Selective Serotonin Reuptake Inhibitors and the Incidence and Outcome of Pulmonary Hypertension FREE TO VIEW

Patients with pulmonary hypertension (PH) have increased circulating levels of serotonin, which is a pulmonary vasoconstrictor and vascular smooth muscle cell mitogen.^1–3 The high levels of serotonin in patients with PH persist after lung transplantation and after treatment with prostacyclins, suggesting that increased serotonin levels are not simply due to pulmonary vascular disease.^4,5 Anorexigens, which increase serotonin levels through the stimulation of the serotonin transporter (5HTT) on the surface of platelets, have also been associated with the development of PH.⁶ In addition, studies of genetic polymorphisms of the serotonin transporter gene have provided further evidence of the role of serotonin in PH.^7–11

The experimental data^12–15 supporting the ability of selective serotonin reuptake inhibitors (SSRIs) to decrease the development of and reverse PH in animal models of PH is compelling. It appears that the central pathogenic mechanism in the serotonin pathway is increased activity of 5HTT.¹ In animal models, overexpression of the 5HTT gene leads to worsening PH,^16,17 and disruption of the gene reduces PH.⁵ Although anorexigens and SSRIs both increase serotonin levels, the former increases the efflux of serotonin through activation of 5HTT in platelets, whereas the latter blocks the reuptake of serotonin by 5HTT.¹ Therefore, SSRIs could be beneficial in patients with PH because they block 5HTT, which is a key regulator in serotonin-induced vascular smooth muscle cell mitogenesis.

The efficacy of SSRIs as a novel therapeutic agent for patients with PH is unknown. In the International Primary Pulmonary Hypertension Study,⁶ which was designed to evaluate the risks of anorexigens as a cause of PH, there was noted to be a possible protective effect of SSRI use against the development of PH (odds ratio [OR], 0.1; 95% confidence interval [CI], 0.01 to 1.1).¹⁸ Consequently, we sought to determine whether additional observational clinical data would support the hypotheses that SSRI use is protective against the development of PH and/or whether SSRI use in patients with PH is associated with improved survival. To study these hypotheses, we examined data from two cohorts, the Surveillance of Pulmonary Hypertension in America (SOPHIA) and the Pulmonary Hypertension Connection (PHC) registries.

The SOPHIA registry (n = 1,335) has been described in detail previously.¹⁹ Briefly, the study consisted of a network of 13 PH centers with the aim of determining whether anorexigen use in the United States was associated with an identifiable increase in the incidence of PH. All patients who were referred to specialists at these 13 centers with suspected PH from 1998 to 2001 were enrolled in the SOPHIA registry if they had undergone a thorough evaluation, including cardiac catheterization.²⁰ PH was diagnosed if the mean pulmonary artery pressure was > 25 mm Hg. All World Health Organization (WHO) categories of PH were included except for category II (pulmonary venous hypertension); these patients were excluded from the SOPHIA analysis. Therefore, WHO categories I, III, IV, and V were classified as PH. Patients referred but found not to have a diagnosis of PH were classified as having no PH. Phone interviews were conducted to ascertain the medications being used by the patients, and the accuracy of medication lists was then confirmed during the study visit with the PH specialist. After the completion of cardiac catheterization and a full clinical evaluation, the presence or absence of PH was documented. The use of the SOPHIA registry database was approved by the steering committee chair on behalf of the investigators who participated in the SOPHIA study.

We treated our analysis of the SOPHIA registry as a case-control study. Cases were patients with confirmed PH and control subjects were patients with no PH. We compared demographics and clinical characteristics between these two groups, and we compared the frequencies of SSRI use between these two groups. Using logistic regression, we compared case patients and control subjects and determined the OR for SSRI use in the case patients (ie, the confirmed-PH group), to investigate whether patients with PH had a lower likelihood of receiving SSRIs. Next, we used multivariable logistic regression to determine the likelihood of SSRI use in the confirmed-PH group after adjusting for age, gender, race/ethnicity, study site, and obesity (defined as a body mass index [BMI] of ≥ 30 kg/m²).

