Dissecting the Heterogeneity of Major Depression

Fri, 02 Aug 2019 00:00:00 +0000

Bradley Jermy and Jonathan Coleman

The World Health Organisation reported in 2018 that approximately 300 million people suffer with major depression worldwide. Despite the impact of the disorder, treatments are not effective enough: only around 30% of participants remitted following treatment with citalopram, commonly regarded as the initial anti-depressant therapy. While research into areas such as genetics and inflammation have recently made great strides, it has been difficult to identify specific biomarkers for major depression. This may be due to many different symptom profiles resulting in the same diagnosis of major depression.

Major depression is diagnosed according to a pool of nine symptoms. Having any five over a two-week period (including at least one of ‘depressed mood’ or ‘anhedonia/loss of interest’) is necessary for a diagnosis. Theoretically, this gives rise to 227 possible combinations, and studies suggest most of these combinations are found in people with depression.

Caption: The nine symptoms of a major depressive episode used in the DSM diagnostic manual.

To consider how research can tackle this issue, the NIHR Maudsley Biomedical Research Centre Biomarkers and Genomics theme (lead: Professor Cathryn Lewis) hosted a symposium entitled ‘Dissecting the Heterogeneity of Depression’ on 24th April 2019. Each of the six speakers discussed how analyses into symptoms, genetics, comorbidities and outcomes can begin to decompose the disorder into more manageable, bitesize chunks.

Talk Summaries

Eiko Fried, Assistant Professor, Leiden University

Title: Depression is a problematic phenotype: The potential benefits to studying symptoms over syndromes.

Summary:

The Problem

Eiko focused on four overarching issues that arise from using major depression as a phenotype: ‘Heterogeneity’, ‘Measurement, ‘Reliability’, and ‘Validity’.

Within the diagnostic criteria for major depression in the DSM, the word ‘or’ is used nine times to aggregate symptoms. Some of these are polar opposites of one another, for example, weight gain or loss. Disaggregating these symptoms increases the number of possible symptom combinations from 227 to 10,377, and also allows for two individuals with a diagnosis of major depression to not share a single symptom. In spite of this massive heterogeneity, we currently add together symptoms, treating them as interchangeable with one another.

This heterogeneity is compounded by the fact that 280 scales have been used to measure major depression, some of which do not correlate with each other. A recent paper by Eiko, which reviewed a series of the seven most commonly used scales, shows 40% of symptoms appear just once and only 12% appear in all scales. Again, there is evidence of many studies combining results using different scales, implicitly assuming such scales are interchangeable, but the evidence doesn’t support that. Many of these scales do not consistently measure the same construct over time. Indeed, it is questionable whether they measure a single construct of major depression at all, as scales often require multiple latent constructs to explain the relationship between items.

It was hoped the introduction of diagnostic criteria would improve reliability through the use of a common language for clinicians, which would then lead to improvements in validity. The inter-rater reliability for major depression, i.e. the agreement of a diagnosis between two clinicians for a given patient, recently obtained as part of the DSM-5 field trials was poor (kappa = 0.28), highlighting the considerable improvements required to our current definition.

Given reliability has shown to be lacking, it is no surprise that the validity of the disorder is in a somewhat similar state. Major depression does not hold up to generally accepted criteria for validity, for example, there is no clear aetiology or diagnostic boundaries, and treatments lack specificity.

The Solution is in the Symptoms

Eiko suggests that, while major depression has some practical value as a clinical grouping, research is better served by examining individual symptoms and their causal relations, a field he coined as ‘Symptomics’ (albeit begrudgingly). He points towards studies supporting the notion that individual symptoms arise from different risk factors which in turn give rise to differing levels of impairment. If this is true, Eiko argues there is a strong case against a single common cause being the sole explanation for this complex disorder.

Eiko then delves into the world of biology, showing c-reactive protein, a promising biomarker for major depression, has a greater relationship with somatic symptoms compared with cognitive or mood related symptoms. These and other studies highlight the benefits of using a more granular, symptom level analysis. Symptom level analysis might create more homogenous comparisons which may improve reliability, validity and replicability within depression research.

Should you wish to explore Eiko’s arguments in more detail:

Slides and audio for this talk are available at: https://osf.io/qdkpw/
A twitter thread providing additional literature can be found here.
Joni Coleman’s live-tweet to this talk can be found here.

Bradley Jermy, BRC PhD Student, Kings College London

Title: Depression: Dimensional or Categorical? A genetic comparison.

Summary:

Bradley’s talk focused on how we conceptualise major depression. He argues that, despite a growing recognition of the continuous or ‘dimensional’ nature of depression, our categorical classification system implies a separation of people with major depression from those without.

