Introduction
Chronic kidney disease (CKD) patients have high risks of death, cardiovascular disease (CVD), and end-stage renal disease (ESRD), which increase with the progression of CKD. Various endpoints have been used for clinical studies. The occurrences of death, CVD, and ESRD are important clinical endpoints for CKD patients, which are true endpoints, and are often used as the endpoints in clinical studies. However, because the events associated with these true endpoints are of limited frequency, the sample sizes based on the endpoints are large to evaluate the effects of a therapy on the endpoints, the study period are long, and a large budget is required. Thus, surrogate endpoints are used instead of these true endpoints for evaluating the effects of a therapy. The Biomarkers Definitions Working Group defined a surrogate endpoint as a parameter expected to predict the clinical benefits of a therapy on the basis of epidemiologic, therapeutic, pathophysiologic, or other scientific evidence [
1]. The use of appropriate surrogate endpoint, which reflects kidney function and can quantify the risk of the true endpoints, may make the sample size of clinical studies smaller and the study period shorter.
The US National Kidney Foundation and the US Food and Drug Administration reported eGFR change as surrogate endpoint of ESRD [
2‐
5]. %eGFR changes of less than −30% over 2 years have been proposed as a surrogate endpoint of ESRD and death [
2,
6‐
8]. Although the applicability of %eGFR change as a surrogate endpoint has been examined in CKD patients in previous studies, it has not been evaluated yet in a general population. For this purpose, we examined whether %eGFR change can be used as a surrogate endpoint of ESRD using data of a community-based prospective cohort study in Okinawa, Japan. Thus, we (1) investigated the association between %eGFR change and the risk of ESRD, (2) determined the cutoff value that represents ESRD, and (3) evaluated the accuracy of %eGFR change for the prediction of ESRD.
Discussion
In this study, the association between %eGFR change and ESRD was shown in the high- and low-eGFR groups. This finding was consistent with those of previous studies [
2,
6,
8]. In Japan, it has been reported that 81.3% of the subjects showed GFRs of higher than 60 ml/min/1.73 m
2, and 18.7% of the subjects showed GFRs of 60 ml/min/1.73 m
2 or lower [
14]. From the public health viewpoint, eGFR change is an important index for the prevention of CKD progression in the general population.
The criteria for the strength of a surrogate endpoint are defined by the statistical principles for clinical trials of the International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH) E9 as follows: (1) the biological plausibility of the relationship, (2) the demonstration of the prognostic value of the surrogate for the clinical outcome in epidemiological studies, and (3) evidence from clinical trials showing that the treatment effects on the surrogate correspond to the effects on the clinical outcome [
15]. Regarding the first criterion, %eGFR change is biologically associated with ESRD because the loss of kidney function leads to ESRD; therefore, the biological plausibility of the relationship is self-evident. Thus, in this study, we mainly evaluated the second criterion.
The association between %eGFR change and ESRD has been examined in previous studies [
2,
6,
8,
16]. This study showed similar relationships, and that the risk of ESRD gradually increased with the decrease in %eGFR change over 2 or 3 years in the high- and low-eGFR groups. On the other hand, in the high- and low-eGFR groups, no relationship between the decrease in %eGFR changes over 1 year and slope and a high risk of ESRD was observed. The results of previous studies and this study taken together suggest that 2 or 3 years can be the baseline period in the high- and low-eGFR groups [
2,
6,
8]. Interventions, such as the use of angiotensin-converting enzyme inhibitors (ACEIs) and angiotensin II receptor blockers (ARBs), may reversibly change GFR acutely, thereby protecting kidney function [
5]. Moreover, it has been reported that acute change in GFR on initiation of renin–angiotensin system blockers is weakly associated with renal outcomes [
17]. Because this change in GFR observed in a short-term clinical trial may be temporary and acute, a long-term observation period is required to accurately evaluate the improvement of GFR. Therefore, the baseline period over 2 or 3 years would be better for the unit of %eGFR change than the baseline period over 1 year. The US National Kidney Foundation and the US Food and Drug Administration reported that “a confirmed decline in estimated GFR of 30% over 2–3 years may be an acceptable surrogate end point” [
7]. Because there is a possibility that the optimal baseline period may differ on the basis of the characteristics of the subjects, more studies of various populations are needed to determine surrogate endpoints.