The PHC registry, which was initiated in March 2004, has been described previously.^21,22 Briefly, all patients who had been evaluated at a single US practice over time at three different university hospitals between 1982 and 2006 were entered into the database. The PHC registry was approved by the respective institutional review boards based on the location of the practice, and all actively seen patients gave informed consent to be entered into the registry.

A total of 1,360 patients who had been referred to our practice from 1982 to 2006 for the evaluation of PH were entered into the registry (812 patients were entered retrospectively, and 548 patients were entered prospectively). We collected baseline data on demographics, clinical phenotype, medication use, exercise test results, and cardiac catheterization results on all patients seen by our practice. From the overall PHC registry, we identified all patients who had documentation of SSRI use (all patients evaluated by the clinic after 1998; n = 571).

Of the 571 patients with PH and documentation of SSRI use, 15 patients were excluded due to age < 18 years (consistent with the SOPHIA registry), and 24 patients were excluded due to lack of follow-up information. The remaining 542 patients comprised our study cohort (55% were entered retrospectively [prevalent cohort], and 45% were entered prospectively [incident cohort]).

For each patient, the medication list from the first visit in our PH clinic was reviewed for SSRI use, and if the patient was receiving SSRIs, the name and type of SSRI was documented. Based on prior studies, each SSRI was categorized as medium-affinity (dissociation constant, > 1 nmol) or high-affinity (dissociation constant, < 1 nmol) based on its ability to bind to and competitively inhibit the serotonin transporter.^23,24 Sertraline, fluoxetine, and paroxetine were classified as high-affinity SSRIs. All others (citalopram and escitalopram) were classified as medium-affinity SSRIs.

We analyzed the following baseline variables at the time of referral for the characterization of clinical phenotype: demographic data, including age and sex; comorbidities; WHO functional class; medications used; and exercise treadmill testing result using the Naughton-Balke protocol.²⁵ Of the 542 subjects included in our analysis, 413 underwent complete baseline hemodynamic testing by right heart catheterization.

Vital statistics were collected for all patients by chart review and by query of the Social Security Death Index. For each death, the date of death was documented. In all patients who were not identified as deceased by the Social Security Death Index, we were able to confirm vital status by chart review.

We first compared groups of patients depending on baseline SSRI use, using t tests (or a nonparametric equivalent when necessary) for continuous variables and χ² or Fisher exact tests for categorical variables. To determine the univariate and multivariate risk of death by SSRI use, we used a Cox proportional hazards analysis, and the proportionality assumption was tested and confirmed for all models. Variables associated with SSRI use at p < 0.05 and known predictors of mortality in PH²⁶ were included in our multivariate models. We also conducted further analyses to determine the effect of SSRI affinity (ability to inhibit the serotonin transporter) on mortality. All statistical analyses were performed using a statistical software package (Stata, version 9; Stata Corp; College Station, TX).

Table 1 shows the demographic and clinical features of patients in the confirmed-PH and no-PH groups in the SOPHIA registry, and the difference in SSRI use between these groups. Patients with confirmed PH were less likely to be receiving SSRIs compared to those patients who did not to have PH (p = 0.003). After multivariable adjustment, patients in the confirmed-PH group still had a decreased OR of SSRI use compared with those in the no-PH group, although this finding did not meet statistical significance (p = 0.09) [Table 2]. The majority of SSRI use in the SOPHIA registry (96%) was with high-affinity SSRIs.

