Change in glomerular filtration rate over time in the Oxford Renal Cohort Study: observational study

Background Decline in kidney function can result in adverse health outcomes. The Oxford Renal Cohort Study has detailed baseline assessments from 884 participants ≥60 years of age. Aim To determine the proportion of participants with a decline in estimated glomerular filtration rate (eGFR), identify determinants of decline, and determine proportions with chronic kidney disease (CKD) remission. Design and setting Observational cohort study in UK primary care. Method Data were used from baseline and annual follow-up assessments to monitor change in kidney function. Rapid eGFR decline was defined as eGFR decrease >5 ml/min/1.73 m2/year, improvement as eGFR increase >5 ml/min/1.73 m2/year, and remission in those with CKD at baseline and eGFR >60 ml/min/1.73 m2 during follow-up. Cox proportional hazard models were used to identify factors associated with eGFR decline. Results There was a net decline in eGFR in the 884 participants over 5 years of follow-up. In 686 participants with >2 eGFR tests with a median follow-up of 2.1 years, 164 (24%) evidenced rapid GFR decline, 185 (27%) experienced eGFR improvement, and 82 of 394 (21%) meeting CKD stage 1–4 at baseline experienced remission. In the multivariable analysis, smoking status, higher systolic blood pressure, and being known to have CKD at cohort entry were associated with rapid GFR decline. Those with CKD stage 3 at baseline were less likely to exhibit GFR decline compared with normal kidney function. Conclusion This study established that 24% of people evidenced rapid GFR decline whereas 21% evidenced remission of CKD. People at risk of rapid GFR decline may benefit from closer monitoring and appropriate treatment to minimise risks of adverse outcomes, although only a small proportion meet the National Institute for Health and Care Excellence criteria for referral to secondary care.

to have the condition at recruitment to the study, whereas 44% were diagnosed through screening. 15 Some OxRen participants now have over 5 years' follow-up data, allowing exploration of the changes in eGFR over time and identification of risk factors associated with eGFR decline. Follow-up data were used to determine the proportion of participants who experienced rapid eGFR decline, describe numbers of people moving in and out of remission, and identify determinants of rapid GFR decline and remission.

METHOD Study population and laboratory methods
Participants aged ≥60 years in UK primary care were screened for CKD by determining their albumin-creatinine ratio (ACR) and Modification of Diet in Renal Disease eGFR. 17 Briefly, participants were screened to identify those who had a CKD diagnosis confirmed by two positive tests a minimum of 90 days apart and those who did not have a CKD diagnosis, but had a single test with either a decreased eGFR (<60 ml/min/1.73 m 2 ) or a raised urinary ACR (>3 mg/mmol). Participants with existing CKD or those with a positive screening test based on eGFR and ACR were invited to a baseline assessment. All laboratory measurements were performed in a single laboratory and those with existing CKD or a positive screening test based on eGFR and ACR attended a baseline assessment. Those who had a baseline assessment by December 2019 (n = 884) were included in the analysis. Full details of OxRen have been reported elsewhere. 15 Data management decisions on laboratory test results made before analysis are detailed in the Supplementary data (see Supplementary Box S1).

Outcomes
Potential CKD progression was calculated in participants who had ≥2 eGFR measurements at least 12 weeks apart. 18 In the primary analysis, rapid GFR decline was defined as a decrease in eGFR of at least 5 ml/min/1.73 m 2 per year in line with the Kidney Disease Improving Global Outcomes (KDIGO) definition. 1 Numbers who met different definitions of CKD progression were also reported, including a decline in eGFR of at least 15 ml/min/1.73 m 2 per year, a 25% decline in eGFR combined with progression to the next CKD stage (indications for referral to a nephrology service), 1 and a decline in the slope of at least 5 ml/min/1.73 m 2 per year through regression of ≥3 readings.
Improvement was defined in this study as both an increase in eGFR of at least 5 ml/ min/1.73 m 2 per year and remission as eGFR >60 ml/min/1.73 m 2 and ACR <3 mg/ mmol. 11 Progression and remission of CKD in 403 people with full baseline and 2-year follow-up results were presented in a flow chart to show the proportions that enter and leave remission at each visit. Movement between each stage of CKD at baseline and at year 2 were presented in a table.

