Blood pressure variability causes spurious identification of hypertension in clinical studies: a computer simulation study.

BACKGROUND

The blood pressure (BP) of an individual varies considerably from day to day. Hypertension is commonly identified based on the average of two BPs taken at each of two visits, a practice consistent with current guidelines. We hypothesized that (i) in the setting of high-normal BP ("prehypertension"), this practice results in frequent spurious detection of hypertension, and (ii) that random, spurious detection of hypertension and flawed study design together explain why in the Trial of Preventing Hypertension (TROPHY) study candesartan appeared to suppress the development of hypertension for 2 years after cessation of therapy.

METHODS

We used Monte Carlo simulation to quantify spurious detection of hypertension at repeated clinic visits in one million subjects with unchanging usual systolic BPs (SBPs) between 130 and 139 mm Hg and normal BP variability. Criteria for identifying hypertension derived from Rosner and Polk, Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC 7), and the Trials of Hypertension Prevention (TOHP), TROPHY, and Framingham studies were applied to Canadian Heart Health Survey data. On the basis of these simulations, we created a three-parameter model of Kaplan-Meier curves recorded in the TROPHY study. We used the model to analyze the rates at which hypertension was identified in the TROPHY study.

RESULTS

TROPHY criteria spuriously identify hypertension in 76% of subjects over 18 clinic visits. Higher rates of spurious detection of hypertension are seen with JNC 7, TOHP and Framingham study criteria. Even highly conservative criteria falsely identify hypertension in 27% of subjects over 18 visits. Our three-parameter model suggests no postdiscontinuation benefit of candesartan.

CONCLUSIONS

When applied over multiple visits, current criteria for detecting hypertension are intolerant of normal BP variation. The use of conservative criteria would reduce spurious identification of hypertension. The apparent long-term beneficial effect of candesartan seen in the TROPHY study can be explained by an inadequate method for detecting hypertension, and flawed study design.