Response to Letter to the Editor concerning manuscript, “Blood pressure variability predicts adverse events and cardiovascular outcomes in sprint”

Dear Editor,

We welcome the comments by Chang et al, and provide this response to the issues raised by them in comparing our study to theirs.1, 2

Firstly, Chang et al observe that we only reported on diastolic BPV and neglected systolic BPV. In fact, we reported on systolic BPV and pulse pressure (PP) BPV. Systolic BPV and PP BPV did not vary between the intensive treatment arm and the standard treatment arm; but diastolic BPV did differ between the groups; hence, we explored diastolic BPV in more detail. One might hypothesize that each investigator's focus on achieving a specific target systolic blood pressure (BP) in each group could suppress the ability of the SPRINT study to reveal an effect of systolic BP variability. In contradistinction, diastolic BP had no specific target. Our analysis showed a strong effect of diastolic BP coefficient of variation (DBP‐CV) on the primary composite end point as well as on common adverse effects. Chang et al did not look for a DBP‐CV effect in their study, so their results neither agree nor disagree in this regard.

Secondly, Chang et al compared their method of using a specifically selected number of measurements during a period they assumed would be relatively free of BP medication titration, ie, at 3‐, 6‐, 9‐, and 12‐month visits, to our method of determining for each patient the landmark date at which he/she achieved the target systolic BP, and then including BP measurements at all subsequent visits in the analysis. We reasoned that including any BP reading prior to achieving the target systolic BP inappropriately inserts variability that arises from the intent to decrease blood pressure further. The methods of Chang et al, which use measurements at months 3, 6, 9, and 12, insert more of that variability in the intensive group than the comparator group, potentially biasing results because at month 3, 28.7% of patients in the intensive group (target systolic BP 120 mm Hg) had not yet achieved target, compared to only 4.1% of those in the standard group (target systolic BP 140 mm Hg). Our analysis included no BP readings prior to achievement of target systolic BP. Figure 1 shows the frequency distribution of the number of BP measurements included in our analysis. Overall, the analysis performed by Chang et al excluded 16% of all SPRINT participants and 42% of end points; our analysis excluded only 5% of participants and only 8% of end points. Chang et al also commented that our regression model did not include the number of BP measurement visits as a covariate. We re‐ran the regression with that parameter included (Table 1) and obtained similar results.

Figure 1.

Frequency distribution of the number of blood pressure observations included in analysis

Table 1.

Hazard ratios and significance levels for predictor variables of interest in the original analysis and the additional analyses

	Originala	Add age, sex, race	Add 7 covariatesb	Add 8 covariatesc
DBP‐CV	1.14 P  < 0.0001	1.13 P < 0.0001	1.12 P < 0.001	1.10 P < 0.0001
Intensive	0.55 P = 0.0022	0.68 P < 0.0001	0.68 P < 0.0001	0.61 P < 0.0001

^aMezue et al AJH 2018.

^bAdd covariates to original model, which already contains randomized treatment group and measures of BP variability: age, sex, race, baseline chronic kidney disease, baseline coronary disease, baseline smoking status, and average systolic blood pressure.

^cThe 8th covariate added to the model is the number of blood pressure measurement visits.

Thirdly, Chang et al observed that we did not include known confounders of BPV in our regression model. We re‐ran the regression, including age, sex, race, baseline coronary disease, baseline chronic kidney disease, baseline smoking status, and average systolic BP (in addition to randomized treatment group and measures of BPV in the original model) as covariates (Table 1) and obtained similar results.

The fundamental difference between our approach and that of Chang et al is that we focused on diastolic BPV where we did find a clinically relevant result, while they did not investigate the potential effect of diastolic BPV.

CONFLICT OF INTEREST

All authors have no conflicts of interest to declare.