Statistical analysis
All analyses were performed using Stata SE16 (Statacorp). Descriptive data on eGFR and numbers who progressed were presented in tables, bar charts, and graphs as mean and standard deviation (SD) of eGFR at baseline and each year of follow-up. Covariates used in the primary analysis included demographic data, anthropometric measurements, and history of CKD (see Supplementary Table S1). An extended analysis was conducted that included history of comorbidities collected at participant's baseline assessment.
The primary analysis was conducted using univariable and multivariable Cox proportional hazards regression analyses to estimate hazard ratio (HR) and 95% confidence intervals (95% CIs) for the association of covariates with time to progression or censoring at the end of follow-up (December 2019). The dependent variable was time to first GFR that was >5 ml/min/1.73 m 2 per year lower than the baseline value. The index date was the date of the participant's baseline eGFR test. Assumption of proportional hazards was verified and no violations were observed for

How this fits in
The prevalence of chronic kidney disease (CKD) increases with age and some people are unaware that they have CKD. It is established that some people with CKD will go on to develop adverse health outcomes, and understanding what makes this more (or less) likely is important in primary care. This study reports the number of people with CKD who experience a rapid deterioration of kidney function measured as a decline in estimated glomerular filtration rate. It identifies factors associated with a rapid decline in kidney function and describes the number of people who enter and leave CKD remission during follow-up.
any of the predictors. Because of skewed data, urinary ACR was logarithmically transformed and alcohol consumption was dichotomised to no alcohol use or any alcohol use before analysis. Improvement in eGFR was analysed using the same covariates using Cox regression. All covariates were included in the multivariable analysis.
To determine whether eGFR variability was associated with rapid eGFR decline, participants' laboratory eGFR results before and up to their baseline assessment date were used to calculate variability. Variability independent of the mean (VIM) 9,19,20 was calculated using the following equation: 100 × SD/mean β . Where β was the regression coefficient, calculated from a simple linear regression of the natural logarithm of the SD against the natural logarithm of the mean within-individual eGFR.
The VIM was classified into four quartiles, the highest quartile representing the participants with the highest eGFR variability, 9 which were used as categorical variables in the analysis. Because this method has not been widely used, sensitivity analyses were performed using mean eGFR and SD, 10,21 and the coefficient of variation (CV) 10,19 to ensure that the choice of methods had not affected the outcome.
There were 547 people (62%) who had a follow-up eGFR measurement 1 year from baseline; 377 (43%) 2 years from baseline; 211 (24%) 3 years from baseline; 65 (7%) 4 years from baseline; and 28 (3%) with an eGFR 5 years from their baseline visit ( In total, 686 participants had ≥2 eGFR measurements during follow-up and were included in the analysis of CKD progression. Mean age and sex were similar to the full cohort (Table 1). Of these 686 participants, 164 (24%) experienced rapid GFR decline  (>5 ml/min/1.73 m 2 eGFR decrease per year), 27 (4%) experienced progression of >15 ml/min/1.73 m 2 /year, and 48 (7%) experienced a 25% decrease in eGFR combined with progression to the next stage. In total, 30 of 423 (7%) people experienced a sustained decline in the slope who had ≥3 eGFR measurements of >5 ml/min/1.73 m 2 . Additionally, 185 (27%) people experienced improvement in eGFR of at least 5 ml/min/1.73 m 2 /year and 82 of 394 (21%) people who fell into CKD stages 1-4 at the baseline visit experienced remission of their CKD at some point during follow-up. These data and numbers of participants with known CKD, CKD identified by screening, and without CKD who progressed are shown in Table 2.
During follow-up, there were 224 withdrawals in the full cohort (n = 884) and 27 deaths, and for those with ≥2 eGFR measurements (n = 686), there were 122 withdrawals and 17 deaths.

