Improving Patient Selection in Clinical Trials

The established clinical end point in clinical trials in CKD is a composite of kidney failure (defined as an eGFR ≤15 ml/min per 1.73 m², chronic dialysis or kidney transplantation, or a 57% eGFR decline). This end point is a late manifestation of CKD. Thus, trials typically enroll participants with severe albuminuria and low eGFR for sufficient end points to occur to robustly assess the efficacy of the intervention. However, recent CKD trials have reported low event rates, with the proportion of patients experiencing a primary kidney outcome ranging between 8% and 11% in the placebo groups.^1–4 Moreover, in the Study Of diabetic Nephropathy with AtRasentan trial, despite recruiting participants with severe albuminuria and low eGFR, nearly 60% showed a stable eGFR during the trial.⁵ These findings highlight the need for alternative strategies to enrich trial populations with patients genuinely at risk of disease progression. One approach is to use historical eGFR data to identify participants with steep declines in eGFR before enrollment in a clinical trial (pretrial eGFR), providing a direct assessment of progression risk. This strategy is examined by Keum et al. in this CJASN issue and is worthy of particular attention.⁶

The study by Keum et al. evaluated whether the pretrial eGFR decline can effectively identify patients at high risk of kidney failure. They analyzed data from the Preventing Early Renal Loss in Diabetes (PERL) trial, which enrolled 530 individuals with type 1 diabetes and CKD. Participants could be enrolled on the basis of either increased albuminuria or a history of rapid eGFR decline. Increased albuminuria was defined as urinary albumin–creatinine ratio between 30 and 5000 mg/g or between 18 and 5000 mg/g if participants were not using renin-angiotensin system inhibitors. In addition, participants who did not qualify on the basis of the albuminuria thresholds could still be enrolled if the eGFR decline in the 3–5 years before the trial was ≥3 ml/min per 1.73 m² per year. The trial recruited 394 participants on the basis of the albuminuria criterion and 124 on the basis of the eGFR decline criterion. The two different trial selection criteria enabled the researchers to compare the efficiency of these approaches to identify participants at highest risk of eGFR decline during the trial. Their results showed that participants with a history of albuminuria experienced a −3.6 ml/min per 1.73 m² per year (95% confidence interval, −3.2 to −4.0) eGFR decline during the trial for participants enrolled on the basis of the albuminuria criterion compared with −2.4 ml/min per 1.73 m² per year (95% confidence interval, −1.9 to −2.8) for those enrolled on the basis of the pretrial eGFR criterion (P = 0.001). These findings imply that while pretrial eGFR decline correlates with increased risk of kidney failure, increased albuminuria seems to be more efficient to identify participants at high risk of kidney function decline.

One would expect pretrial eGFR decline to identify patients with fast-progressing CKD, but the PERL trial showed a less steep eGFR decline in those selected on the basis of historical data compared with those selected on the basis of albuminuria. Intriguingly, the eGFR decline in the placebo group was also less steep during the intervention period compared with the preintervention period. How could this be? The pretrial eGFR decline may have been imprecise because of a short preintervention period to collect eGFR data or because of too few measurements. In the PERL study, the pretrial eGFR decline was estimated using a median of seven central laboratory measurements obtained over a median period of 3.3 years, an interval that should theoretically provide sufficient accuracy and precision to estimate the eGFR slope. The less pronounced eGFR decline observed during the trial might be attributed to close monitoring of patients and improved care resulting from trial participation. It may also simply reflect regression to the mean, which is a tendency of extreme values to move closer to the average upon subsequent measurement, and thus extreme eGFR declines during the pretrial period might have affected the subsequent eGFR slope estimate during the trial. Despite excluding eGFR data during hospitalizations or known AKI, undetected AKI episodes or medication effects may have skewed the eGFR slope.

