The ALLHAT Report: A Case of Information and Misinformation
Michael A. Weber, MD, State University of New York, Downstate College of Medicine, 450 Clarkson Avenue, Box 97, Brooklyn, NY 11203
The announcement of the results of the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT)1 was literally front page news. After all, this very large clinical outcomes trial in hypertension comparing the diuretic chlorthalidone with newer agents, the angiotensin-converting enzyme (ACE) inhibitor lisinopril and the calcium channel blocker (CCB) amlodipine, had concluded that the diuretic was superior to the other drugs in preventing major cardiovascular events. More than that, the ALLHAT authors pointed out that because thiazide-like diuretics like chlorthalidone are so inexpensive, they have the double advantage of being cheaper as well as better than the other drug classes.
Right from the moment of publication, though, experts in hypertension were surprised at a result that appeared to be in conflict with data from previous carefully conducted clinical trials. And, as it has become possible to digest the lengthy and detailed ALLHAT report, serious questions have arisen not only as to the accuracy of the original claims, but also as to the propriety of announcing them in so flamboyant a fashion. Those of us who have advocated the value of diuretics, either as single agents or in combination with other drugs, can now feel reassured that they will continue to have a key role in hypertension management. At the same time, it is important to take a closer look at the ALLHAT data and evaluate the validity of the claims and conclusions published in the formal report,1 not to mention the accompanying press releases.
THE MYSTERY OF THE PRIMARY END POINT
From the beginning, the chief focus of this study was intended to be coronary events. So much so, that the “HAT” in the title, “ALLHAT,” stands for Heart Attack Trial. The results of the study showed that for the formal primary end point of fatal coronary heart disease and nonfatal myocardial infarction, there were no meaningful differences among the drugs. The study event rate for chlorthalidone was 11.5%, with fractionally lower point estimates for amlodipine (11.3%) and lisinopril (11.4%) despite the fact that systolic blood pressure was not as well controlled in the latter two groups as in the diuretic group.
In an almost unprecedented departure from scientific probity, the authors of the ALLHAT report omitted this apparently inconvenient fact from their Conclusion. It is noteworthy that even the New York Times, which in its initial front page coverage of ALLHAT proclaimed the superiority of the diuretic, felt compelled to publish a formal Correction four days later acknowledging that the primary end point of the study was, in fact, similar among the three drugs.
The main basis for the claim that chlorthalidone was better than the other two drugs depended on secondary end points. Most notably, the stroke rate with the diuretic was claimed to be lower than with lisinopril, and heart failure was claimed to be reduced when compared with both lisinopril and amlodipine. Close scrutiny of the data supporting these claims has raised some doubts, and it is revealing to explore them further. However, before doing so, it is important to look at how study design issues affected patient management and blood pressure control.
STUDY DESIGN AND BLOOD PRESSURE CONTROL
Since the goal of ALLHAT was to compare the effects of different antihypertensive agents on clinical end points, it was critical to achieve equal blood pressure effects in each of the three treatment groups so as to ensure that the outcomes benefits of the three drug classes could be validly compared. This intention was reinforced quite clearly in a publication by the ALLHAT study leaders about 1 year before ALLHAT's results were announced.2 So, the difference in systolic blood pressure of 2 mm Hg (actually, slightly higher when weighted for patient-years of treatment exposure) favoring chlorthalidone over lisinopril was not trivial, but instead has clouded interpretation of many of ALLHAT's clinical outcomes. A recent comprehensive study of the relationships between blood pressure and clinical events, based on observations in one million persons, indicated that blood pressure differences closely similar to those observed in ALLHAT could account for powerful effects on stroke and coronary mortality rates.3
The reason for the blood pressure problem in ALLHAT is easy to explain. Because the study was originally set up to compare the outcomes effects of a diuretic, a CCB, an ACE inhibitor and an α blocker (which was discontinued during the trial and is not discussed further here), the research protocol prohibited the use of agents from these classes when additive treatments were required in patients whose blood pressures did not respond adequately to their primary drug. Beta blockers, which were the agents most commonly chosen for this purpose, as well as other agents affecting adrenergic mechanisms, were the main drugs that could be added. This situation clearly helped chlorthalidone, for addition of a β blocker to a diuretic provides a logical and effective blood pressure-lowering combination. Even for those patients assigned to amlodipine, the addition of a β blocker is useful. But, for the lisinopril group, adding a β blocker is clearly less helpful than a lowdose diuretic or a CCB. For black patients, in whom neither ACE inhibitors nor β blockers are drugs of choice for blood pressure control,4 this caused an even greater shortfall in blood pressure control.
