap with schizophrenia and bipolar disorder.

Genome-Wide Association Study of HIV Whole Genome Sequences Validated using Drug Resistance.

BACKGROUND: Genome-wide association studies (GWAS) have considerably advanced our understanding of human traits and diseases. With the increasing availability of whole genome sequences (WGS) for pathogens, it is important to establish whether GWAS of viral genomes could reveal important biological insights. Here we perform the first proof of concept viral GWAS examining drug resistance (DR), a phenotype with well understood genetics. METHOD: We performed a GWAS of DR in a sample of 343 HIV subtype C patients failing 1st line antiretroviral treatment in rural KwaZulu-Natal, South Africa. The majority and minority variants within each sequence were called using PILON, and GWAS was performed within PLINK. HIV WGS from patients failing on different antiretroviral treatments were compared to sequences derived from individuals naïve to the respective treatment. RESULTS: GWAS methodology was validated by identifying five associations on a genetic level that led to amino acid changes known to cause DR. Further, we highlighted the ability of GWAS to identify epistatic effects, identifying two replicable variants within amino acid 68 of the reverse transcriptase protein previously described as potential fitness compensatory mutations. A possible additional DR variant within amino acid 91 of the matrix region of the Gag protein was associated with tenofovir failure, highlighting GWAS's ability to identify variants outside classical candidate genes. Our results also suggest a polygenic component to DR. CONCLUSIONS: These results validate the applicability of GWAS to HIV WGS data even in relative small samples, and emphasise how high throughput sequencing can provide novel and clinically relevant insights. Further they suggested that for viruses like HIV, population structure was only minor concern compared to that seen in bacteria or parasite GWAS. Given the small genome length and reduced burden for multiple testing, this makes HIV an ideal candidate for GWAS.

Polygenic risk scores for schizophrenia and bipolar disorder predict creativity.

We tested whether polygenic risk scores for schizophrenia and bipolar disorder would predict creativity. Higher scores were associated with artistic society membership or creative profession in both Icelandic (P = 5.2 × 10(-6) and 3.8 × 10(-6) for schizophrenia and bipolar disorder scores, respectively) and replication cohorts (P = 0.0021 and 0.00086). This could not be accounted for by increased relatedness between creative individuals and those with psychoses, indicating that creativity and psychosis share genetic roots.

Genome-wide estimates of inbreeding in unrelated individuals and their association with cognitive ability.

The consequence of reduced cognitive ability from inbreeding has long been investigated, mainly restricted to cousin-cousin marriages. Molecular genetic techniques now allow us to test the relationship between increased ancestral inbreeding and cognitive ability in a population of traditionally unrelated individuals. In a representative UK sample of 2329 individuals, we used genome-wide SNP data to estimate the percentage of the genome covered by runs of homozygous SNPs (ROH). This was tested for association with general cognitive ability, as well as measures of verbal and non-verbal ability. Further, association was tested between these traits and specific ROH. Burden of ROH was not associated with cognitive ability after correction for multiple testing, although burden of ROH was nominally associated with increased non-verbal cognitive ability (P=0.03). Moreover, although no individual ROH was significantly associated with cognitive ability, there was a significant bias towards increased cognitive ability in carriers of ROH (P=0.002). A potential explanation for these results is increased positive assortative mating in spouses with higher cognitive ability, although we found no evidence in support of this hypothesis in a separate sample. Reduced minor allele frequency across the genome was associated with higher cognitive ability, which could contribute to an apparent increase in ROH. This may reflect minor alleles being more likely to be deleterious.

Are genetic risk factors for psychosis also associated with dimension-specific psychotic experiences in adolescence?