Table 3 lists the demographic, clinical, and hemodynamic characteristics of patients by SSRI use in the PHC registry. Approximately one-half of the patients (49%) had WHO group I pulmonary arterial hypertension (PAH). Sixty-nine patients (13%) in the entire cohort were receiving SSRIs at baseline. Of the 69 patients receiving SSRIs, 48 patients (70%) were receiving high-affinity SSRIs (sertraline, 24 patients [35%]; paroxetine, 14 patients [20%]; and fluoxetine, 10 patients [15%]), 18 patients (26%) were receiving medium-affinity SSRIs (citalopram, 9 patients [13%]; escitalopram, 9 patients [13%]), and 3 patients (4%) were receiving an unknown type of SSRI. Patients who were receiving SSRIs were less likely to have WHO group I PAH and were more likely to be obese, but they were otherwise similar to the group of patients who were not receiving SSRIs. Importantly, PH-specific medication use, WHO functional class, treadmill exercise time, and hemodynamics were similar between the two groups, underscoring the similarities in the risk profile between patients receiving SSRIs and those not receiving SSRIs.

We found a stepwise decrease in mortality rate from the no SSRI group to the medium-affinity SSRI group to the high-affinity SSRI group (Fig 1A) in the entire cohort, and in the subgroup of patients with WHO group I PAH (Fig 1B). There was no significant difference in mean (± SD) follow-up time in the SSRI group (3.1 ± 2.4 years; maximum follow-up time, 8.0 years) and the no-SSRI group (2.5 ± 2.1 years; maximum follow-up time, 7.9 years; p = 0.12 [Mann-Whitney U test]). However, to account for the possibility of lead-time bias, we repeated our analyses in the incident cohort (n = 237) to determine whether patients who were receiving high-affinity SSRIs were less likely to die compared to those not receiving SSRIs. We found that only 4% of the patients receiving high-affinity SSRIs died; whereas, 18% of the patients receiving no SSRIs died (p = 0.089).

On univariate and multivariate analysis (Fig 2), SSRI use emerged as an independent predictor of decreased mortality (univariate hazard ratio [HR] 0.35 [95% CI, 0.14 to 0.87]; multivariate HR, 0.35 [95% CI, 0.14 to 0.88]). This was especially true when patients receiving high-affinity SSRIs were compared with patients receiving no SSRIs (multivariate HR including hemodynamic variables, 0.22; 95% CI, 0.05 to 0.91). Figure 3 shows the cumulative hazard of death by SSRI use in the PHC registry over the 8-year time period of the study. The curves separate early, suggesting a protective benefit with SSRI use.

We have shown that SSRI use is associated with a decreased incidence of PH and decreased mortality in PH patients by using two large cohorts of patients, the SOPHIA registry and the PHC registry. In support of the beneficial effects of SSRIs in humans, we demonstrated a reduced incidence of PH in patients with risk factors (eg, connective tissue diseases and anorexigens) for PH and an improved survival time in patients with newly diagnosed PH. Our finding that increasing affinity of SSRIs for the serotonin transporter results in a stepwise decrease in mortality adds further support to the hypothesis that blockade of the serotonin transporter in humans can mimic the benefits of SSRIs in animal models of PH. We also found that patients receiving SSRIs had lower mortality in the subgroup of patients with PAH and in the subgroup of patients who were studied prospectively (incident cohort). Although the association between SSRIs and improved outcomes in the incident subgroup did not achieve statistical significance due to the smaller numbers of subjects and decreased follow-up time (with decreased statistical power), the finding in this subgroup was in the same direction as the findings in the entire PHC registry cohort and the subgroup of patients with PAH. Therefore, our subgroup analyses strengthened our observations by addressing the fact that fewer patients receiving SSRIs had PAH and that a large number of patients were entered retrospectively, two potential factors that could have biased the results of our study.

Decreased smooth-muscle cell proliferation via effects on the serotonin transporter is the most likely mechanism by which SSRIs are protective in PH patients. Other potential mechanisms for the decreased risk of mortality in patients receiving SSRIs include an antiplatelet effect and mood enhancement. SSRIs are well known to decrease platelet reactivity and aggregation²⁷ because serotonin plays a critical role in the thrombotic process.²⁸ Because hypercoagulability and thrombosis play an important role in the pathogenesis and progression of PH, the antiplatelet effects of SSRIs could be driving their association with decreased mortality. We also cannot exclude the possibility that SSRIs have a mood-enhancing effect in PH patients, thereby leading to a healthier lifestyle, and better compliance with clinic visits and medications. This may be important in PAH patients, in whom depression is common, and compliance with complicated and risky IV therapies is difficult.

