‘Treatment-resistant depression’ (TRD) is defined as major depressive disorder (MDD) for which two different classes of antidepressants have not been effective (Souery et al., 1999). Patients with TRD are reported to have 40% higher odds of hospitalisation and 30-50% higher mortality than patients for whom current first- and second-line antidepressants are effective (Chan et al., 2022). In the UK, around 1 in 6 adults have been estimated to have depression (Office for National Statistics, 2022), of which 10-30% would be estimated to have TRD.
However, the causes of TRD as a distinct subgroup of MDD remain unknown. Whilst TRD has been shown to be associated with other psychiatric conditions such as anxiety disorders, stress disorders, substance use disorder, and attention deficit hyperactivity disorder (ADHD), these studies have used survey response and other observational data, meaning there would be limited evidence to suggest causality. By utilising genetic data, polygenic scores (PGS) can be derived to determine the effect genetic variants may have on somebody’s likelihood of developing a specific disease. You can read more about PGS in previous Mental Elf Blogs (e.g., Staines, 2025; Hagenberg, 2024).
This study from Xu et al. uses data from the “All of Us Research Program”, a US nationwide cohort linked to electronic health records (EHRs), where patients’ medical history including demographics and medication history is updated by the healthcare provider. The data also includes whole genome sequencing (WGS), a process which identifies the entire DNA order of a genome. From this data, the researchers made PGS analyses to help determine whether TRD has separate biological mechanisms from MDD which responds to treatment.
Genetics may help reveal why some forms of depression don’t respond well to currently available treatments.
Methods
The study used data from over 400,000 participants, separating those with genetic similarities to people of European descent (N = 124 945) from those of diverse ancestry (N = 104 388) and those without WGS but with genotype data (N = 63 330) due to differences in allele frequencies and allele combinations (linkage disequilibrium) across populations. Identification of MDD and TRD status were determined through diagnostic codes and prescription data from the linked electronic health records, and alternative definitions were used in sensitivity analyses. These definitions included requiring clinical visits, allowing for schizophrenia and/or bipolar diagnoses, using different treatment lengths, allowing for different time windows, and allowing for special treatments such as electroconvulsive therapy or antipsychotics.
The authors used logistic regression to test whether PGS were associated with treatment-resistant depression, incorporating principal components (uncorrelated new variables created as a linear combination of variables from the original dataset as dimension reduction technique), biological age, and sex as covariates. In total, 61 traits from 7 categories were selected for the PGS analysis. Traits spanned topics such as intelligence/cognition, inflammation, personality, psychiatric disorders, sleep pattern, substance use, and temperament.
Results
After removing those with missing data, 292,663 participants remained (60% were female; median age of 57). Of the 61 PGS traits, 42 showed significant links with treatment-responsive MDD compared to those without MDD. When comparing those with TRD and treatment-responsive MDD, 11 PGS stood out, including domains of education/cognition, personality, sleep, and temperament.
Notably, genetic liability to insomnia and neuroticism both increased one’s risk of TRD (both with an odds ratio of 1.11), whilst higher education (odds ratio: 0.88) and intelligence (odds ratio: 0.91) were shown to be protective against TRD development. These findings remained consistent across different TRD definitions and ancestries, with no significant sex differences.
The study also looked at how MDD progressed to TRD. Among 28,964 participants with detailed genetic and diagnostic data available, 3,854 (~13.3%) progressed to TRD in an average time of 2.6 years. Those with higher polygenic scores for education (PGS-EDU) were shown to have slower progression rates than those with lower PGS-EDU. Similarly, individuals with higher genetic risk for insomnia progressed to TRD faster than those with a low PGS, with an estimated 21% increased risk of progression. These results suggest that a genetic liability to these traits may affect the likelihood of responding to currently available antidepressant treatments for MDD.
Finally, the study looked at different definitions of treatment-responsive MDD and TRD to see if these impacted the results. With three definitions of treatment-responsive MDD and five definitions of TRD (giving 15 possible combinations), results were highly consistent with the main findings, and 8 of the 11 PGS traits found in the main study were present in every combination of definitions. These were neuroticism and its subscales, insomnia, recent affects, educational attainment, and cognitive performance.
Genetic risk for insomnia and neuroticism may increase the likelihood and speed of developing treatment-resistant depression, while higher genetic scores for education may offer some protection.
Conclusions
The research found that genetic liability to neuroticism is associated with an increased risk of TRD, while traits related to education and cognition were shown to be protective of TRD. Insomnia also emerged as a genetic risk factor for TRD, a condition for which treatment options have advanced in recent years (Rosenberg, R et al. 2021). The authors concluded:
These insights not only advance our understanding of the genetic underpinnings of TRD but also highlight potential ways to improve outcomes for those experiencing this challenging condition.
These new genetic insights into treatment-resistant depression reveal key risk and protective factors that could shape future treatment approaches.
Strengths and limitations
The paper used a large cohort of 292,663 participants and had three separate groups – those of White European ancestry with whole-genome sequencing (WGS), a diverse ancestry group with WGS, and people without WGS but genotyped through a microarray dataset. The results being highly consistent across these three groups gives a stronger indication that the traits showing associations with TRD may be causal rather than being driven by other factors or entirely due to chance. These results were also consistent when looking at different definitions of treatment resistance. Another strength of the research was the use of different methods to evaluate associations; as well as the PGS analysis, the paper used survival analysis to look at associations of time-to-onset for TRD with the selected PGS. Using methods with different limitations and assumptions can provide greater confidence that there are differences between treatment-resistant depression and depression which responds to treatment.
