The aim of the present study was to determine contemporary survival in pulmonary arterial hypertension (PAH), and to investigate whether or not the National Institutes of Health (NIH) equation remains an accurate predictor of survival.
In 576 patients with PAH referred during 1991–2007, observed survival was described using the Kaplan–Meier method. In patients with idiopathic, familial and anorexigen-associated PAH (n = 247), observed versus NIH equation predicted survival was compared. A new survival prediction equation was developed using exponential regression analysis.
The observed 1-, 3- and 5-yr survival in the total cohort were 86, 69 and 61%, respectively. In patients with idiopathic, familial and anorexigen-associated PAH, the observed 1-, 3- and 5-yr survival (92, 75 and 66%, respectively) were significantly higher than the predicted survival (65, 43 and 32%, respectively). The new equation (P(t) = e-A(x,y,z)t, where P(t) is probability of survival, t the time interval in years, A(x,y,z) = e(−1.270–0.0148x+0.0402y–0.361z), x the mean pulmonary artery pressure, y the mean right atrial pressure and z the cardiac index) performed well when applied to published contemporary studies of survival in PAH.
Contemporary survival in the PAH cohort was better than that predicted by the NIH registry equation. The NIH equation underestimated survival in idiopathic, familial and anorexigen-associated PAH. Once prospectively validated, the new equation may be used to determine prognosis.
Pulmonary arterial hypertension (PAH), a debilitating disease characterised by progressive obstruction and obliteration of the pulmonary arteries, eventually leads to right ventricular failure and death 1. PAH can be idiopathic, familial or associated with other conditions, including connective tissue disease, congenital heart disease, portal hypertension, HIV and anorexigen exposure 1. A landmark National Institutes of Health (NIH) registry study, published in 1987, described the clinical characteristics and natural history of patients with primary pulmonary hypertension (PPH), which included idiopathic, familial and anorexigen-associated PAH 2. The NIH registry proposed an empirically derived equation, based on baseline haemodynamics, for the estimation of survival in patients with PPH 3. Subsequently, many clinical trials in PAH have used the NIH equation to suggest improvements in survival by comparing the observed survival rates on a study drug versus the survival rates predicted by the NIH equation 4–9.
The NIH registry was initiated during a time when there were no US Food and Drug Administration-approved therapies for PAH. Patients in the NIH registry were treated only with conventional therapy, which included diuretics, digoxin, supplemental nasal oxygen and, in a minority of cases, anticoagulation with warfarin and/or vasodilators, such as calcium-channel blockers and hydralazine 2, 3. Since the mid-1980s, however, the management and therapies of PAH have changed significantly 10.
The aim of the present study was to: 1) characterise contemporary survival in patients with World Health Organization (WHO) category I PAH; and 2) determine whether or not the NIH equation remains an accurate predictor of survival in patients with idiopathic, familial and anorexigen-associated PAH (previously classified as PPH).
Patients in the Pulmonary Hypertension Connection (PHC) registry, which we initiated in March 2004, were studied. This database has been described in detail previously 11. Briefly, the PHC registry was created as a customised patient database in order to longitudinally collect specific variables on every patient treated in the authors’ practice. The practice has been sequentially based at three university medical centres in Chicago since 1982. Patients were entered into the database retrospectively from 1982 to February 2004 and prospectively from March 2004 onwards. The recruitment period for the current analysis ended in August 2007. All new patients gave informed consent for participation in the registry during the initial evaluation. Patients who were already being treated at the practice prior to the initiation of the registry gave informed consent during a routine office visit. Each local institutional review board approved the registry.
