Association between a selective 5-HT4 receptor agonist and incidence of major depressive disorder: an emulated target trial

Background The serotonin 4 receptor (5-HT4R) is a promising target for the treatment of depression. Highly-selective 5-HT4R agonists, such as prucalopride, have antidepressant-like and pro-cognitive effects in pre-clinical models, but their clinical effects are not yet established. Aims To determine whether prucalopride (a 5-HT4R agonist and licensed treatment for constipation) is associated with reduced incidence of depression in individuals with no past history of mental illness, compared to anti-constipation agents with no effect on the central nervous system. Method Using anonymised routinely collected data from a large-scale US electronic health records network, we conducted an emulated target trial comparing depression incidence over one year in individuals without prior diagnoses of major mental illness who initiated treatment with prucalopride versus two alternative anti-constipation agents which act by different mechanisms (linaclotide and lubiprostone). Cohorts were matched for 121 covariates capturing sociodemographic factors, and historical and/or concurrent comorbidities and medications. The primary outcome was a first diagnosis of major depressive disorder (ICD-10 code F32) within one year of the index date. Robustness of the results to changes in model and population specification was tested. Secondary outcomes included a first diagnosis of six other neuropsychiatric disorders. Results Treatment with prucalopride was associated with significantly lower incidence of depression in the following year compared to linaclotide (HR 0.87, 95% CI 0.76-0.99, p=0.038, n=8572 in each matched cohort) and lubiprostone (HR 0.79, 95% CI 0.69-0.91, p<0.001, n=8281). Significantly lower risks of all mood disorders and psychosis were also observed. Results were similar across robustness analyses. Conclusions These findings support pre-clinical data and suggest a role for 5-HT4R agonists as novel agents in the prevention of major depression. These findings should stimulate randomised controlled trials to confirm if these agents can serve as a novel class of antidepressant within a clinical setting.

The cohorts included adults of any age or sex.Patients were censored at their last visit or when they died.Patients with pre-existing diagnoses of common and / or serious mental illness within their health records before the index date were excluded; that is with either code F20-F29 (schizophrenia, schizotypal, delusional, and other non-mood psychotic disorders), F30-F39 (mood [affective] disorders), F40-F48 (anxiety, dissociative, stressrelated, somatoform and other nonpsychotic mental disorders), or F01-F09 (mental disorders due to known physiological conditions including cognitive disorders).
All three medications (prucalopride, linaclotide, and lubiprostone) are approved by the FDA for use in the United States for chronic constipation and all three drugs are available under Medicaid / Medicare.5-HT 4 agonists, linaclotide and lubiprostone are third-line pharmacotherapy options in United States and other IBS-C guidelines (5,6).
It is important to note that prucalopride, linaclotide and lubiprostone are all indicated for chronic constipation, and prucalopride is mentioned in UK irritable bowel syndrome (IBS) guidelines (6), but only linaclotide and lubiprostone have IBS as a specific indication.Nonetheless, matching for IBS achieved across analyses in this study guarantees that the prevalence of IBS was similar between cohorts.Ideally, we would also have selected people with the indication for these drugs (i.e.chronic constipation) coded (7)(8)(9).

Methods 3 -Covariates
All cohorts were matched for disorders and medications that could be associated with differences in choices of anti-constipation treatment and / or with psychiatric disorders according to expert opinion.In addition, matching occurred for any diagnoses or medication where the baseline incidence was greater than 5% in both cohorts and the difference between cohorts greater than 2% at the pre-matching stage.Propensity score matching was achieved within TriNetX using a greedy 1:1 nearest neighbour and a caliper of 0.1.Matching for a covariate was considered appropriate if a standardised mean difference of 0.1 or less was observed between matched cohorts (10).Baseline characteristics of cohorts prior to matching (lifetime, 5 years and 1 year before inclusion) were checked (Supplementary Tables 3-5).

Negative control outcomes
To assess for potential unmeasured confounding, matched cohorts were also compared for a range of negative control outcomes, i.e. outcomes that are not expected to be influenced by differences in anti-constipation medications (11).20 negative control outcomes were selected, adapted from previous similar analyses using the TriNetX database (12).These were designed to include both emergency and primary care, and various body systems and mechanisms: cutaneous abscess, ganglion, ingrowing nail, onycholysis, sebaceous cyst, senile keratosis, trigger finger, viral warts, insect bite, adhesive capsulitis of shoulder, blepharitis, fracture of hand, folliculitis, knee injury, paronychia, lateral epicondylitis (tennis elbow), otitis externa, nasal polyp, seborrheic dermatitis, and finger laceration.The incidence of a first diagnosis of each negative control outcome was assessed individually within a cohort, and also as a composite of any of these outcomes (to increase statistical power and thereby increase the chance of detecting small but significant effects of residual confounding).