Psychosis has been hypothesised to be a continuously distributed quantitative phenotype and disorders such as schizophrenia and bipolar disorder represent its extreme manifestations. Evidence suggests that common genetic variants play an important role in liability to both schizophrenia and bipolar disorder. Here we tested the hypothesis that these common variants would also influence psychotic experiences measured dimensionally in adolescents in the general population. Our aim was to test whether schizophrenia and bipolar disorder polygenic risk scores (PRS), as well as specific single nucleotide polymorphisms (SNPs) previously identified as risk variants for schizophrenia, were associated with adolescent dimension-specific psychotic experiences. Self-reported Paranoia, Hallucinations, Cognitive Disorganisation, Grandiosity, Anhedonia, and Parent-rated Negative Symptoms, as measured by the Specific Psychotic Experiences Questionnaire (SPEQ), were assessed in a community sample of 2,152 16-year-olds. Polygenic risk scores were calculated using estimates of the log of odds ratios from the Psychiatric Genomics Consortium GWAS stage-1 mega-analysis of schizophrenia and bipolar disorder. The polygenic risk analyses yielded no significant associations between schizophrenia and bipolar disorder PRS and the SPEQ measures. The analyses on the 28 individual SNPs previously associated with schizophrenia found that two SNPs in TCF4 returned a significant association with the SPEQ Paranoia dimension, rs17512836 (p-value = 2.57×10⁻4) and rs9960767 (p-value = 6.23×10⁻4). Replication in an independent sample of 16-year-olds (N = 3,427) assessed using the Psychotic-Like Symptoms Questionnaire (PLIKS-Q), a composite measure of multiple positive psychotic experiences, failed to yield significant results. Future research with PRS derived from larger samples, as well as larger adolescent validation samples, would improve the predictive power to test these hypotheses further. The challenges of relating adult clinical diagnostic constructs such as schizophrenia to adolescent psychotic experiences at a genetic level are discussed.

Fecundity of patients with schizophrenia, autism, bipolar disorder, depression, anorexia nervosa, or substance abuse vs their unaffected siblings.

CONTEXT: It is unknown how genetic variants conferring liability to psychiatric disorders survive in the population despite strong negative selection. However, this is key to understanding their etiology and designing studies to identify risk variants. OBJECTIVES: To examine the reproductive fitness of patients with schizophrenia and other psychiatric disorders vs their unaffected siblings and to evaluate the level of selection on causal genetic variants. DESIGN: We measured the fecundity of patients with schizophrenia, autism, bipolar disorder, depression, anorexia nervosa, or substance abuse and their unaffected siblings compared with the general population. SETTING: Population databases in Sweden, including the Multi-Generation Register and the Swedish Hospital Discharge Register. PARTICIPANTS: In total, 2.3 million individuals among the 1950 to 1970 birth cohort in Sweden. MAIN OUTCOME MEASURES: Fertility ratio (FR), reflecting the mean number of children compared with that of the general population, accounting for age, sex, family size, and affected status. RESULTS: Except for women with depression, affected patients had significantly fewer children (FR range for those with psychiatric disorder, 0.23-0.93; P < 10-10). This reduction was consistently greater among men than women, suggesting that male fitness was particularly sensitive. Although sisters of patients with schizophrenia and bipolar disorder had increased fecundity (FR range, 1.02-1.03; P < .01), this was too small on its own to counterbalance the reduced fitness of affected patients. Brothers of patients with schizophrenia and autism showed reduced fecundity (FR range, 0.94-0.97; P < .001). Siblings of patients with depression and substance abuse had significantly increased fecundity (FR range, 1.01-1.05; P < 10-10). In the case of depression, this more than compensated for the lower fecundity of affected individuals. CONCLUSIONS: Our results suggest that strong selection exists against schizophrenia, autism, and anorexia nervosa and that these variants may be maintained by new mutations or an as-yet unknown mechanism. Bipolar disorder did not seem to be under strong negative selection. Vulnerability to depression, and perhaps substance abuse, may be preserved by balancing selection, suggesting the involvement of common genetic variants in ways that depend on other genes and on environment.

A genome-wide association study of a sustained pattern of antidepressant response.