From the PHC registry, all adult patients (aged ≥18 yrs at the time of referral) with PAH (n = 654) were identified. In brief, the diagnosis of PAH required the following: 1) a mean pulmonary artery pressure (P̄pa) of >25 mmHg at rest, with a pulmonary capillary wedge pressure of <15 mmHg; and 2) the exclusion of other WHO categories of pulmonary hypertension by clinical evaluation and objective tests 12. In order to make the study cohort more comparable to the NIH registry, 52 patients who were on approved PAH therapy at the time of referral (17 on prostacyclins, 27 on bosentan and eight on sildenafil) were excluded. In addition, 26 patients diagnosed with PAH before 1991 were excluded in order to avoid entering patients who may have been in the NIH registry cohort (recruitment 1981–1985, with follow-up to 1988 and publication in 1991). The remaining 576 patients formed the study group, 100 (17%) of whom were studied prospectively (enrolled after initiation of the registry in March 2004). Of the 576 patients in the study group, 282 (49%) with idiopathic, familial and anorexigen-associated PAH formed a subgroup that matched the NIH registry. Patients with anorexigen-associated PAH were included in the subgroup since the NIH registry included patients exposed to anorexigens, and based on the recent study of Souza et al. 13 demonstrating no significant survival difference between anorexigen-associated PAH and idiopathic and familial PAH patients. Figure 1⇓ illustrates the study flow chart.
The following baseline variables were analysed at the time of referral for characterisation of clinical phenotype: demographic data, including age and sex; comorbid conditions; WHO functional class; medications; and exercise treadmill testing (ETT) using the Naughton–Balke protocol as a measure of exercise capacity 14. It has previously been shown that the ETT is comparable to the 6-min walking test and a predictor of mortality in patients with pulmonary hypertension 15, 16.
Among the patients, 521 (91%) of the 576 in the study group, and 270 (96%) of the 282 with idiopathic, familial and anorexigen-associated PAH underwent baseline haemodynamic testing by means of right heart catheterisation. More than 95% of catheterisations were performed at the University of Chicago Medical Center (Chicago, IL, USA) by pulmonary hypertension specialists. Haemodynamic testing included measurement of mean right atrial pressure (P̄RA), P̄pa, pulmonary capillary wedge pressure, cardiac index (CI) and pulmonary vascular resistance (PVR). An acute vasodilator challenge with adenosine was performed during right heart catheterisation, as described previously 17.
All patients who responded to acute vasodilator challenge were treated with calcium-channel blockers. Initially, a positive response was defined as a 20% decrease in P̄pa with an increase in CI, but, after 2005, the following definition was used: a decrease in P̄pa of >10 mmHg, and to <40 mmHg, with unchanged or increased CI 18. Patients who did not respond to the acute vasodilator challenge received either monotherapy or combination therapy with endothelin antagonists, phosphodiesterase inhibitors or prostacyclins, based on the severity of symptoms. All patients in the present cohort without contraindications were offered anticoagulation with warfarin in order to achieve a target international normalised ratio of 2–3. Patients with an arterial oxygen saturation of <90%, either at rest or during exercise, were prescribed supplemental nasal oxygen. In addition, patients received diuretics and digoxin as needed to treat symptoms of right heart failure. Patients were followed closely every 6–12 months on an outpatient basis, and more frequently if medically necessary.
Vital statistics were obtained for all patients by chart review and Social Security Death Index (SSDI). For each death, the date of death was collected. In all patients who were not identified as deceased using the SSDI, it was possible to establish vital status by chart review.
Baseline categorical variables are expressed as frequencies and proportions, whereas continuous variables are presented as mean±sd. Survival rates were calculated using the Kaplan–Meier method and standard life table analyses. The date of baseline right heart catheterisation was used as the date of entry into the study, as for the NIH registry 3. For patients who survived during follow-up, the date of data cut-off (August, 10 2007) was used as the censoring date. For patients who did not undergo initial right heart catheterisation, the date of referral to the practice was used as the enrolment date. Survival before and after 2002 were compared because bosentan, the first oral therapy for PAH, was approved for use in the USA in 2002. Prior to 2002, treatment consisted of conventional therapy and epoprostenol. For idiopathic, familial and anorexigen-associated PAH patients with available baseline haemodynamic data, expected survival was calculated for each patient based on the NIH equation:
where P(t) is the probability of survival, H(t) = 0.88−0.14t+0.01t2, t the time interval in years, A(x,y,z) = e(0.007325x+0.0526y–0.3235z), x the P̄pa, y the P̄RA and z the CI. The probabilities of survival at 1, 3 and 5 yrs are as follows: P(1) = 0.75A; P(3) = 0.55A; and P(5) = 0.43A.