Consistent with our observations is a prior study²⁹ that examined the association of SSRIs and mortality in patients with PH by examining mortality risk in a small cohort of 84 patients, 13 of whom (15%) were receiving SSRIs. Although this study found an HR for mortality of 0.53 in patients receiving SSRIs vs those not receiving SSRIs, the 95% CIs were wide (0.07 to 3.9; p = 0.53), which may have been due to the fact that this study was underpowered.

There have been reports^30,31 of an increased incidence of PH in newborns whose mothers were receiving SSRIs during late pregnancy. The mechanism of this effect of SSRIs is unclear, but it is likely due to increased fetal levels of circulating serotonin that may cause pulmonary vasoconstriction in utero or shortly after birth.³² In adults, SSRIs block the serotonin transporter, which decreases smooth muscle cell mitogenesis and proliferation (thereby improving PH); whereas, in the neonate SSRIs most likely cause persistent pulmonary vasoconstriction due to increased fetal serotonin levels (thereby inducing PH).

Although the findings of our study are provocative, several limitations should be considered when interpreting the results. Because only baseline SSRI use was available in both the SOPHIA and PHC cohorts, we cannot calculate “exposure time” for SSRIs. However, any misclassification of SSRI use (due to the initiation of SSRI therapy after the baseline visit) would have biased our study in an opposite direction. The control group from the SOPHIA registry was composed of patients who were referred for the evaluation of suspected PH, but who, on evaluation, were found not to have it. We believed that this control group was more appropriate than a group of age- and sex-matched healthy control subjects because these patients should have a similar referral bias. Also, SOPHIA was powered to detect an altered risk from exposure to anorexigens, and not from antidepressants. However, it is possible that the control group in SOPHIA had a higher incidence of depression-related somatization (eg, unexplained fatigue or dyspnea) that led to evaluation for PH, in which case SSRI use might be a marker of depression.

The PHC registry portion of our study was also limited by decreased statistical power in the incident subgroup analysis. Therefore, although it appeared that the use of high-affinity SSRIs were associated with decreased mortality in the incident cohort subgroup (n = 237), our findings did not achieve statistical significance (p = 0.09). Like all association studies, we cannot prove causation, and residual confounding cannot be excluded. However, we controlled for several important factors, such as the etiology of PH and abnormal hemodynamics, which are known to affect prognosis. Finally, based on the observational nature of our study, our results should be viewed as hypothesis generating, and further study of SSRIs in PH is warranted.

In summary, we found that patients with PH associated with several types of risk factors were less likely to be receiving SSRIs compared with patients who were found not to have PH. We also found that in patients with established PH, SSRI use was independently associated with decreased mortality. The consistency of our findings in the two cohorts we studied, coupled with compelling animal data and data from other observational studies,^18,29 point to SSRIs as potential therapies for the prevention and treatment of PH. We believe that prospective trials with these drugs are warranted.

Author contributions: Drs. Shah, Gomberg-Maitland, and Rich conceived and designed the study. Dr. Shah performed the statistical analyses and drafted the article. Drs. Gomberg-Maitland, Thenappan, and Rich acquired the study data. All authors participated in interpreting the data and revising the manuscript for important intellectual content. All authors approved the final version of the manuscript.

Financial/nonfinancial disclosures: Dr. Gomberg-Maitland has received research grant support from Actelion, Gilead, Lilly/Icos, Pfizer, and United Therapeutics, and he has served as a consultant for and/or on advisory boards for Biomarin, Gilead, Medtronic, and Pfizer. Drs. Shah, Thenappan, and Rich have reported to the ACCP that no significant conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article.

Other contributions: We thank Dr. James Korelitz for his thoughtful analysis of the SOPHIA database to look at the effects of SSRIs.