Regression models
The univariate Cox model revealed significant associations of rapid eGFR decline with being a former smoker, higher systolic and diastolic blood pressure, higher eGFR, and higher log ACR at baseline. In the multivariable analysis, being a former smoker (HR 1.457, 95% CI = 1.033 to 2.054), higher systolic blood pressure (HR 1.013, 95% CI = 1.004 to 1.022), and being known to have CKD at cohort entry (HR 1.670, 95% CI = 1.140 to 2.446) were associated with rapid GFR decline (Table 3). Those with CKD stages 3a or 3b at baseline were less likely to evidence rapid GFR decline, compared with those with normal kidney function. Extending the model to include history of comorbidities at baseline did not identify any other factors to be significantly associated with rapid eGFR decline (see Supplementary Table S2).
There were no associations between variability in eGFR before the baseline visit and rapid eGFR decline. Including eGFR variability in the multivariable analysis resulted in female sex, having higher body mass index (BMI), being a former smoker, and having stage 3a CKD at baseline being significantly associated with progression of  CKD. Results did not change substantially when variability was defined as mean and SD or CV of eGFR in place of VIM (see Supplementary Table S3).
Lower systolic blood pressure and lower log ACR were significantly associated with CKD improvement in the univariable analysis. Only lower systolic blood pressure remained significant in the multivariable analysis (see Supplementary Table S4). Those with lower ACR were more likely to experience remission and those with known CKD were less likely to experience remission in the univariable analysis. Only lower ACR remained significant in the multivariable analysis (see Supplementary  Table S4). Figure 3 shows that people move across the CKD diagnostic threshold over a 2-year follow-up in both directions, with 30%-40% of the cohort falling below the eGFR/ACR threshold for CKD at any one time and 139 participants remaining below CKD diagnostic thresholds over the entire follow-up. When CKD stage changes between baseline and year 2 were mapped, it was found that,  although 59% of people remained in the same stage, 41% had changed stage (see Supplementary Table S5). Of participants who were classified as having CKD at baseline, 83% were also classified as having CKD at their 2-year follow-up, but 17% fell into the non-CKD range after 2 years (see Supplementary Table S6). Of those with CKD at baseline, 8% of people moved to a less advanced CKD stage and 18% moved to a more advanced CKD stage after 2 years.

DISCUSSION Summary
In this older primary care population with early-stage CKD or some renal impairment, the current study determined that 24% of people experienced rapid eGFR decline of at least 5 ml/min/1.73 m 2 /year. It was also identified that having a higher BMI, being a smoker, having a higher systolic blood pressure, and being known to have CKD at cohort entry were all associated with rapid GFR decline in this population. In addition to eGFR and CKD progression, it was found that 27% of people experienced an improvement in eGFR, and that having a lower systolic blood pressure showed the strongest association with this improvement. Those who experienced remission were most likely to have a lower ACR. During 2 years of follow-up, at any one time, approximately 30% of people were in remission, despite many of them meeting the definition of CKD at some point.

Strengths and limitations
This article provides reliable, prospectively collected data from an observational study that will help clinicians to prioritise those at greatest need of closer monitoring or medication reviews. Although eGFR decline is indicative of deterioration in kidney function and CKD, the data in the current study demonstrate that, in some instances, it may be a temporary decrease, such as an episode of acute kidney injury following which kidney function recovers. This and other potential reasons for a temporary decline in kidney function include biological and analytical variability, which are outlined in Supplementary Table S7. The KDIGO guidelines 1 state that confidence in the precision of the decline increases with an increasing number of serum creatinine tests and duration of follow-up. The impact of a larger number of eGFR results on this analysis were explored, but fewer events meant that it was not possible to conduct a full analysis. Furthermore, those with a more advanced stage of CKD at baseline were less likely to experience eGFR progression, which represent 'regression to the mean'. This has been noted in other studies, 22,23 and arises when day-to-day biological variations and imprecision in the measurement result in some people by chance having a spuriously high eGFR at their baseline assessment, meaning they may be more likely to meet the definition of progression in subsequent measurements. In the current study, therefore, numbers of participants who meet different definitions of progression and sustained progression were also reported. In addition, the numbers of people whose eGFR improved and the number who experienced remission, of those who met the definition of having CKD, have been reported.
The follow-up time for this cohort was relatively short, which may have had an impact on the numbers of people who experienced progression. Whereas the proportion of participants experiencing a short-term rapid decline in GFR was relatively high (24%), the proportion with GFR decline sufficient to prompt referral to secondary care was low (7%).
This analysis has found that those who were already known to have CKD may be more likely to experience rapid eGFR decline than those who were identified to have CKD on screening but, depending on the statistical method used, this was not consistent, which may result from a lack of statistical power. However, characteristics of the OxRen baseline population have already been reported and those with existing CKD had lower eGFR than those who were diagnosed through screening, 15 so may have been living longer with CKD.