The proportion of observed survival for each time period was compared with expected survival using Chi-squared tests. Univariate and multivariate Cox’s proportional hazards analyses were performed in order to determine the independent predictors of survival in the total study cohort, as well as in the subgroup of patients with idiopathic, familial and anorexigen-associated PAH. The proportional hazards assumption was tested in all models. Univariate analysis was performed with all baseline characteristic variables. For the final multivariate analyses, any variables with a p-value of <0.1 on univariate analysis were retained. In the analysis of the total PAH cohort, the subgroup of patients with idiopathic, familial and anorexigen-associated PAH was used as the referent group, to which all other aetiologies of PAH were compared. Only connective tissue disease was significant and retained in the final multivariable models. Functional class and treadmill exercise capacity were significantly correlated (correlation coefficient = 0.73, p<0.0001), and so only functional class was retained in the final models in order to avoid multicollinearity. In addition, separate models were created for each haemodynamic variable to avoid multicollinearity.
In the subgroup of patients with idiopathic, familial and anorexigen-associated PAH, a new survival prediction equation was developed, using first a Weibull model followed by an exponential regression model for ease of use 19. It was predetermined a priori that the three haemodynamic predictors (P̄pa, P̄RA and CI) from the NIH model were clinically relevant for inclusion in the model. The variables that were significant on univariate analysis were then added to these three haemodynamic variables in order to complete the model. Variables were then removed from the model one at a time, using backward elimination. Similar to the NIH equation, only P̄pa, P̄RA and CI were retained in the final model, and all three were significant.
In addition, in order to address concerns about inclusion of patients who responded to calcium-channel blocker (n = 11; all showed a positive response to the vasodilator study during the invasive haemodynamic study), a separate equation was developed for responders. The data were also analysed to determine whether exclusion of responders to calcium-channel blocker significantly altered the overall survival of the cohort.
In order to validate the new survival equation, it was applied to three published idiopathic and familial PAH patient cohorts, and the published observed survival compared with that predicted by the new survival equation based on the baseline mean haemodynamic data (P̄pa, P̄RA and CI) of these study cohorts 8, 20. All statistical analyses were performed using Stata version 9 (StataCorp, College Station, TX, USA).
All PAH patients
The mean age upon entry into the registry was 48±14 yrs, and 77% of patients were female (table 1⇓). The majority (80%) of subjects showed WHO functional class III or IV symptoms and poor baseline exercise capacity, achieving a maximum workload of only 3.6±2 metabolic equivalents of the task using the Naughton–Balke protocol ETT. At the time of referral, subjects were on conventional therapy and gave severe haemodynamic results (table 1⇓).
The median follow-up time was 3.9 yrs (interquartile range 1.7–7.8 yrs; maximum follow-up time 16.6 yrs). Of the 576 study patients, 307 (53%) died during follow-up. Vital status was available for all patients, but survival time was missing for 30 and one died on the date of enrolment; therefore, these patients were excluded, leaving 545 patients for the final survival analyses. The observed 1-, 3- and 5-yr survival rates for the total PAH cohort were 86, 69 and 61%, respectively. The 1-, 3- and 5-yr survival rates were not affected by exclusion of patients who exhibited a positive acute vasodilator response to adenosine during invasive haemodynamic testing (Fig. I of online supplementary material).
On univariate analysis in the total cohort (n = 576), age, connective tissue disease aetiology, worse functional class, decreased exercise capacity, increased P̄RA, increased PVR and decreased CI were all associated with an increased risk of death (table 2⇓). On multivariable analysis, age, connective tissue disease aetiology, functional class, P̄RA and CI all remained independent predictors of death (table 3⇓).
