This study was a post hoc analysis, utilizing a prespecified analysis plan, of the PRIME Care study. Briefly, the PRIME Care trial was a randomized pragmatic clinical trial in which patients with MDD and at least one treatment failure were randomized to receive either PGx-guided treatment or usual care for a period of 24 weeks at the beginning of an antidepressant treatment episode. Baseline assessments were performed prior to randomization, with post-randomization follow-up visits scheduled at 4, 8, 12, 18, and 24 weeks. A full description of the trial, including eligibility and exclusion criteria, is available (Oslin et al., 2021; Oslin et al., 2022). This post hoc study was reviewed by the Advarra Institutional Review Board and determined to qualify as exempt research per 45 CFR 46.104(d)(4).

Relevant clinical and demographic information collected at the baseline or follow-up visits included age, race (Black/African American, White, Other/unspecified), Hispanic ethnicity, sex (female, male), smoking status (cigarettes per day over the prior 30 days), patient-reported psychotropic treatment history, reported visit date (days since randomization), and practice location (primary care, mental healthcare, or integrated care) (Table 1).

Assessments relevant for this analysis, conducted at baseline and each follow-up visit, included the Patient Health Questionnaire-9 (PHQ-9) (Spitzer et al., 1999) and the Post Traumatic Stress Disorder (PTSD) Checklist for DSM-5 (PCL-5) (Weathers et al., 2013).

The primary endpoints of this analysis were the time of the first instance of remission (defined as PHQ-9 ≤5) and response (defined as ≥50% reduction from baseline PHQ-9 score). PHQ-9 scores range from 0–27 points.

Baseline characteristics were compared between study arms using chi-squared tests for categorical variables and Wilcoxon rank-sum test for baseline PHQ-9 score to assess significant differences, with a two-sided α level of 5%.

In the primary analysis, two separate multivariate Cox proportional hazards models (one for remission, one for response) were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for associations between study arm (PGx-guided or usual care) and the first instance of either PHQ-9 remission or PHQ-9 response, respectively, measured at each follow-up visit (Klein and Moeschberger, 2003). Baseline PHQ-9 score was included as the only additional covariate in each model as an explanatory variable. The Efron method was selected to handle tied event times. The statistical significance of individual explanatory variables was determined using likelihood ratio tests. The proportional hazard assumption for the study arm variable was assessed using the Schoenfeld residuals test. Two-sided tests of significance were used, with a Bonferroni-adjusted α level of 0.025 (0.05/2) to account for testing two outcomes. Cox models used study week (4, 8, 12, 18, or 24 weeks) as the time scale in accordance with the timing of data collection in the PRIME Care study design (set prior to this post hoc analysis). Patients were right-censored at the first instance of a missing follow-up visit (defined as any visit without an associated PHQ-9 score). Patients were also right-censored at the first instance of a follow-up visit occurring outside of the expected time window based on reported visit date (14–42 days [week 4], 42–70 days [week 8], 70–105 days [week 12], 105–147 days [week 18], 147–189 days [week 24]). These windows were defined post hoc, based on the midpoints between the weeks when follow-up visits were scheduled to occur. Patients were excluded from the analysis if they only had baseline data remaining after right-censoring (Supplementary Figure S1). As a sensitivity analysis, models were re-analyzed without the right-censoring based on reported visit date. As an additional sensitivity analysis, models were re-analyzed with the inclusion of additional prespecified covariates (age [60+], race, Hispanic ethnicity, sex, presence of PTSD at baseline, history of treatment refractory depression, smoking status [≥1 cigarette per day over the prior 30 days], and practice location).

All analyses were conducted using R software version 4.4.1 (R Foundation for Statistical Computing, 2021).

Out of the 1,944 patients in the PRIME Care study, 180 (9.3%) were excluded due to having only baseline data remaining after right censoring, leaving 1,764 patients available for analysis (90.7%; n = 884 in the PGx-guided arm and n = 880 in the usual care arm), with the majority being under 60 years of age (73% in the PGx-guided arm, 73% in the usual care arm) and male (77% in the PGx-guided arm, 73% in the usual care arm) (Table 1; Supplementary Figure S1). This cohort was also mostly White (67% in the PGx-guided arm, 70% in the usual care arm), non-Hispanic (88% in the PGx-guided arm, 89% in the usual care arm), and non-smoking (82% in the PGx-guided arm, 81% in the usual care arm). Both arms had equal proportions of individuals with moderately severe depression (mean PHQ-9 values of 17.5 and 17.4, respectively), treatment refractory depression (31%), and comorbid PTSD (58%).

