GFR slope surrogate analysis corrected after multi-arm trial misclassification

Nature Medicine has published an addendum to the 2023 meta-analysis of GFR slope as a surrogate endpoint for kidney failure, correcting how shared control groups were handled in one of the included multi-arm trials. The correction applies to the African American Study of Kidney Disease and Hypertension, where the original publication had incorrectly stipulated the shared control groups in the AASK trial. After the dataset was re-run with the correct designation, the summary estimates changed only in the hundredths place, leaving the overall message of the analysis intact.

The adjustment may look minor on paper. In surrogate endpoint work, however, even small bookkeeping errors in trial structure can ripple through a Bayesian meta-regression because correlated sampling errors are part of the model itself. Multi-arm studies are especially sensitive to this issue: when a single control group is reused across several comparisons, the statistical dependency must be specified correctly or the precision of the pooled estimate can shift. In a field where kidney failure progression is often slow and event counts are limited, exact trial accounting is not cosmetic; it is the foundation of the inference.

What changed in the trial-level model

The original article by Inker and colleagues examined the association between treatment effects on GFR slope and treatment effects on clinical endpoints across randomized controlled trials. In the addendum, the investigators repeated the trial-level Bayesian meta-regression using the proper designation of the multi-arm trials. The resulting updated summaries differed only slightly from the published values, with changes confined to the hundredths place. The addendum also points readers to the updated trial-level association summaries in Figure 1 and Table 1.

One sentence captures the practical consequence: the correction altered the arithmetic, not the direction of the signal.

Across 66 studies, the relationship between treatment effects on GFR slope and clinical kidney endpoints remained strong. The figure shows that greater treatment effects on GFR slope correspond to lower hazard ratios for the clinical endpoint, with the association tighter for the 3-year total slope than for the chronic slope. The reported meta-regression slopes were minus 0.35 and minus 0.33, and the R squared values were 0.95 for the total GFR slope and 0.55 for the chronic GFR slope. Those values matter because they suggest that the earlier part of the slope signal may track clinical benefit more cleanly than the chronic phase alone, where slower drift and more background noise can dilute the relationship.

Why the surrogate endpoint question still matters in CKD

The appeal of GFR slope is straightforward: kidney failure develops over years, while changes in filtration rate appear earlier and can be measured repeatedly. In chronic kidney disease, waiting for hard renal outcomes can make trials long, expensive, and underpowered, especially when modern background therapy reduces event rates. A surrogate endpoint can shorten development timelines, but only if the surrogate reliably mirrors the endpoint clinicians care about. That is why the original meta-analysis drew attention in the first place, and why the addendum, despite its narrow correction, still deserves notice.

At the bedside, this statistical issue has a direct clinical translation. If GFR slope can stand in for kidney failure with enough fidelity, investigators may be able to evaluate renoprotective therapies sooner and with fewer patients. If it cannot, early declines in slope could mislead drug development by exaggerating benefits that never translate into fewer dialysis starts, transplants, or severe loss of kidney function. The present correction does not reopen that debate; it simply strengthens confidence that the pooled model was rebuilt with the proper trial structure.

A correction that strengthens confidence in the original framework

The addendum also reminds readers how fragile confidence can be when a surrogate endpoint depends on cross-trial synthesis. The AASK trial has long held an important place in nephrology because it examined blood pressure lowering and antihypertensive drug class in hypertensive kidney disease, and its data were among the studies folded into the broader evidence base. When a study like that is embedded in a multi-arm structure, the handling of the shared comparator affects the covariance matrix, the weighting of comparisons, and the final summary estimates. Correcting that structure is the right statistical move, even when the numerical shift is tiny.

For clinicians and trialists, the message is not that the prior conclusion was wrong. It is that surrogate validation in nephrology depends on details that are easy to overlook and hard to recover once published. A small indexing error in the trial architecture can change the decimal point while leaving the clinical interpretation stable. In this case, the updated analysis still supports the view that GFR slope tracks renal outcomes closely enough to remain a credible surrogate candidate, particularly when the total slope over three years is used.

Published evidence and the record now

The addendum was published on 16 November 2026 and refers back to the original Nature Medicine article published online on 17 June 2023. The corrected paper continues to anchor the field’s discussion of surrogate endpoints in CKD, and the accompanying references include both the 2023 Nature Medicine article and the 2002 JAMA AASK trial report. No new clinical trial was introduced here, and no new patient cohort was added; this was a methodological correction to the existing evidence synthesis.

For nephrology researchers, the most important point may be the quietest one: a corrected model can preserve a result while making it more defensible. In surrogate endpoint science, that is not a trivial edit. It is the difference between a plausible shortcut and a statistically unstable promise.


Reporting by Dr. A. Nikhil Krishnan, MBBS for Medicine World Foundation, with additional data provided by Nature Medicine.
PRIMARY CLINICAL EVIDENCE & REGISTRY DATA