Subgroup (idiopathic, familial and anorexigen-associated PAH)
The mean age upon entry into the registry was 46±14 yrs (∼10 yrs older than the NIH registry subjects). The baseline demographic, clinical and haemodynamic characteristics of the idiopathic, familial and anorexigen-associated PAH subgroup were comparable to those of patients in the NIH registry (table 4⇓).
The median follow-up time in the subgroup of patients with idiopathic, familial and anorexigen-associated PAH was 4.9 yrs (interquartile range 2.3–8.7 yrs; maximum follow-up time 16.6 yrs), and, during follow-up, 149 (53%) out of 282 died. The survival time was missing for 16 patients, leaving 266 patients for survival analyses. The observed 1-, 3- and 5-yr survival rates were 91, 75 and 65%, respectively, and were better than for patients with PAH associated with connective tissue disease, congenital heart disease, portal hypertension and HIV (fig. 2⇓). Patients diagnosed with idiopathic, familial and anorexigen-associated PAH after 2002 exhibited better survival than those diagnosed before 2002 (fig. II of online supplementary material). Of the 282 subgroup patients with idiopathic, familial and anorexigen-associated PAH, all of the three baseline haemodynamic data (P̄RA, P̄pa and CI) required to calculate survival using the NIH equation were available for 247 patients (P̄pa 55±12 mmHg, P̄RA 11±6 mmHg and CI 2.0±0.6 L·min−1·m−2). The observed 1-, 3- and 5-yr survival of these patients with available haemodynamic data were 92, 75 and 66%, respectively, which did not differ significantly from that of those without haemodynamic data. In contrast, the predicted 1-, 3- and 5-yr survival rates, calculated using the NIH equation, were 65, 43 and 32%, respectively. The observed survival was significantly greater than the predicted survival at 1, 3 and 5 yrs (p<0.0001 for all).
On univariate analysis in the subgroup (n = 282), age, worse functional class, decreased exercise capacity, increased P̄RA and decreased CI were all associated with an increased risk of death (table 2⇑). On multivariable analysis, age, functional class, P̄RA and CI remained as independent predictors of death (table 5⇓).
Development of a new survival equation: the PHC equation
In the subgroup of patients with idiopathic, familial and anorexigen-associated PAH, the following equation was derived to predict an individual patient’s chances of survival based on exponential regression analysis:
where A(x,y,z) = e(−1.270–0.0148x+0.0402y–0.361z) in nonresponders to calcium-channel blockers and A(x,y,z) = e(−3.012–0.0148x+0.0402y–0.361z) in responders to calcium-channel blockers.
A patient with a P̄pa of 40 mmHg, P̄RA of 3 mmHg and CI of 3.5 L·min−1·m−2, and who did not respond to an acute vasodilatory challenge, would have 1-, 2- and 3-yr survival estimates as follows:
giving probabilities of survival at 1, 2 and 3 yrs of P(1) = e-0.049544×1 (0.952 or 95%); P(2) = e-0.049544×2 (0.905 or 90.5%); and P(3) = e-0.049544×3 (0.861 or 86%).
A patient with exactly the same haemodynamics as the patient in the first example, but who showed a positive response to an acute vasodilatory challenge, would have 1-, 2- and 3-yr survival estimates as follows:
giving P(1) = e-0.008679×1 (0.991 or 99%); P(2) = e-0.008679×2 (0.983 or 98%); and P(3) = e-0.008679×3 (0.974 or 97%).
Sensitivity analyses demonstrated that the Cox's proportional hazard estimates of the effects of the predictor variables were virtually identical to those from the Weibull or exponential models of the equation. Figure 3⇓ displays the observed versus predicted survival (using both the new equation and the NIH equation) in the subgroup of patients with idiopathic, familial and anorexigen-associated PAH. When the new equation was applied to data from other published PAH patient cohorts, the predicted survival calculated using the new equation was comparable to the observed survival (fig. 4⇓).