This discrepancy cannot be shrugged off as an accidental or unintended result of the study. It was predictable from the time the study organizers decided to attempt multiple comparisons in a single trial rather than the more direct and clinically relevant approach of separate head-to-head comparisons. This decision may have been largely driven by considerations of cost and logistics, but obviously it also contrived to prevent equal blood pressure effects in the treatment groups, in essence benefiting chlorthalidone but putting lisinopril at a disadvantage.
STROKES AND THE AFRICAN AMERICAN PATIENTS
Stroke is perhaps the most feared outcome of hypertension. Compared with the diuretic, the stroke event rate during treatment with amlodipine was actually 7% lower. Although this advantage to the CCB was not statistically significant, it was observed across almost all subgroups in the diverse population included in ALLHAT.
On the other hand, chlorthalidone reduced stroke event rate by 15% when compared with the ACE inhibitor lisinopril. But herein lie some of the most interesting and contentious issues in ALLHAT. For a start, stroke event rates were virtually identical for the diuretic and the ACE inhibitor in the nonblack patients, meaning that the entire overall difference between chlorthalidone and lisinopril could be accounted for by the dramatic 40% greater event rate in black patients randomized to lisinopril.
There are two explanations for this result in the African American patients. First, due to the treatment selection problems discussed earlier, there was a discrepancy of 4 mm Hg in systolic blood pressure favoring chlorthalidone. This difference by itself could explain much or even all the stroke excess in these high risk patients.3 And, second, there is the fact that the majority of those patients who required additional therapy finished with a combination of an ACE inhibitor and a β blocker, thus getting two drugs with overlapping neurohormonal actions that might not provide the same additive target organ benefits that might be expected when drugs with complementary properties are combined.
THE HEART FAILURE CONUNDRUM
The overall claim for chlorthalidone's superiority over the other two drugs in preventing clinical end points in ALLHAT was based predominantly on the difficult-to-diagnose secondary end point of heart failure. To cardiovascular experts, this finding came as a surprise, particularly the claimed superiority over the ACE inhibitor lisinopril which, based on copious clinical trial evidence as well as its known actions in the circulation, is regarded as the primary treatment for heart failure and would be expected to be highly effective in preventing this condition. Some important questions should be addressed.
Is the Heart Failure Claim Credible?
ALLHAT is not the first clinical trial to compare the effects on heart failure of diuretic-based treatment with treatments based on ACE inhibitors or CCBs in hypertensive patients. A meta-analysis carried out by the Blood Pressure Lowering Treatment Trialists' Collaboration,5 which had the strong authority of being based on prospectively designated clinical trials, compared heart failure rates in patients treated with ACE inhibitors with those treated with conventional therapies (diuretics and β blockers). With a total of more than 8000 patients in each group, the investigators found an 8% lower event rate with the ACE inhibitor-based treatment. Although this difference was not significant, it goes clearly in the opposite direction to that reported in ALLHAT. In a similar comparison between CCBs and conventional therapy, with over 11,000 patients in each group, the heart failure event rate this time was found to be lower in the diuretic-based group by 12%, though it did not reach statistical significance. Moreover, while favoring the diuretic-based treatment, the point estimate of this difference was substantially lower than that reported in ALLHAT. These unexpected differences in event rates between the previous studies and ALLHAT could possibly be explained by the blood pressure problems in ALLHAT, but it is also worth considering the accuracy of the diagnosis.
Were the Heart Failure Findings Real?