Additional comparisons
The two anti-constipation comparators (linaclotide and lubiprostone) were compared with each other in terms of their effect on mental health and negative control outcomes to assess potential equivalence in incidence.
Prucalopride was licensed by the FDA more recently than linaclotide and lubiprostone in the US (prucalopride 2018; lubiprostone 2006; linaclotide 2012), and the prescription of prucalopride may be affected by social factors relating to Medicaid / Medicare availability.In order to control for this, linaclotide was compared with plecanatide.Plecanatide is a drug with the same mechanism of action and indication as linaclotide, and a similar Medicaid approval year as prucalopride: both prucalopride and plecanatide were approved for Medicaid / Medicare between 2018 and 2022, compared to several years earlier for linaclotide and lubiprostone (both available from 2014).To assess whether these factors played a role in the observed association, we ran additional analyses aiming to isolate the effect of these additional elements from the effect of 5-HT 4 agonism.Cohort limitations prevented direct comparison of prucalopride and plecanatide (there were too few participants in each cohort to allow for them to be well-matched as a comparison).For analyses involving plecanatide, plecanatide use was not matched across cohorts.

Interrupted time series analysis
We complemented the emulated target trial analysis with an interrupted time series analysis comparing the trend in depression incidence in the 12 months after versus before the first prescription of prucalopride.The incidence was reported as the total number of cases of depression (N d ) relative to the total number of people who had at least one health encounter within each month (N v ).The linear regression used was: where I d =N d /N v is the incidence of depression measured each month, T is the time in months parameterised so that T=0 is the time of prescription of prucalopride, T=0.5, 1.5, 2.5, etc represent the time in months since prescription, and T=-0.5, -1.5, -2.5, etc represent the time in months before prescription, and Xt is a binary variable representing the period after (X t =1) or before (X t =0) prescription of prucalopride.The coefficient of interest was the change in slope .We hypothesised a change in slope with a progressive reduction in the  3 number of depression diagnoses after initiation of prucalopride.This analysis was also conducted for negative control outcomes.
Auto-correlation was assessed by examining the auto-correlation function and formally tested using the Breusch-Godfrey test.(13) Seasonality was not deemed relevant as time relative to prescription was used rather than calendar time.Incidences in the month just before and the month just after initiation of prucalopride were ignored to avoid possible contamination by increased depression incidence resulting from untreated constipation and to account for delays between prescription and initiation of treatment.We tested the robustness of the findings to inclusion of these two months, as well as to the use of absolute counts rather than relative incidences (i.e.N d rather than N d /N v ) in the time series (and using a quasi-Poisson link function in the regression).

Methods 6 -Statistical analysis
Cumulative incidences over the one-year follow-up were estimated using the Kaplan-Meier estimator.The logrank test was used to compare survival between matched cohorts and the Cox proportional hazard model was used to estimate hazard ratios (HR) with 95% confidence intervals (CI) (computed using the "survival" package in R 4.2.1).The proportional hazard assumption was tested with the generalised Schoenfeld approach, and time-varying hazard ratios were calculated using natural cubic splines fitted to the log-cumulative hazard to identify where there was evidence of non-proportionality of hazards in the main comparison between the cohort of patients receiving prucalopride and a matched cohort receiving an alternative anti-constipation agent.Evalues were calculated for all comparisons in the primary analysis.(14) Statistical significance was set at two-sided p<0.05.Except for propensity score matching estimated within TriNetX, all statistical analyses were conducted using R version 4.2.1 from 25 th January 2024.Table 1: Specification of the target trial and emulation of this using electronic health records

Protocol components Target trial specification Emulation of the target trial Eligibility criteria -inclusion
Individuals over the age of 18 Individuals over the age of 18 Eligibility criteria -exclusion Pre-existing diagnoses of common and / or serious mental illness, as defined as any of F20-F29 (schizophrenia, schizotypal, delusional, and other non-mood psychotic disorders), F30-F39 (mood [affective] disorders), F40-F48 (anxiety, dissociative, stress-related, somatoform and other nonpsychotic mental disorders), or F01-F09 (mental disorders due to known physiological conditions including cognitive disorders).
We also ran robustness analyses with additional exclusions of contraindications for study medication, previous use of prucalopride in comparator cohorts, overlap with other study arm following index prescription, neuropsychiatric disease, and recent prescription of the two most frequently prescribed SSRI antidepressants.

Treatment assignment
Patients randomly assigned to prucalopride, linaclotide or lubiprostone We assume random assignment of treatment as all three medications are third line approved medications for chronic constipation.
All cohorts were also matched using propensity score matching for disorders and medication that could be associated with differences in choices of anticonstipation treatment and / or with psychiatric disorders.In addition, matching occurred for any diagnoses or medication where the baseline incidence was greater than 5% in both cohorts and the difference between cohorts greater than 2% at the prematching stage.
Baseline characteristics of cohorts prior to matching for 121 variables (lifetime, 5 years and 1 year before inclusion) were checked.Effectiveness of prucalopride on longitiduinal incidence of depression compared to comparators estimated as a cumulative incidence over the one-year outcome period using the Kaplan-Meier estimator.The Cox proportional hazard model and the log-rank test would be used to compare matched cohorts in terms of hazard ratios (HR) with 95% confidence intervals.
Effectiveness of prucalopride on future incidence of depression compared to comparators estimated as a cumulative incidence over the one-year outcome period using the Kaplan-Meier estimator.The Cox proportional hazard model and the log-rank test are used to compare matched cohorts in terms of hazard ratios (HR) with 95% confidence intervals.For individuals within each cohort, the subsequent prescriptions that were recorded are shown.The inner ring shows the prescription immediately after the index; the outer "bursts" show third and subsequent prescriptions.