It has been shown that survival rates in patients with WHO category I PAH have improved considerably compared to those in the NIH registry cohort. The observed 1-, 3- and 5-yr survival rates for the total PAH cohort were 86, 69 and 61%, respectively. This is the first evaluation of survival of WHO category I PAH patients. In patients with idiopathic, familial and anorexigen associated PAH, the observed survival rates at 1, 3 and 5 yrs were significantly higher than the predicted survival calculated using the NIH equation, and patients diagnosed after 2002 appeared to show better survival than those diagnosed before 2002. From the present cohort, a new regression equation was developed to estimate survival, based on baseline haemodynamic data, in patients with idiopathic, familial and anorexigen-associated PAH. The survival calculated using this new equation performed well when applied to other published patient cohorts 8, 20.
The NIH registry, which prospectively collected data on 187 patients with PPH from 32 centres in the USA between July 1981 and September 1985, described the clinical characteristics of PPH and its natural history over a 7-yr period in an era when there were no approved PAH-specific therapies. In the NIH registry, patients showed a median survival time of 2.8 yrs. The 1-, 3- and 5-yr survival rates were only 68, 48 and 34%, respectively 3. Mortality correlated with baseline P̄RA, P̄pa and CI. The NIH registry proposed an equation for prediction of survival, which was subsequently validated by Sandoval et al. 21 in a small Mexican cohort of 61 patients with PPH. In this study, the predicted and observed survival were closer in the patients who received no long-term oral vasodilatory therapy, but the predicted survival underestimated the observed survival in the group as a whole.
The improved survival in the present patients with WHO category I PAH, and the subgroup analysis, compared to the NIH registry cohort may be due to a multitude of synergistic factors. The number of PAH patients referred to tertiary care centres has increased compared to the NIH registry time period, and physicians are much more familiar with treating right heart failure. However, since most patients were still referred late (functional class III or IV), it may be that the natural history of the disease itself has changed. Aetiology had a small impact on survival. This may be due to the interaction with the other variables in the multivariable model lessening its effect, and the other factors may be stronger predictors. Another factor for improved survival may be the greater use of anticoagulation with warfarin. PAH treatment guidelines recommend warfarin for patients with idiopathic, familial and anorexigen-associated PAH 10 based on two small retrospective clinical trials showing improved survival in PAH patients treated with long-term warfarin therapy 22, 23, but it has not been validated in a prospective randomised clinical trial. Early detection with echocardiography and heightened awareness of PAH may have increased the number of patients evaluated and treated earlier in the course of the disease, explaining their presentation to a referral centre a decade after the NIH registry. Although earlier intervention may translate into improved survival, the possibility cannot be excluded that lead-time bias is also a factor. However, patients in the present cohort were older than those studied in the NIH registry, and showed a similar severity of functional class and haemodynamic abnormalities at the time of initial presentation, arguing against lead-time bias due to earlier diagnosis. This also argues against the possibility of a selected population of less severely diseased patients surviving, a cohort effect as an explanation for the improved survival.
The present analyses included patients with a positive vasodilatory response, an independent predictor of survival 7, 18, 23, 24. However, the number of responders was significantly lower in the total study cohort (13 (2.3%) out of 576), as well as in the subgroup of patients with idiopathic, familial and anorexigen-associated PAH (11 (3.9%) out of 282). In addition, the number of responders across eras in the present database was consistent and low (3.7, 4.2 and 5.8%; p = ns)) 11. Excluding responders did not affect survival rates significantly, which is probably due to the small numbers of patients who showed a positive acute vasodilatory response during invasive haemodynamic testing (fig. I of online supplementary material). We, therefore, do not believe that a positive acute vasodilatory response has played a significant role in improved survival. However, given the much better survival in patients who show a positive acute vasodilatory response, a separate new equation was derived for responders.