Genetic research can be used to test which conceptualisation is most appropriate. Bradley stated that should a clear separation exist between major depression cases and controls, a genetic difference would be evident when comparing depression analysed as a continuum of symptoms with depression analysed categorically (as cases and controls). He shows this not to be the case through exploring genetic correlations between two of the largest genome-wide association studies (GWAS) conducted on major depression as a category and on a continuum.

He builds on the evidence for dimensionality, suggesting that a single dimension of symptom severity is too simplistic. Instead, he proposes establishing a multidimensional substructure of major depression through data-driven methods such as factor analysis that investigate covariation of symptoms.

To support his claim of multidimensionality, he summarised a study that investigated 7,500 twins to show three uncorrelated genetic factors are required to explain the covariation across the nine DSM symptoms. Through exploring the genetic variants underlying these factors, Bradley believes a greater level of specificity can be achieved relative to the current construct of major depression, which can be used to enhance prediction of course, outcome and treatment response.

Should you wish to discuss these ideas further Bradley is contactable through:

e-mail: bradley.jermy@kcl.ac.uk
twitter: @brad_jermy
Joni’s live tweet of the talk can be found here.

Anamaria Brailean, Postdoctoral Researcher, King’s College London

Title: Characteristics, comorbidities and correlates of atypical depression: Evidence from the UK Biobank Mental Health Survey.

Summary:

A key goal in identifying reliable subtypes of major depression is to improve prediction of symptom presentation, course of illness and treatment response. Possibly one of the most established distinctions for major depression is the separation of typical and atypical depression.

Atypical depression is characterised clinically through mood reactivity and any two of: increased appetite or weight gain, hypersomnia, heaviness in the arms or legs, and interpersonal rejection sensitivity. This is complex to identify clinically. However, Anamaria argues that the key separator from typical depression is the ‘reversed neurovegetative symptoms’ (hypersomnia and increased appetite or weight gain). As this is the best predictor of the subtype, it is possible to identify cases with atypical depression in community samples thus increasing sample sizes from self-report data.

Previous literature has suggested atypical depression differs in terms of chronicity (younger age of onset as well as severity and recurrence of episodes), risk factors (lower socioeconomic status) and psychiatric and physical comorbidities such as obesity and addictions. In her talk, Anamaria considered the current literature surrounding the subtype, highlighting the limitations of small sample size, lack of consistency regarding comparison groups and a focus on cross-sectional rather than lifetime data. Anamaria goes on to discuss how she has recently sought to tackle these problems and thus replicate previous literature through analysing the large population cohort, the UK Biobank.

Should you wish to discuss these ideas further Anamaria is contactable through anamaria.1.brailean@kcl.ac.uk

Anamaria’s paper has since been published.
Joni’s live tweet of the talk can be found here.

David Howard, Sir Henry Wellcome Fellow, King’s College London

Title: “What do the 102 genetic variants identified for depression reveal about its heterogeneity?”

Summary:

David began the session describing his recently published large-scale GWAS analysis, combining data from the Psychiatric Genomics Consortium, UK Biobank and 23andMe. David’s work combined cohorts where cases were defined using detailed clinical assessments with other, larger cohorts where participants self-reported their illness. However, there were strong genetic correlations between the different cohorts, suggesting that these different case definitions had a similar genetic component (although see Na Cai’s talk below for further discussion of this).

David’s analyses identified over 100 regions of the genome associated with depression, and most of these also showed evidence of association in a further independent cohort from 23andMe. Further analyses show evidence for the involvement of specific genes, and implicate certain regions of the brain. Identifying genes associated with depression can provide new targets for anti-depressant drugs, and further analyses demonstrate that specific drugs and drug classes can be linked to depression in this way.

Following on from this research, David then described his ongoing efforts to dissect the heterogeneity of genetic influences on depression, as part of his fellowship research. Initial results seem promising for identifying specific subgroups of individuals with depression, and David hopes to extend this research into larger samples.

Should you wish to discuss these ideas further David is contactable through david.howard@kcl.ac.uk

David’s paper is available here.
Joni’s live tweet of the talk can be found here.

Saskia Hagenaars, MRC Skills Development Fellow, King’s College London

Title: “Genetic comorbidity between depression and cardio-metabolic disease: stratified by depression age at onset”

Summary:

There is a known co-occurrence of depression and cardiometabolic disease, particularly in older individuals. However, the correlation of genetic associations with depression and with cardiometabolic diseases is less well-described – while there is good evidence for small positive genetic correlations between depression and cardiac illness and body mass index, no such evidence has been found for other disorders such as stroke or type 2 diabetes.

Saskia sought to shed further light on the relationship between depression and cardiometabolic illness, and in particular to assess whether this relationship differs for depression beginning later, rather than earlier, in life. Using polygenic risk scores, she showed that depression is associated with a variety of cardiometabolic traits, including stroke and type 2 diabetes. However, no clear effects by age-of-onset are apparent, although some very weak evidence for such an effect on the relationship between depression and BMI has support from previous epidemiological research. It remains to be seen whether the disentangling the heterogeneity of depression might shed further light on this relationship.