In addition to using the pretrial eGFR slope as a risk enrichment criterion, comparison of the pretrial eGFR slope with the eGFR slope during the trial can also provide insight into the kidney protective efficacy of an intervention. Unfortunately, allopurinol did not reduce eGFR decline compared with placebo in the PERL study, preventing assessment of whether it altered the pretrial GFR slope. This issue was recently assessed by Miyamoto et al. in the CANPIONE trial, an open-label trial of the sodium glucose cotransporter 2 inhibitor canagliflozin. Their trial enrolled patients with type 2 diabetes and microalbuminuria and included a preintervention period of 24 weeks during which eGFR was centrally measured, supplemented by retrospective eGFR data spanning 3 years before the study. The results demonstrated a large between-patient variability in the pretrial eGFR slope. Nevertheless, 52 weeks of treatment with canagliflozin compared with care as usual reduced the participants' individual pretrial eGFR decline. This benefit was detected in a small 96-participant trial at early CKD stages. Importantly, post hoc analyses suggested that the benefit of canagliflozin may be particularly present in patients with a rapid deterioration of kidney function before trial enrollment: those with a pretrial eGFR decline >1 ml/min per 1.73 m² per year.⁷ These data suggest that the use of a pretrial eGFR slope may be an efficient approach to detect treatment effects in participants with early stages of CKD. However, similar to the PERL study, a large variability in eGFR slope was observed, particularly when the pretrial slope was determined over a short period of 6 months.⁷

The findings of the PERL and CANPIONE trials underscore the challenge of variability and uncertainty in eGFR assessments. Sufficient precision in pretrial eGFR decline is vital if it is to be used as a risk enrichment criterion. This challenge was similarly evident in an analysis of pretrial eGFR data from the Dapagliflozin and Prevention of Adverse Outcomes in CKD trial.¹ Using medical history data collected up to 3 years before inclusion, the pretrial eGFR slope for patients in the dapagliflozin group of the trial showed a much larger variation (SEM, 0.4) compared with the eGFR slope during the trial when eGFR was measured under standardized conditions in a central laboratory (SEM 0.1; Figure 1).

Figure 1.

eGFR trajectory in 412 participants assigned to dapagliflozin in the DAPA-CKD trial. eGFR data from medical electronic records and measured in local laboratories were collected to calculate the pretrial eGFR slope. eGFR data during the trial were collected at four monthly intervals and measured in a central laboratory. The CKD-EPI 2009 equation was used to estimate GFR. The mean eGFR slope attenuated from pre- to postrandomization to dapagliflozin. The SEM of the eGFR slope was also lower during the trial compared with the pretrial period. CKD-EPI, CKD Epidemiology Collaboration; DAPA-CKD, Dapagliflozin and Prevention of Adverse Outcomes in CKD.

Given these challenges, the key question is whether the pretrial eGFR slope should be considered as a risk enrichment criterion for clinical trials. To use pretrial eGFR slope effectively, we must determine the minimum number of observations needed and assess whether historical data suffices. Key considerations include the time frame for eGFR collection, inclusion of old eGFR measurements obtained years before inclusion in a clinical trial, and the interval between eGFR measurements to assess the decline. In addition, understanding the impact of concomitant medications and capturing comorbidities such as AKI episodes is crucial. An alternative to using the historical data could be a run-in period where patients measure eGFR under standardized conditions, possibly using innovative at-home technologies. This strategy could collect high-resolution data over a short period (up to 6 months) to derive an accurate GFR estimate. Before adopting these approaches, rigorous evaluation of their validity and reliability is needed. If effective, at-home eGFR measurements could reduce logistical burdens and costs while standardizing data collection.

In conclusion, while traditional inclusion strategies using albuminuria and eGFR have limitations, emerging approaches show promise. The PERL study provides important insights into this issue. Although using pretrial eGFR slope as a risk marker is appealing, collecting pretrial eGFR data is more complex than measuring albuminuria at a single visit. Nevertheless, improving pretrial eGFR estimates using more frequent or remote data collection could advance patient selection strategies for future trials.

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/CJN/C107.

Funding

None.

Author Contributions

Conceptualization: Hiddo J.L. Heerspink, Niels Jongs.

Writing – original draft: Niels Jongs.

Writing – review & editing: Hiddo J.L. Heerspink.