While the results are consistent across different definitions, TRD was defined using medication treatments and changes from electronic health records where the reasons for change of drug were not given. There may, therefore, be people classified as having TRD who switched medication due to other reasons, e.g. unwanted side effects such as weight gain or sexual dysfunction (Keks et al. 2016). The EHR would also not allow for data on other treatment methods shown to be effective for TRD such as cognitive behavioural therapy (CBT) (Wiles et al. 2016). As mentioned in the limitations section of the paper, there may be residual confounding for factors not available in the dataset, such as the barrier of financial cost of treatment in the US and adverse childhood experiences (Farooq et al. 2024). This means that factors not used in the analysis could potentially explain the elevated risk shown in the results.
This large, diverse genetic study found consistent results across groups, strengthening the case that certain traits may causally influence treatment-resistant depression.
Implications for practice
This study’s results suggest that neuroticism, insomnia, and low educational attainment may be predictive risk factors for an increased likelihood of developing treatment-resistant depression. Early treatment of neuroticism and insomnia may have the potential to lower a patient’s risk of having TRD later in life. We could perhaps envisage a future where clinicians could use knowledge of these risk factors to tailor medication selection or non-pharmaceutical treatment options such as CBT. However, we must acknowledge that the raised risks for developing TRD based on these traits are relatively modest and much more work is needed before this future becomes a reality.
The PGS for major depressive disorder was not significantly associated with TRD. This may suggest that treatment-resistance is more complicated than just being a subgroup, or more severe form, of MDD and may require additional screening information to determine the likelihood of treatment responsiveness.
While genetic tests are not generally provided in healthcare settings (PHG Foundation, 2021), there have been developments that could lead this to change in the near future (Polygenic Risk Score Task Force of the International Common Disease Alliance, 2021), meaning it may soon be possible to calculate polygenic scores for patients. For the time being, however, predictive models based on known risk factors will be more clinically useful.
A potential avenue for future research could be to see whether early interventions on the associated traits could reduce a person’s likelihood of developing TRD. For example, could CBT for insomnia mitigate future TRD diagnosis? It would also be useful to link electronic health records to survey or Census data to account for other environmental factors such as socio-economic status or stress.
Despite the consistency of the results being shown across diverse ancestry groups, it would be useful for a similar analysis to be done using data from countries where treatment-seeking behaviours, barriers, and adherence may differ. The cost barriers to healthcare in the US may limit the participants based on socio-economic position, which could bias the results.
Traits like neuroticism, insomnia, and low education may predict treatment-resistant depression, offering potential for early intervention, though clinical application is still developing.
Statement of interests
I had no involvement in the study and have no conflicts of interest to declare.
Links
Primary paper
Xu B, Forthman KL, Kuplicki R, et al. Genetic Correlates of Treatment-Resistant Depression. JAMA Psychiatry. 2025;82(5):505–513. doi:10.1001/jamapsychiatry.2024.4825
Other references
D Souery, J Amsterdam, C de Montigny, et al., Treatment resistant depression: methodological overview and operational criteria, European Neuropsychopharmacology, Volume 9, Issues 1–2, 1999, Pages 83-91,
Chan VK, Cheung EC, Chan SS, et al., Mortality-causing mechanisms and healthcare resource utilisation of treatment-resistant depression: A six-year population-based cohort study. Lancet Reg Health West Pac. 2022 Mar 15;22:100426. doi: 10.1016/j.lanwpc.2022.100426.
Cost of living and depression in adults, Great Britain: 29 September to 23 October 2022, Office for National Statistics, 6 December 2022. Accessed 05/09/2025
Staines L, Naturalistic use of psychedelics in the general population: what are the mental health risks?, National Elf Service, Jan 23 2025, Accessed Sep 05 2025.
Hagenberg J, Can proteomics improve our prediction of depression remission?, National Elf Service, Feb 15 2024, Accessed Sep 05 2025.
Rosenberg R, Citrome L, Drake CL. Advances in the Treatment of Chronic Insomnia: A Narrative Review of New Nonpharmacologic and Pharmacologic Therapies. Neuropsychiatr Dis Treat. 2021;17:2549-2566.
doi:10.2147/NDT.S297504
Keks N, Hope J, Keogh S. Switching and stopping antidepressants. Aust Prescr. 2016;39(3):76-83. doi:10.18773/austprescr.2016.039
Wiles NJ, Thomas L, Turner N, et al. Long-term effectiveness and cost-effectiveness of cognitive behavioural therapy as an adjunct to pharmacotherapy for treatment-resistant depression in primary care: follow-up of the CoBalT randomised controlled trial. Lancet Psychiatry. 2016;3(2):137-144. doi:10.1016/S2215-0366(15)00495-2
Farooq, B., Russell, A.E., Howe, L.D. et al., The relationship between type, timing and duration of exposure to adverse childhood experiences and adolescent self-harm and depression: findings from three UK prospective population-based cohorts. J Child Psychol Psychiatr, 65: 1369-1387, 2024.
Moorthie S, Hall A, Janus J, Polygenic scores and clinical utility, PHG Foundation, January 2021, Accessed 05 Sep 2025.
Polygenic Risk Score Task Force of the International Common Disease Alliance. Responsible use of polygenic risk scores in the clinic: potential benefits, risks and gaps. Nat Med 27, 1876–1884 (2021). https://doi.org/10.1038/s41591-021-01549-6
