Background
Glomerular filtration rate (GFR) estimating equations based on creatinine and cystatin C perform well overall but yield different and sometimes contradictory results in a subset of individuals [
1‐
10]. Clinicians are chiefly concerned with the quality of estimated GFR (eGFR) for individual patients for drug dosing, diagnosis of kidney disease, and follow-up of kidney function. To improve GFR estimation methods at the individual level it is first necessary to understand which, if any, particular subgroups are prone to poor estimation using one or both biomarkers. Biomarker concentrations are affected by GFR and non-GFR determinants, some of which are corrected for in current estimation equations, for example with coefficients for race and sex. Better understanding of affected subgroups could lead to recommendations on which biomarker to favor when choosing an equation for a patient in a particular subgroup, or could lead to development of additional correction coefficients for non-GFR determinants of biomarker concentrations. Some recent studies have also indicated that a large difference between cystatin C based and creatinine based eGFR can be a marker for increased risk for morbidity and mortality [
11‐
13]. The pathophysiology of this association has not been determined although glomerular pore size has been hypothesized to play a role. Knowledge of which groups may be more prone to large differences in GFR estimates may give insight into this phenomenon.
Most previous studies have focused on determining which eGFR equation is optimal compared to a gold standard. Several studies have examined factors other than GFR affecting creatinine or cystatin C [
14‐
33]. Few have looked specifically at factors affecting consistency between different eGFR equations [
34‐
38].
The first objective of this study is to determine the prevalence and size of differences in GFR estimation using creatinine-based (eGFRcreat) or cystatin C-based (eGFRcys) equations in a community-dwelling population of older adults. The main objective of the study is to explore correlations between non-GFR determinants of biomarkers and differences in eGFR with the aim of finding subgroups for further study.
Discussion
This study demonstrates that in a community-dwelling population of older adults, eGFRcreat and eGFRcys yield estimates that differ by more than 10% in nearly two-thirds of cases. In nearly 20 % of cases the two estimates differ by more than 30%. Our results show that smoking, age, BMI, CRP and glucocorticoid use are correlated with increasing eGFRcreat or decreased eGFRcys while mean eGFR displays the inverse correlation.
Although it is not possible from this study to know whether the findings are chiefly attributable to effects on creatinine or cystatin C, these findings are not in conflict with previous research on biomarker determinants [
15‐
18,
21‐
23,
30,
32,
34,
36,
38,
44]. Hypothetical explanations for the correlations seen in this study include a decrease in serum creatinine due to loss of muscle mass in the case of increasing age, prolonged glucocorticoid use or inflammatory disease; the latter being reflected by increased CRP. There have been numerous hypotheses regarding the influence of non-GFR determinants, including body composition, on cystatin C but the physiological mechanisms are not as clear as in the relationship between muscle mass and creatinine levels. Results have often been conflicting regarding the extent of influence of lean body mass, adiposity, diabetes, smoking, level of inflammation, and thyroid function on the cystatin C concentration in blood [
16,
19‐
21,
23,
25,
26,
28,
33,
35‐
37,
48]. In addition to the potential effect of non-GFR determinants on cystatin C and creatinine production, these biomarkers may also be differentially eliminated in the glomeruli. Previous research has hypothesized the existence of a shrunken pore syndrome, wherein various pathological factors could lead to changes in glomerular membrane pore diameter [
11,
49‐
51]. This could in turn explain differing filtration rates of differently-sized macromolecules, in this case creatinine (113 Da) and cystatin C (13.3 kilodaltons). Recent studies in select adult populations have shown that close to 10% of patients studied display a ratio of eGFR
cys to eGFR
creat less than or equal to 0.6, which has been defined by researchers as indicative of shrunken pore syndrome [
49]. These patients are at generally increased risk for morbidity and mortality, and at higher risk for right ventricular dysfunction and for death after coronary artery bypass grafting [
11‐
13]. It is unknown to what extent the hypothesized shrunken pore syndrome may explain differences in GFR estimates in elderly populations compared to other non-GFR determinants of cystatin C and creatinine that were found to be correlated to differences in eGFR in the current study. The above hypothetical pathophysiological models for the observed differences in GFR estimates can be kept in mind when designing future studies.
The results of this study should prompt clinicians to consider whether one or both biomarkers should be used for GFR estimation in older adults. A clinically important difference between eGFRcys and eGFRcreat is not unusual and should be anticipated when the patient profile includes factors known to affect biomarkers.
A strength of this study is its sampling from a general community-dwelling population of elderly both with and without chronic kidney disease.
Choice of risk variables was based on factors known to affect non-GFR determinants of creatinine and cystatin C. One of the strengths of this study is that most factors known to significantly affect biomarkers are included as variables. However, chiefly for reasons of power, not all potential variables were included. For instance, usage of cimetidine or trimethoprim, which have been associated with changes in creatinine metabolism [
24,
45,
47], were not included as the total number of study participants taking one of these medications was only six (0.2%).
This study is subject to the common self-selection bias of participants that are healthier on average than the general population of older adults, despite efforts to minimize bias by offering home visits to participants unable to come to the study center. Our assumption is that a healthier population with fewer individuals in the disease-defined biomarker determinant categories decreases the chances of finding correlations between risk variables and the outcome variable. In this age cohort in the study region there are few, if any, non-white individuals. This could affect the generalizability of these results. Race was not recorded during the study, so the exact number of non-white participants is not known.
Although efforts were made to decrease misclassification of exposure variables by accessing both medical records and participant self-report, some of the variables may be subject to bias, most notably variables that rely on medical diagnoses. These are the categories of treatment for hypertension, diabetes, and thyroid function. Hypertension, diabetes, and thyroid function are not routinely screened for in Sweden and therefore it is likely that some proportion of participants were undiagnosed, potentially leading to a misclassification bias. It is also important to mark a distinction between the presence of illness and the presence of treated illness. However, generally speaking, in the Swedish healthcare system all patients with diagnosed diabetes, current imbalances in thyroid hormone production or clinically relevant hypertension receive treatment. The above limitations would tend to bias towards unity and may explain why we were unable to find a correlation between these factors and differences in eGFRcreat and eGFRcys.
A factor limiting the scope of our study is the lack of measured GFR, meaning we are limited to exploratory analyses of the correlations between biomarker determinants and inconsistencies in estimated GFR. In addition, our consideration of anthropometrics was limited to the use of BMI. Specific consideration of muscle mass and adiposity in relation to the biomarkers and eGFR would require a more accurate measure of body composition, such as Dual-Energy X-ray Absorptiometry (DEXA).
Acknowledgements
Our thanks go to the administrative and clinical staff of the GÅS study for their invaluable work, to Ole Larsen for his vital assistance in quality control of the data, and to the study participants who have made this work possible.