Findings from the ALLHAT extension study show that over the entire follow-up period, the only major outcomes that were significantly different were a higher HF rate with amlodipine compared with chlorthalidone (HR=1.12, P
=.01) and a higher stroke mortality rate with lisinopril compared with chlorthalidone (HR=1.20, P
=.03). The former result was mostly attributable to the in-trial difference and the latter result was due to the addition of events post-trial which converted a not quite significant in-trial result (HR=1.25, P
=.05) to a significant one. Thus neither for this nor any other outcome, such as CHD, was there evidence of any lagged (late emerging) effect. In the post-trial period, there were only two differences in major outcomes: a lower HF mortality rate in the amlodipine group compared with the chlorthalidone group (HR=0.71; 95% CI, 0.51–0.98, P
=.02 for heterogeneity comparing in-trial with post-trial); and a lower CVD rate in the lisinopril group compared with the chlorthalidone group (HR=0.92; 95% CI, 0.85–1.00, P
=.02 for heterogeneity comparing in-trial with post-trial). These results could be consistent with many other post-trial results wherein the medications used, including the use of diuretics, likely became more similar across the randomized groups or could be due to chance. There was no difference in HF mortality (HR=.94; 95% CI 0.77–1.17) or in CVD (HR=0.99; 95% CI, 0.94–1.05) over the entire follow-up period. Although the apparent persistence or emergence (a legacy effect) of post-trial differences may seem plausible (as discussed below), it must be noted that post-trial comparisons are no longer protected by blinded randomized therapy. Legacy effects for antihypertensive treatment due to prevention of nonfatal events, attenuation of left ventricular modeling or prevention or regression of pathological or functional changes caused by hypertension have been shown for mortality, but mainly where the comparator is placebo or usual care.6
In-trial results in this reduced cohort for all-cause mortality, stroke mortality, fatal and non-fatal CVD events, and renal outcomes were similar to what had been reported previously from the entire trial population. Notably, such similarities for lisinopril compared with chlorthalidone included higher HRs for stroke mortality (HR of 1.26 originally, 1.25 here), combined CVD (CVD plus revascularizations and hospitalized angina) (HR 1.07 originally and 1.05 as CVD was defined in this analysis), and HF (1.11 originally and 1.11 here), although the CVD and HF HR’s were not statistically significant in this smaller cohort. The similarities for amlodipine compared with chlorthalidone included the significantly higher HR for HF (1.35 originally vs. 1.37 here), significantly lower non-CV mortality (0.90 originally and 0.89 here), and significantly lower trauma mortality (0.49 originally and 0.51 here). These non-CV and trauma mortality differences disappeared post-trial, suggesting that they were related to randomized treatment, but we have no plausible explanation; these associations may merit further study.
The overall results suggest that observed in-trial differences dissipated over time as participants were taken off blinded study medications and put on open-label therapy. With the exception of the HF and stroke mortality results, there was no evidence of a legacy effect. Unfortunately, post-trial antihypertensive medications usage is unknown. It is likely that treatments became similar across randomized groups, which would cause post-trial HRs to be close to one. For the lisinopril-chlorthalidone comparison, another possibility is that participants may have received thiazide-type diuretics added to ACEIs—which could lead to decreases in CVD rates compared with those who were simply continued on a thiazide. If such altered regimens were more common among Black participants, this could explain the proportionately greater narrowing of differences for stroke, and perhaps for HF, in Black compared with non-Black participants. These sub-group results contrast with those in the amlodipine-chlorthalidone comparison, where lowered post-trial HRs for HF were proportionately similar in Black and non-Black participants.
An additional possibility is that the in-trial to post-trial difference for CVD suggests some delayed effects (or a legacy effect) for lisinopril that may make it (or any ACEI) a desirable adjunct in antihypertensive regimens.20–23
ACEIs have shown beneficial CVD effects in some other trials, especially in combination with thiazide-type diuretics, in treating patients with hypertension, diabetes, high CVD risk, or after strokes.23–26
Overall, these long-term results from passive follow-up suggest that differences in major CVD outcomes between regimens diminished once the clinical trial stopped and the antihypertensive regimens most likely became similar. In addition, for CVD and all-cause mortality, as well as most other secondary outcomes (except for stroke mortality for lisinopril versus chlorthalidone), there were no new differences that appeared as a net result of any possible intermediate effects observed or not recognized during the trial. For example, differences in effects on glucose, lipids, HF or stroke did not apparently result in overall net differences in CVD or all-cause mortality. Specific effects on post-trial outcomes will be reported further in separate papers addressing participants who developed diabetes or HF, or changes in glucose or renal function, within the ALLHAT trial; long-term renal outcomes from the United States Renal Data System (USRDS) will also be further reported.
The extended follow-up study had several strengths. ALLHAT was the largest antihypertensive randomized controlled outcome trial conducted to date, with 32,804 participants in the 3 arms. Also, ALLHAT was a double-blind, randomized, active-controlled clinical trial with well-documented procedures for outcome ascertainment, which were supplemented by national databases to capture outcomes for both post-trial and in-trial events.
The analyses of the extended data set had several limitations. Participants were taken off blinded therapies at the trial’s end, and information about medications participants used post-trial was not available. BP and laboratory data were not obtained post-trial. It should be noted that medication use and/or blood pressures in other major randomized trials with post-trial follow-up have tended to converge.8, 10
All participants were not included in the analyses, i.e., 533 Canadian participants were in neither the mortality nor combined mortality/morbidity analyses, and neither VA participants (n = 5558) nor those without a Medicare number (n=5623) were in the combined mortality/morbidity analyses except for ESRD outcomes. Event ascertainment, except for ESRD, was not the same in-trial and post-trial (but when analyses were done capturing information from databases for both in-trial and post-trial periods, results were remarkably similar). Finally, many analyses were performed and only one was significant by a strict multiple comparison standard. Given these limitations, it is noteworthy that the in-trial HRs were quite similar to those observed in the original analyses that included all the participants.
In conclusion, these long-term follow-up results from the ALLHAT trial show that significant cardiovascular outcome differences observed during the trial did not persist except for an excess of HF in the amlodipine group compared with the chlorthalidone group. No new significant differences were observed over the entire follow-up period that had not been present in-trial with the exception of a higher stroke mortality rate for lisinopril compared with chlorthalidone. Lastly, no new differences in major outcomes developed post-trial, except for slightly lower major CVD events in the lisinopril group compared with the chlorthalidone group. However, there was no difference in major CVD events or CVD mortality over the entire follow-up period. These findings, therefore, suggest that neither calcium channel blockers nor angiotensin converting-enzyme inhibitors are superior to diuretic in long-term prevention of major cardiovascular complications of hypertension.27, 28