Genome-wide association studies (GWAS) have failed to replicate common genetic variants associated with antidepressant response, as defined using a single endpoint. Genetic influences may be discernible by examining individual variation between sustained versus unsustained patterns of response, which may distinguish medication effects from non-specific, or placebo responses to active medication. We conducted a GWAS among 1116 subjects with Major Depressive Disorder from the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) trial who were characterized using Growth Mixture Modeling as showing a sustained versus unsustained pattern of clinical response over 12 weeks of treatment with citalopram. Replication analyses examined 585 subjects from the Genome-based Therapeutic Drugs for Depression (GENDEP) trial. The strongest association with sustained as opposed to unsustained response in STAR*D involved a single nucleotide polymorphism (SNP; rs10492002) within the acyl-CoA synthetase short-chain family member 3 gene (ACSS3, p-value=4.5×10(-6), odds ratio=0.61). No SNPs met our threshold for genome-wide significance. SNP data were available in GENDEP for 18 of the top 25 SNPs in STAR*D. The most replicable association was with SNP rs7816924 (p=0.008, OR=1.58); no SNP met the replication p-value threshold of 0.003. Joint analysis of these 18 SNPs resulted in the strongest signal coming from rs7816924 (p=2.11×10(-7)), which resides in chondroitin sulfate N-acetylgalactosaminyltransferase 1 gene (CSGALNACT1). An exploratory genetic pathway analysis revealed evidence for an involvement of the KEGG pathway of long-term potentiation (FDR=.02). Results suggest novel genetic associations to sustained response.

The interaction between child maltreatment, adult stressful life events and the 5-HTTLPR in major depression.

Both childhood maltreatment and adult stressful life events are established risk factors for the onset of depression in adulthood. However, the interaction between them can be viewed through two conflicting frameworks. Under a mismatch hypothesis stressful childhoods allow 'adaptive programming' for a stressful adulthood and so can be protective. Only when childhood and adulthood do not match is there a risk of behavioural problems. Alternatively, under the cumulative stress hypothesis we expect increased risk with each additional stressor. It has also been suggested that an individual's genetic background may determine the extent they undergo adaptive programming, and so which of these two hypotheses is relevant. In this study we test for an interaction between exposure to childhood maltreatment and adult stressful life events in a retrospective sample of 455 individuals, using major depression as the outcome. We also test whether this interaction differs by genotype at the 5-HTTLPR, a candidate for an individual's plasticity to adaptive programming. Early maltreatment and stressful life events in adulthood interacted to produce increased risk for depression over each individually (p = 0.055). This supports the cumulative stress hypothesis over the mismatch hypothesis, at least with respect to severe environmental risk factors. This effect was not altered by 5-HTTLPR allele, suggesting there was no difference by genotype in adaptive programming to these events. We suggest that the apparent additional vulnerability to stressful events of those who have experienced maltreatment has clinical relevance, highlighting the importance of providing support beyond the immediate aftermath of maltreatment into adulthood.

Non-random dropout and the relative efficacy of escitalopram and nortriptyline in treating major depressive disorder.

Most comparisons of the efficacy of antidepressants have relied on the assumption that missing data are randomly distributed. Dropout rates differ between drugs, suggesting this assumption may not hold true. This paper examines the effect of non-random dropout on a comparison of two antidepressant drugs, escitalopram and nortriptyline, in the treatment of major depressive disorder. The GENDEP study followed adult patients with major depressive disorder over 12 weeks of treatment, and the primary analysis found no difference in efficacy of the two antidepressants under missing at random assumption. By applying the recently developed Muthén-Roy model, we compared the relative efficacy of these two antidepressants taking into account non-random distribution of missing outcomes (NMAR). Individuals who dropped out of the study were those who were not responding to treatment. Based on the best fitting NMAR model, it was found that escitalopram reduced symptom scores by an additional 1.4 points on the Montgomery-Åsberg Depression Rating Scale (p = 0.02), equivalent to 5% of baseline depression severity, compared to nortriptyline. We conclude that association between dropout and worsening symptoms led to an overestimate of the effectiveness of treatment, especially with nortriptyline, in the primary analysis. These findings review the primary analysis of GENDEP and suggest that, when non-random dropout is accounted for, escitalopram is more effective than nortriptyline in reducing symptoms of major depression.