When You SPRINT, It's Good to Know the Goal as Well as the Goal Line

For many patients and their physicians, the Systolic Blood Pressure Intervention Trial (SPRINT) changed the goal line: It provided impressive evidence that in the population studied, intensive blood pressure management (IBPM)—lowering systolic blood pressure to ≤120 mm Hg—reduced the risk of cardiovascular events and mortality.¹ But applying these important findings to patients with CKD has not been straightforward.

In this issue of Journal of the American Society of Nephrology (JASN), Elaine Ku and colleagues report a valuable, pooled secondary analysis of seven randomized clinical trials (RCTs) of IBPM in CKD.² The authors obtained patient-level data either from data repositories or directly from the original authors. The seven analyzed trials they analyzed all tested the impact of IBPM on kidney outcomes and death, and all reported at least 15 kidney failure events, reflecting a high-risk population. The pooled analysis includes almost 6000 patients, all with an eGFR <60 ml/min per 1.73 m², of whom SPRINT provided almost half, but only 3% of kidney outcome events.

The primary question Dr. Ku and coauthors address, whether IBPM slows the progression of CKD and reduces the need for kidney replacement therapy (KFRT), is an important and by no means new question for the nephrology community. Almost 30 years ago, two major clinical trials—the African American Study of Kidney Disease (AASK) and the Modification of Diet in Renal Disease study (MDRD) trials^3,4—were undertaken to address it. AASK contributes 16.5% of the patients, and 33% of the kidney outcome events of the patients analyzed in the current report; MDRD contributes 14% of the patients and 30% of the events. Neither AASK nor MDRD investigators found statistically significant effects of intensive BP control on their primary renal outcomes, defined as a reduction in kidney function or need for KFRT.^3,4

Currently, approximately 25 years later and in light of five other trials increasing by half the number of kidney outcome events, Dr. Ku and colleagues ask whether a negative conclusion remains warranted. In brief, in both their unadjusted primary and baseline-adjusted analyses, the risks of both the kidney outcome and death were approximately 13% lower with intensive BP control, but these differences did not achieve statistical significance. Many sensitivity analyses produced similar results for the overall IBPM effect. In broad strokes, the findings of AASK and MDRD still held.

However, the authors prespecified tests of the interactions of intensive blood pressure control with baseline eGFR and proteinuria, and the results of these analyses are intriguing. In both primary analysis and sensitivity analysis excluding the single pediatric trial, the estimated IBPM effect on the rate of kidney disease progression varied statistically significantly with baseline eGFR. In secondary analyses adjusted for baseline covariates including eGFR, this interaction remained significant in the adult though not the full collection of trials. When the nature of this interaction, a.k.a. effect modification, was explored in conventional CKD stages, IBPM was associated with a statistically significant 20% reduced rate of kidney disease progression in the stages 4–5 CKD subgroup—that is with eGFR <45 ml/min per 1.73 m²—but not with a meaningful difference in the CKD stage 3 subgroup. Ku and colleagues recommend appropriately powered trials in patients with high-risk advanced CKD to potentially confirm the suggested benefit for patients with CKD stages 4–5.

Where does this leave us? The perils of jumping to conclusions about subgroups are well established and not inherently diminished in composite datasets drawn from several RCTs. These perils can be controlled by prespecification of the scope and strategy of statistical testing, including the use of hierarchical testing sequences and/or tests that limit overall or familywise false positive or false discovery rates. However, such false positive error control lowers power to detect true subgroup effects. Hence, many investigators prefer a more liberal tradeoff in exploratory subgroup analyses. Ku and colleagues follow a common strategy for individual RCTs, by limiting the scope of their subgroup analyses to those defined by eGFR and proteinuria, clearly delineating their statistical tests as primary, secondary, or sensitivity analyses, but omitting formal multiple comparison methods to control an overall false positive rate. Their testing strategy thus has a familywise false positive error rate closer to 3×5=15% than to the conventional 5% statistical significance threshold, leaving chance as a plausible competing explanation of the eGFR interaction and apparent CKD stages 4–5 benefit.

CKD stages are ultimately arbitrary quantitative categories. The simplest explanation consistent with the data is an IBPM effect on kidney outcomes that is increasingly beneficial with diminishing eGFR. But with no established explanatory mechanism for this and in view of the above, neither absence of any effect nor a uniform benefit independent of eGFR (such as the overall roughly 13% observed reduction in adverse kidney outcomes) can be ruled out. Bias in the subgroup analysis of kidney outcomes is also possible, if risks of death and kidney outcomes conditional on eGFR and other controlled risk factors are positively associated. While the pooled analysis refines our estimates of IBPM's effects, more data are needed to resolve the decades old conundrum.

Reviewers of this manuscript and the JASN editors agree that this work should not be framed as advice for clinical management. Reducing cardiovascular risk provides a well-justified rationale for using the SPRINT targets in patients with CKD, a conclusion supported by current guidelines.⁵

Nevertheless, practical questions remain. As all clinicians know and as guidelines emphasize, therapeutic goals for every patient need to balance patient characteristics, patient preferences, and an assessment of individual risk and benefit. To provide wise counsel, it helps to know the goal as well as the goal line. This study again reminds us that we lack definitive trial-grade evidence to guide hypertension management for patients with CKD, particularly the important group of patients with CKD stages 4 and 5, a group still frequently excluded from trials.^6,7 For many patients with CKD stages 4 and 5, slowing progression and reducing the need for KFRT remain important therapeutic goals. The work by Ku and colleagues tells us that we do not know the impact of IBPM on these goals but that there is a good reason to keep looking, particularly to see if there is possible benefit for the most vulnerable patients.

And a final editorial note. This study reinforces the value of second looks at the data for major clinical trials. The primary report of a large RCT, no matter how carefully analyzed and presented, rarely tells us all we can learn from these important experiments. JASN editors require that all authors submitting a trial for possible publication in JASN provide a data sharing plan. We look at these plans with flexibility, recognizing the need for protections dictated by proprietary or regulatory concerns and the needs to protect patient privacy, but we require a plan.⁸

Often, as in this case in the study reported here, secondary analyses clarify what we do not know and where we should go next. The analysis by Ku and colleagues provides useful insights on work needed to optimize blood pressure management for our most vulnerable patients.

Disclosures

J.P. Briggs reports Advisory or Leadership Role: peer review editor for PCORI (paid honorarium) and JASN EIC. P.B. Imrey reports Consultancy: Colgate Palmolive.

Funding

None.

Author Contributions

J.P. Briggs and P.B. Imrey conceptualized the study and wrote the original draft.