Finally, the improved survival may be due to the availability of approved PAH medications, which include prostacyclin and its analogues, endothelial antagonists and phosphodiesterase inhibitors 10. It remains unknown whether or not the current PAH-specific therapies improve long-term survival, as there are no long-term randomised treatment trials with any therapy for PAH. The present study is an observational study, and thus the true effect of new therapies on survival cannot be evaluated. A recent meta-analysis by Galie et al. 25 suggested an improvement in survival in patients treated with PAH-specific therapies (relative risk of all-cause mortality 0.57 (95% confidence interval 0.35–0.92); p = 0.023). In contrast, in a meta-analysis by Macchia et al. 26, none of the currently available PAH-specific therapies were associated with long-term survival benefit 26. Thus, in order to ascertain whether or not there are any long-term survival benefits with the currently available newer PAH-specific therapies, prospective randomised placebo-controlled trials, which evaluate clinical outcomes, including true surrogates of death if not death itself, stratified to specific aetiologies of PAH are required 26. A new formula gives us a new baseline against which to make comparisons once it has been validated in independent cohorts.
It was decided to revise the NIH formula only when better-than-predicted survival was found, understanding that this was only a registry. The equation included anorexigen-associated PAH since the NIH registry included patients exposed to anorexigens, and based on the recent Souza et al. 13 study demonstrating no significant survival difference between anorexigen-associated PAH and idiopathic and familial PAH patients. If prospectively validated in another idiopathic, familial and anorexigen-associated PAH cohort, the new equation could be used to estimate contemporary survival in this patient population. In addition, the new survival equations could potentially be used in clinical and research settings in order to determine response to therapy and changes in prognosis that occur with changes in invasive haemodynamic profile. Survival in WHO category I PAH patients as a whole should not be evaluated using these equations since congenital heart disease-associated PAH carries a better prognosis, and connective tissue disease-associated PAH a worse prognosis than other forms of PAH 27, 28.
Data collection started in 2004, and thus most patients were entered retrospectively. All patients in the cohort were cared for by a single tertiary referral practice; therefore, the results may not be generalisable. The cohort consisted of both incident and prevalent cases, which may have contributed to a survival bias. Subsequent treatment data were not collected for individual patients; hence the proportion of patients treated with different PAH-specific therapies is not available. Thus, the study does not address the key issue of whether or not the current PAH-specific therapies affect survival. A total of 9% of the study cohort did not have baseline invasive haemodynamic data; however, the majority of these patients underwent follow-up right heart catheterisation, which confirmed the diagnosis of PAH. Finally, the new equation was developed based on a heterogeneous patient collection on a variety of PAH-specific therapies. Thus it may not be useful for demonstrating improved survival with a new drug in clinical trials, but may be helpful for clinical prognostication if prospectively validated.
Contemporary survival in patients with PAH is significantly better than that in the NIH registry cohort. The NIH equation underestimates survival in patients with idiopathic, familial and anorexigen-associated PAH in the present PHC registry. If prospectively validated, the new regression survival equations (PHC equations) should assist in determining prognosis in patients with idiopathic, familial and anorexigen-associated PAH.
This study was supported by grant number UL1RR024999 from the National Center for Research Resources (Bethesda, MD, USA). S.J. Shah was supported by a Heart Failure Society of America (Saint Paul, MN, USA) Research Fellowship Award, Actelion Entelligence (San Francisco, CA, USA) Young Investigator Award and an American Heart Association (Dallas, TX, USA) Scientist Development Grant. S.L. Archer's research is supported by the Canadian Institutes for Health Research (Ottawa, ON, Canada) and National Institutes of Health (Bethesda, MD, USA) grant HL071115. The content of the present article is solely the responsibility of the authors and does not necessarily represent the official views of the National Center for Research Resources or the National Institutes of Health. The funding sources played no role in the collection, analysis and interpretation of data, the writing of the report or the decision to submit the paper for publication.
Statement of interest
Statements of interest for M. Gomberg-Maitland, S. Rich and S.J. Shah can be found at www.erj.ersjournals.com/misc/statements.dtl
We thank R.A. Thisted (Dept of Health Studies, University of Chicago, Chicago, IL, USA) for expert statistical review of the survival analysis.
For editorial comments see page 958.
This article has supplementary material accessible from www.erj.ersjournals.com
- Received May 4, 2009.
- Accepted December 10, 2009.
- © ERS Journals Ltd