Baseline patient characteristics were compared by study arm (Table 1). No statistically significant differences were observed across study arms for any baseline characteristic.

Cox proportional hazards models and likelihood ratio tests revealed that patients who received PGx-guided treatment had higher rates of remission (HR [95% CI] = 1.27 [1.05, 1.53]; p = 0.015) and response (HR [95% CI] = 1.21 [1.05, 1.40]; p = 0.010) at any time relative to patients receiving usual care. Kaplan-Meier plots, using reported visit date as the timescale, show that the cumulative incidence of remission and response appeared higher in the PGx-guided arm starting early during the study period and continued through the end of the study period (Figure 1).

The Schoenfeld residuals test of the study arm variable was not statistically significant for remission (p = 0.931) or response (p = 0.112), indicating no evidence of time-dependent effects of study arm, consistent with the proportional-hazards assumption. Schoenfeld residual plots did not show any significant fluctuation of the remission or response hazard ratios over time (Figure 2). These results indicate that the benefit due to PGx-guided treatment did not change significantly over the 24-week period.

As a sensitivity analysis, models were re-analyzed without right-censoring based on reported visit date, resulting in fewer (n = 125 [6.4%]) patients excluded. Among patients included in the sensitivity analysis (n = 1819 [93.6%]), the associations between study arm and time-to-remission (HR [95% CI] = 1.24 [1.03, 1.49]; p = 0.022) or response (HR [95% CI] = 1.20 [1.05, 1.38]; p = 0.010), as well as Schoenfeld residuals test results for study arm and remission (p = 0.699) or response (p = 0.227), were similar to those seen in the primary analysis. Models were also re-analyzed with the inclusion of additional prespecified covariates (age [60+], race, Hispanic ethnicity, sex, presence of PTSD at baseline, history of treatment refractory depression, smoking status [≥1 cigarette per day over the prior 30 days], and practice location). After adjusting for these prespecified covariates, associations between study arm and time-to-remission (HR [95% CI] = 1.30 [1.07, 1.57]; p = 0.008) or response (HR [95% CI] = 1.23 [1.07, 1.43]; p = 0.004), as well as Schoenfeld residuals test results for study arm and remission (p = 0.990) or response (p = 0.092), were similar to those seen in the primary analysis. Primary analysis findings were robust to changes in right-censoring and to the inclusion of additional covariates.

The PRIME Care study, on which this analysis was based, was conducted on a large cohort of depressed patients and the design was pragmatic to represent real-world clinical practice (Oslin et al., 2022). The PRIME Care study also followed a prespecified analysis plan, and included multiple time points to assess depression outcomes between the PGx-guided and usual care arms over 24 weeks after randomization (Oslin et al., 2022). Building upon the original study, which assessed outcome differences at each time point individually, the current analysis was able to directly assess timing of initial response and remission across the entire study duration. Moreover, this analysis expands upon the non-significant interaction between arm and time reported in Oslin et al. (2022) by using statistical methods that directly assess time-to-event, providing a more statistically principled and clinically interpretable framework for evaluating the persistence of study arm effects.

While the PRIME Care study was the largest randomized clinical trial evaluating depression outcomes with and without PGx testing, the study design did not include time points after 24 weeks. As such, this analysis could not evaluate the persistence of the impact of PGx-guided treatment for longer than 24 weeks. Moreover, only time-to-first response or remission was evaluated, and any response/remission events occurring after relapse were not included in the analysis. Additionally, the outcomes observed in this study were limited to a specific PGx test and may not apply to all PGx tests. Finally, this study was conducted in a cohort of veterans, which may not be representative of the MDD patient population in the United States or worldwide; for example, the frequencies of PTSD and male sex were higher than in other large US depression cohorts.