For %eGFR change to be used as a surrogate endpoint of ESRD, its cutoff value should be determined. In this study, the number of the subjects with %eGFR changes of less than −40% was small. The distribution of the risk of ESRD showed that in the high- and low-eGFR groups, the risk of ESRD was higher in %eGFR changes of less than −30% over 2 or 3 years than in other ranges of %eGFR changes. These findings were in accordance with those of previous studies [
2,
6,
8]. These lines of evidence indicate that persons with %eGFR changes of less than −30% over 2 or 3 years have a high risk of ESRD.
To predict accurately a true endpoint, it is better that the specificity of a surrogate endpoint should be high. In this study, in the high- and low-eGFR groups, the specificities of %eGFR changes of less than −30% were more than 90%. Moreover, the relationship between a surrogate endpoint and a true endpoint is related to the proportion of persons with the surrogate endpoint who will have the true endpoint in the future. That is, PPV is based on a similar concept to a surrogate endpoint, and a good surrogate endpoint has a high PPV. In this study, the PPVs based on %eGFR changes of −30% over 2 or 3 years were higher than those based on other %eGFR changes in the high- and low-eGFR groups. These findings suggest that %eGFR changes of −30% over 2 or 3 years had the highest substitutability for ESRD. Therefore, −30% can be a candidate cutoff value for ESRD in the high- and low-eGFR groups. The doubling serum creatinine level is equivalent to −53% decline in eGFR, as determined on the basis of the JSN’s eGFR equation. Because the subjects with %eGFR changes of less than −53% were very rare in our study, the observation of doubling serum creatinine level may be difficult in a community-based study.
This study lacked information on the participants’ death, and the persons who died during the course of this study were not included in the analysis. CKD in the majority of affected patients does not progress to ESRD because such patients die before the onset of ESRD [
18‐
21]. Because we treated ESRD only as the outcome, we were unable to control for competitive risk bias. However, the subjects of this study were persons who underwent annual community-based health examinations and were healthier than usual CKD patients. The number of subjects with a high risk of death and with comorbidities, such as cardiovascular disease and diabetes mellitus, and the number of elderly CKD patients was small: therefore, the competitive risk bias might be small. Although the %eGFR changes in this study might have been different from those in CKD patients with various comorbidities, our study showed similar relationships to previous studies [
2,
6,
8,
16]. Regardless of the bias, our findings suggest that the relationships between the risk of ESRD and %eGFR change are robust. Moreover, these findings showed the existence of high risk persons in the general Japanese population who needed care of their kidney function.
In this study, in the low-eGFR group, a high risk of ESRD was associated with increases in %eGFR changes. This suggests that some of the subjects showing increases in %eGFR changes may have a high risk of ESRD. eGFR is estimated from serum creatinine level. Serum creatinine level is associated with skeletal muscle mass, and the weight-loss-associated decrease in skeletal muscle mass increases eGFR [
22‐
24]. Moreover, a cohort study of a nondiabetic population showed that prediabetes is associated with glomerular hyperfiltration [
25]. Considering these lines of evidence, some of the subjects showing increases in %eGFR changes are not always healthy but may have some comorbidities that lead to a decrease in muscle mass or diabetes mellitus.
This study has several limitations. First, because of the observational nature of this study, the results may be biased owing to unmeasured confounders. Second, we did not include the patients with missing data in this study. This might have caused selection bias. Third, because the number of ESRD patients was small, we were unable to investigate stratifications of baseline characteristics such as gender, old age, and urinary protein level. Fourth, comorbidities such as malnutrition and diabetes mellitus and medications such as ACEIs and ARBs were not investigated in this study. The effects of these factors on eGFR were not evaluated. Fifth, the number of subjects with %eGFR change of 0 to less than 10% over 1–3 years was smaller than those of the subjects with other %eGFR changes over 1–3 years in the high- and low-eGFR groups. There was a possibility that the rounding error of serum creatinine level might have caused the unbalanced distribution of %eGFR changes over 1–3 years. Sixth, the prediction capability of %eGFR change is not necessary applicable in the clinical practice as long as 15 years in the future, because interventions to the many risk factors of the CKD progression may be developed currently.