Should you wish to discuss these ideas further, Saskia is contactable through saskia.hagenaars@kcl.ac.uk

Joni’s live tweet of the talk can be found here.

Na Cai, EBI-Sanger Postdoctoral Fellow, Wellcome Sanger Institute

Title: “Definitions of depression used in GWAS”

Summary:

To close the symposium, Na, the second keynote speaker, addressed strategies for phenotyping used in genetic studies of depression, and their impact on genetic discoveries.

When the first large-scale meta-analysis of GWAS of major depression did not implicate specific regions of the genome with depression, researchers estimated the genetic architecture of major depression to be like that of weight, where many genetic and environmental factors each contribute a small effect. The combination of these factors in each individual is highly heterogeneous. They estimated that at this level of heterogeneity, 50000 cases and matched controls were needed for a GWAS discovery, more than all existing cohorts combined. Researchers therefore faced the decision of investing in larger cohorts, or in those with lower heterogeneity.

Na first described one attempt to reduce heterogeneity in data collection for a GWAS on major depression: the CONVERGE study. This study, in Han Chinese women with severe recurrent major depression, was the first to identify genomic regions associated with the disease (identifying two regions). The success of CONVERGE was attributed to its detailed, specific approach to defining major depression. CONVERGE has continued to provide valuable insights in depression genetics, including details of the relationship of depression with childhood trauma, as well as novel ways to study complex genetic traits and diseases.

She followed by describing successful examples of the “large cohorts” approach – using the low-cost approach of collecting self-reported diagnosis of depression by European customers, 23andME performed a GWAS on more than 75,000 cases and 230,000 controls, and identified 15 genomic regions. Efforts from the Psychiatric Genomics Consortium, as well as David’s work on combining definitions of depression in the UK Biobank with that in other cohorts (described earlier in the symposium), brought the total count of genomic regions associated with depression up to 112.

Both approaches are successful in finding GWAS hits – but are they accessing the same thing? Na argued that the answer is “not necessarily”, and demonstrates this through her ongoing work in the UK Biobank comparing different definitions of depression. She found that as the stringency of defining depression decreases, such as when self-report is used rather than the full DSM diagnostic criteria, the apparent genetic contribution to the condition decreases too. This dilution of genetic effects may be due to a variety of other factors, including misdiagnoses and misclassification of other psychiatric conditions as major depression, Furthermore, less strict definitions implicate regions of the genome with more general psychiatric association (rather than being specific to depression).

Na concluded that both approaches are useful in discovering genetic loci relevant to major depression, but if we relax the strictness of our definitions of the disease in data collection, we must be careful to ensure what we are studying still is depression, or communicate our findings to reflect their non-specificity.

Should you wish to discuss these ideas further, Na is contactable through caina@ebi.ac.uk

Na preprint can be found here.
Joni’s live tweet of the talk can be found here.

Symptoms on Bradley S. Jermy

Dissecting the Heterogeneity of Major Depression

Talk Summaries

Eiko Fried, Assistant Professor, Leiden University

Title: Depression is a problematic phenotype: The potential benefits to studying symptoms over syndromes.

Summary:

The Problem

The Solution is in the Symptoms

Should you wish to explore Eiko’s arguments in more detail:

Bradley Jermy, BRC PhD Student, Kings College London

Title: Depression: Dimensional or Categorical? A genetic comparison.

Summary:

Should you wish to discuss these ideas further Bradley is contactable through:

Anamaria Brailean, Postdoctoral Researcher, King’s College London

Title: Characteristics, comorbidities and correlates of atypical depression: Evidence from the UK Biobank Mental Health Survey.

Summary:

Should you wish to discuss these ideas further Anamaria is contactable through anamaria.1.brailean@kcl.ac.uk

David Howard, Sir Henry Wellcome Fellow, King’s College London

Title: “What do the 102 genetic variants identified for depression reveal about its heterogeneity?”

Summary:

Should you wish to discuss these ideas further David is contactable through david.howard@kcl.ac.uk

Saskia Hagenaars, MRC Skills Development Fellow, King’s College London

Title: “Genetic comorbidity between depression and cardio-metabolic disease: stratified by depression age at onset”

Summary:

Should you wish to discuss these ideas further, Saskia is contactable through saskia.hagenaars@kcl.ac.uk

Na Cai, EBI-Sanger Postdoctoral Fellow, Wellcome Sanger Institute

Title: “Definitions of depression used in GWAS”

Summary:

Should you wish to discuss these ideas further, Na is contactable through caina@ebi.ac.uk