Figure 2 :
Figure 2: Kaplan-Meier curves showing outcome probability over 12 months for prucalopride versus lubiprostone for each secondary mental health outcome and composite negative control outcome

Figure 3 :
Figure 3: Sunburst plots showing participants in primary analysis cohorts who had subsequent prescriptions of the same and / or comparator agents over the following 1 year Statement-Checklist of items that should be included in reports of cohort studies Other informationFunding 22 Give the source of funding and the role of the funders for the present study and, if applicable, for the original study on which the present article is based 16 *Give information separately for exposed and unexposed groups.Note: An Explanation and Elaboration article discusses each checklist item and gives methodological background and published examples of transparent reporting.The STROBE checklist is best used in conjunction with this article (freely available on the Web sites of PLoS Medicine at http://www.plosmedicine.org/,Annals of Internal Medicine at http://www.annals.org/,and Epidemiology at http://www.epidem.com/).Information on the STROBE Initiative is available at http://www.strobe-statement.org.

Table 2 :
Baseline comparison statistics used for matching in primary analyses

Table 3 :
Baseline characteristics before matching for cohorts involved in primary analyses.Comorbidities and medications are lifetime prevalence

Table 4 :
Comorbidities and medications before matching for prucalopride versus linaclotide cohorts involved in primary analyses, restricted to 5 years and 1 year before analysis

Table 5 :
Comorbidities and medications before matching for prucalopride versus lubiprostone cohorts involved in primary analyses, restricted to 5 years and 1 year before analysis

Table 6 :
Results for the different robustness analyses in terms of incidence and hazard ratios of depression diagnosis.values for hazard ratios and p-values highlight statistically significant results.

Table 7 :
Excluding those with contraindications for study medications prior to index dateOutcomes for prucalopride versus linaclotide / lubiprostone: percentage with each diagnosis during the 1-year outcome period and hazard ratio.Incidence shown with prior diagnosis of mental illness.Bold indicates significant difference between cohorts in HRs.NCO = composite negative control outcome

Table 8 :
Excluding those with a prescription of prucalopride prior to initiation of linaclotide or lubiprostoneOutcomes for prucalopride versus linaclotide / lubiprostone: percentage with each diagnosis during the 1-year outcome period and hazard ratio.Incidence shown with prior diagnosis of mental illness.Bold indicates significant difference between cohorts in HRs.NCO = composite negative control outcome

Table 9 :
Excluding those with recorded overlap of comparison medication for 1 year following index date (i.e.prucalopride for comparator cohorts; linaclotide or lubiprostone for prucalopride cohort)Outcomes for prucalopride versus linaclotide / lubiprostone: percentage with each diagnosis during the 1-year outcome period and hazard ratio.Incidence shown with prior diagnosis of mental illness.Bold indicates significant difference between cohorts in HRs.NCO = composite negative control outcome

Table 10 :
Exclusion of additional neuropsychiatric illnessOutcomes for prucalopride versus (i) linaclotide and (ii) lubiprostone: percentage with each diagnosis during the 1-year outcome period and hazard ratio.Incidence shown with no prior neuropsychiatric illness.Bold indicates significant difference between cohorts in HRs.NCO = composite negative control outcome

Table 11 :
Exclusion of two most common SSRIsOutcomes for prucalopride versus (i) linaclotide and (ii) lubiprostone: percentage with each diagnosis during the 1-year outcome period and hazard ratio.Incidence shown with exclusion of two most common SSRIs for 6 months prior to index date.Bold indicates significant difference between cohorts in HRs.NCO = composite negative control outcome

Table 12 :
Negative control outcomesPrucalopride versus (i) linaclotide and versus (ii) lubiprostone for each negative control outcome over 12 months follow up

Table 13 :
Lubiprostone versus linaclotideOutcomes for lubiprostone versus linaclotide: percentage with each diagnosis during the 1-year outcome period and hazard ratio.Incidence shown without previous mental illness.Bold indicates significant difference between cohorts in HRs

Table 14 :
Plecanatide versus linaclotideOutcomes for plecanatide versus linaclotide: percentage with each diagnosis during the 1-year outcome period and hazard ratio.Incidence shown without previous mental illness.Bold indicates significant difference between cohorts in HRs

Table 15 :
Interrupted time series analysisTrend in depression diagnosis or any negative control outcomes (NCOs) incidence (and absolute counts) in the 12 months after versus before the first prescription of prucalopride.Incidences are reported as number of cases divided by the total number of people who had at least one health encounter within each month.Bold indicates significant difference.Negative slope value indicates the outcome was less likely after versus before prucalopride prescription; positive slope value indicates the outcome was more likely after versus before prucalopride prescription.