Prospective longitudinal cohort study
Data will be analysed to address three main questions:
1.
Which of the GFR-estimating equations is the most accurate assessment of reference GFR?
2.
Which GFR-estimating equation most accurately reflects change in GFR?
3.
Which GFR-estimating equation, together with ACR, or ACR alone, most accurately predicts those people that have progressive loss of kidney function (CKD progression)?
In each case, data will be further analysed to assess whether observed relationships amongst African-Caribbean and South-Asian subjects differ from those observed amongst Caucasians, and whether diabetes and proteinuria are predictors.
1. Which of the GFR-estimating equations is the most accurate assessment of reference GFR?
Accuracy will be assessed by establishing the proportion of GFR estimates within 30% (P30) of iohexol GFR, using baseline measures. P30 values will be compared between GFR estimating equations using McNemar’s test for paired data. Additional analysis will consider the mean and median, interquartile range and root mean square error of the distribution of differences.
2. Which GFR-estimating equation most accurately reflects change in GFR?
Rate of change in eGFR will be established by linear regression [
54] utilising all available (maximum 7) eGFR time points, and will be compared with the difference between final and baseline reference GFR values. Differences in large error rates (greater than 3 mL/min/1.73 m
2/year, or greater than 5%/year difference in slope) will be compared using McNemar’s test. Additional analyses will consider the predictive ability of the tests to detect i) a change in iohexol GFR of greater than 25%; ii) a decline in GFR of greater than or equal to 10 mL/min/1.73 m
2 over the three years; and iii) a change that exceeds the RCV derived for the reference GFR test in the measurement variability study described below.
3. Which GFR-estimating equation, together with ACR, or ACR alone, most accurately predicts those people that have progressive loss of kidney function (CKD progression)?
Models will be constructed to predict time to progression based on baseline eGFRs and ACR. Progression will be defined in terms of decline in reference GFR (change in iohexol GFR > 10 mL/min/1.73 m
2) or an increase in albuminuria category, as suggested by KDIGO [
38]. Progression will only be detected at one of 6 time points, hence piecewise survival models will be fitted to determine whether the prognostic value of ACR and the estimated GFRs is independent of other risk factors. We will develop a prognostic model utilising age, gender, ethnicity, body mass index, waist circumference, mean arterial blood pressure, diabetes mellitus, smoking status, and presence of vascular disease in addition to baseline ACR and the various eGFRs. Both proportional and non-proportional hazards will be considered. Bootstrap validation will be used with these prediction models.
Study of intra-individual biological variation
Pre-analytical variables will be standardised as described above. All samples for all analytes will be assayed in duplicate and the analytical variance (SD
A
2) will be calculated from the differences between the duplicate measurements. The total (CV
T), analytical (CV
A) and within-individual (CV
I) components of variation will be calculated using nested ANOVA [
19]. The critical difference (reference change value, RCV) for significant changes in serial results (P less than 0.05) and the number of specimens required to estimate the homeostatic set-point of an individual (within ±10% with a confidence of 95%) will also be estimated. The derived RCV for the reference GFR will be used to test the ability of estimated GFR equations to detect a true change in GFR (see study 1, part 2 above).
Modelling monitoring strategies
Whilst our longitudinal cohort will not have adequate power to detect differences in progression, our estimates of the accuracy of eGFR (study 1), patterns and determinants of progression (study 2), and intra-individual biological variation (study 3) can be combined in a model to evaluate the impact of alternative monitoring strategies on detection of progression to stage 4 CKD. True GFR values will be modelled over time for representative cohorts of people, and performance of alternative monitoring strategies in detecting progression to stage 4 CKD (varying in timing and choice of eGFR equation) will be simulated utilising estimates of measurement error and accuracy. Outcome variables which will be assessed will include false positive progression rates, and the sensitivity and delays in detecting progression.
Health economics study
The aim of the health economics evaluation is to determine the cost-effectiveness of implementing cystatin C-based eGFR or a combination of both cystatin C and creatinine-based eGFR in subjects that are initially stage 3 CKD compared with MDRD (creatinine-based) eGFR alone. The cost-effectiveness analysis will take the form of a cost-utility analysis in which the outcome measure will be the cost per quality adjusted life year (QALY). This will be undertaken by extending the monitoring strategy by extrapolating the rate of change in GFR beyond the end of the trial through the use of secondary data sources to link the error in estimated GFR to patient outcomes, which will include myocardial infarction, kidney transplant, and established renal failure. It is unlikely that differences in quality of life will be seen for patients receiving alternative monitoring strategies during the period of the trial, and therefore quality of life data (e.g. EQ-5D) is not being collected during this study. Instead secondary sources will be used to inform the impact of the long term outcomes of CKD on quality of life.
Cost data collection will be undertaken prospectively for all subjects in the cohort study in order to inform the cost component of the cost-effectiveness analysis. The main resource uses monitored during the trial will include the following:
1.
Diagnostic testing procedures implemented.
2.
Resource uses involved in the diagnostic testing procedures (e.g. urine ACR).
3.
Other related procedures, including level of health care professional involvement in the procedure, equipment required, overheads, consumables etc.
The costs obtained will be dictated by the recommended investigations and interventions by GFR category in NICE Clinical Guideline 73 [
6]. The difference in cost will be that between GFR category and rate of change in GFR assessed by creatinine-based estimating equations versus cystatin C-based equations.
The model-based analysis will be carried out following the conclusion of the data collection undertaken during the cohort study. A decision analytic model will be used to allow extrapolation of the cost and effectiveness parameters beyond the data observed during the trial (and to allow extrapolation to other settings). The model will consider the impact of the error in GFR measures on patient outcomes. The results of the economic analysis will be presented using cost-effectiveness acceptability curves to reflect sampling variation and uncertainties in the appropriate threshold value by which the cost-effectiveness of the different diagnostic strategies will be judged.