Comparison with existing literature
A previous study reported that 18% of people experienced CKD progression over 5 years when defining progression as a 25% decline in eGFR combined with moving to a higher CKD stage. 11 Using this definition, it was found that only 7% of this cohort progressed over a shorter follow-up time. Furthermore, if the slope of ≥3 eGFR measurements was used to define progression, only 7% of the OxRen cohort progressed, suggesting that some of those with rapid decline may have had more variability in their eGFR measurements. Other studies reported progression in 15% to 20% of participants, 24,25 with higher rates of progression in those with the most severe CKD. 24 The proportion who experienced an improvement in eGFR was similar to other studies, ranging from 12% to 27% depending on stage of CKD. 11,24,25 The current study found that kidney function fluctuated during follow-up as people move back and forth across the 60 ml/min/1.73 m 2 eGFR threshold over time and, therefore, that a substantial minority of participants evidenced 'remission' of their CKD. This type of variability in eGFR has been described previously, 24,25 with one study in UK primary care reporting a similar observation, 11 such that over 25% of participants were 'in remission' at any one time. A slightly higher proportion (30%-40%) demonstrating remission after 1 and 2 years of follow-up was found in the current study, but some people in the cohort had an earlier stage of CKD, or did not fully meet the definition for CKD on cohort entry. Many of these fluctuations will result from normal biological variability and measurement error in people who were near measurement thresholds. However, those who remain in remission at baseline and year 1 have lower mortality (3.2%) compared with those who meet criteria for CKD at both visits (15.7%). 11 The issue of CKD 'remission' and the associated clinical implications warrants further investigation. However, regardless of individuals who cycle back and forth between CKD and 'remission', the overall outcome during 2 years of follow-up was an overall trend towards a lower eGFR over a 2-year period. Whether those with rapid GFR decline in this study go on to develop adverse health outcomes is yet to be determined. With continued follow-up of this cohort, data on long-term health and mortality can be evaluated in relation to CKD progression.
In the current analysis of variability in eGFR before entering the cohort, no association with subsequent eGFR decline was found. To the authors' knowledge, only one previous study has used similar methods to explore eGFR variability and found that it was associated with increased mortality in a population with existing risk factors for cardiovascular disease. 10

Implications for practice
This work adds to existing knowledge of CKD progression and GFR decline as well as tools that can be used to identify risk of progression. The findings will help clinicians identify patients who may be at greatest risk of deteriorating renal function in order to prioritise closer monitoring or medication reviews to reduce cardiovascular risk including blood pressure management and lifestyle interventions. 1 Particular attention should perhaps be directed towards females who are overweight with a history of smoking. Nevertheless, it should also be appreciated that GFR may fluctuate over time and that only a small proportion of people with a sustained and large GFR decline warrant referral to a nephrology service.
In the OxRen cohort, those known to have CKD on cohort entry were more likely to experience rapid GFR decline than those who were only found to have CKD on screening and those who did not meet the KDIGO criteria for CKD at their baseline assessment. This preliminary analysis casts some doubt on the value of screening for CKD in older populations as those at highest risk may be those already known to have CKD.
In conclusion, this work has improved understanding about how eGFR fluctuates over time and will help clinicians identify those who may benefit most from closer monitoring and appropriate treatment to minimise risks of adverse outcomes.