Heart failure is a difficult diagnosis to make, even by experienced cardiologists participating in formal heart failure trials, let alone a study like ALLHAT that was conducted to a large extent in community-based settings. A potentially important problem with the diagnosis of heart failure in ALLHAT was related to the masking effect of diuretic treatment on major fluid-dependent clinical signs such as rales and peripheral edema and symptoms like dyspnea. For this reason, heart failure in hypertensive patients receiving diuretics can go unrecognized for a considerable period. In ALLHAT, the majority of patients entering the study were already receiving a diuretic, so that those individuals with unsuspected heart failure randomized to the ACE inhibitor or CCB would have been at risk of rapidly losing the masking effects of their previous diuretics and manifesting their heart failure early in the trial. On the other hand, in those patients with hidden heart failure who were randomized to the powerful diuretic chlorthalidone, fluid-dependent clinical signs and symptoms might have remained suppressed. In fact, examination of the Kaplan-Meier curves for heart failure event rates with chlorthalidone and lisinopril shows that much of the separation between their effects takes place during the early stages of the study.
The situation comparing chlorthalidone with amlodipine is not so clear, but again the masking effect of the diuretic might have played a substantial role in the different event rates between the two treatments. Moreover, one of the common side effects of the CCB is peripheral edema, which is not related to fluid retention but which can misleadingly suggest the appearance of heart failure. This, again, could have added to the possibility of misdiagnosis in ALLHAT, where peripheral edema was regarded as a key physical finding of heart failure and where rigorous confirmation of clinical events was carried out in only a small sampling of patients.
One final point should be noted. Heart failure is a condition with a high case-fatality rate, and the substantial benefits claimed for chlorthalidone in preventing heart failure (not to mention some of the other cardiovascular end points) should have resulted in a clear trend toward lower mortality in patients treated with the diuretic. As discussed later, this was definitely not the case. In view of these questions and uncertainties, heart failure seems to be a rather soft and uncertain secondary end point upon which to base the major justification for chlorthalidone's overall superiority claim in ALLHAT.
DIABETES: GOOD NEWS AND BAD NEWS
One of the interesting outcomes of ALLHAT was that in the comparisons of end points between the diuretic and the other agents there were no major differences in event rates between diabetic and nondiabetic patients. Because of data showing the specific benefits of drugs that interrupt the renin-angiotensin system in patients with diabetic nephropathy,6,7 experts had started to recommend that such drugs as ACE inhibitors should be used in all diabetic patients. In ALLHAT, however, there did not appear to be much of an advantage to lisinopril over chlorthalidone in diabetic (as compared with nondiabetic) patients, suggesting that a thiazide agent, as monotherapy, could be as acceptable as ACE inhibitor monotherapy in these high-risk patients. However, in the absence of detailed renal data in ALLHAT we cannot be certain that this applies to kidney protection.
It is possible, though, that the blood pressure difference between lisinopril and chlorthalidone in ALLHAT could have influenced cardiovascular outcomes in chlorthalidone's favor, particularly as these outcomes are so blood pressure-sensitive in diabetic patients.8 We should also not forget that strong cardiovascular and renal benefits in diabetic patients have occurred when blockers of the reninangiotensin system are combined with diuretics. For these reasons, despite some reassurance from ALLHAT, physicians should generally continue to employ this combination approach for their diabetic patients with hypertension.
As would be expected, blood glucose concentrations rose more with the diuretic than with either of the other two agents. For nondiabetic patients entering the study, the 4-year incidence of new-onset diabetes in the chlorthalidone group was 11.6%, which represented an 18% increase in relative risk compared with amlodipine (4-year rate of 9.8%) and a 43% increase in relative risk compared with lisinopril (4-year incidence of 8.1%). Both of these differences were significant. This increase in new-onset diabetes in the diuretic group did not translate into increased cardiovascular events during the relatively short period of observation following diagnosis, but this finding should prompt the recommendation that patients at risk of developing diabetes should not be treated with a thiazide alone but rather with a regimen based on a drug that interrupts the renin-angiotensin system.
DEATH: THE INCONTROVERTIBLE END POINT
Mortality is the most definite as well as the most important of the secondary end points in ALLHAT, yet it received surprisingly little emphasis in the ALLHAT report and press releases. Compared with chlorthalidone, all-cause mortality was identical with lisinopril and 4% lower with amlodipine (nonsignificant). For nonblack patients, mortality was 3% and 6% lower with lisinopril and amlodipine. Mortality differences favoring the CCB and especially the ACE inhibitor very likely would have been greater had it not been for the blood pressure discrepancies in ALLHAT.3
This is a pivotal issue that speaks directly to the principal conclusion stated in the ALLHAT report. It seems inconsistent, perhaps almost absurd, to have claimed cardiovascular outcomes superiority for chlorthalidone—which can only have real meaning if the drug displays life-saving attributes—when its effects on mortality were actually heading in the wrong direction.
IMPLICATIONS OF ALLHAT
The ALLHAT results publication1 is not only one of the longest study reports ever published, it is also detailed and complex. It will take more time to fully define the true meaning of its data. But since the study, in its final form, compared the effects of three antihypertensive agents, it may be helpful to look briefly at the impact of ALLHAT on each of them.
THE INDIVIDUAL AGENTS
According to the ALLHAT authors, this study demonstrated that a thiazide diuretic was superior to an ACE inhibitor or a CCB in cardiovascular protection. For the reasons already detailed, this conclusion is debatable at best. Even so, nothing in ALLHAT has hurt chlorthalidone's role as an integral part of hypertension therapy and as an agent strongly to be considered as first-step therapy in the elderly and particularly in black patients. The ALLHAT experience has also shown that concerns about the use of thiazide diuretics in diabetic patients may be largely unfounded.
Lisinopril fully equaled chlorthalidone for the primary end point of coronary events and for the key secondary end point of mortality, despite the deficiencies of study design that put the ACE inhibitor at a blood pressure disadvantage. Black patients did not seem to do as well with lisinopril as nonblack patients, although it is likely that the use of these agents, when combined appropriately with such drug classes as diuretics or CCBs, could also be of considerable value in preventing renal and cardiovascular events in African Americans. Apart from the controversial heart failure findings, advocates of ACE inhibitors could still argue that ALLHAT was not able to disprove the notion that ACE inhibitors, used in an optimal fashion, might be the first-line therapy of choice for many hypertensive patients, particularly nonblacks.
Amlodipine performed well in ALLHAT. While the heart failure findings need further clarification, it was notable that amlodipine not only equaled chlorthalidone for the primary coronary end point, but actually appeared to have a small advantage (albeit not significant) for both mortality and stroke prevention.
Impact On Guidelines
Although not presented in the ALLHAT report, the relative cost of drugs somehow became part of its published conclusion. No doubt, thiazide diuretics are relatively inexpensive, at least as far as the cost of acquisition is concerned, but so are other antihypertensive agents. Many of the ACE inhibitors, including lisinopril, as well as some of the long-acting dihydropyridine CCBs, are also generic and well priced. Because hypertension is such an aggressive market place, several of the newer branded drugs are priced competitively and allow physicians reasonable flexibility in choosing treatment based on therapeutic need. Most importantly, we all now recognize that effective blood pressure control in most hypertensive patients calls for logical drug combinations that typically will include all the drug types examined in ALLHAT as well as other classes.
The National Heart, Lung, and Blood Institute (NHLBI) was responsible for organizing and conducting the ALLHAT study and for writing its report and conclusions. Since the NHLBI also organizes a hypertension guidelines committee (the forthcoming Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure [JNC VII]) and appoints its members, there is little doubt that the agenda expressed in the ALLHAT report and the NHLBIs press releases will affect the recommendations. Experts in hypertension concerned about providing the best possible treatment for the diverse hypertension population in the United States will hope fervently that the Committee will thoughtfully take into account the full array of available clinical trials data, including responsible interpretation of ALLHAT, when writing their report.