Our study shows that moderate to severe kidney disease occurs commonly among individuals with anaemia. The prevalence of kidney disease among anaemic persons was greater with eGFRCYSC compared with eGFRSCR and eGFRSCR + CYSC. Overall, persons who were differentially classified as having moderate to severe kidney disease by eGFRSCR and eGFRCYSC had similar prevalence of kidney disease risk factors; however, those who were classified as having moderate to severe kidney disease by eGFRCYSC alone were more likely to have microalbuminuria and elevated CRP, two conditions which have been linked to anaemia, compared to individuals who were classified as having kidney disease by eGFRSCR only.
Prior studies have demonstrated the inverse association between kidney function and anaemia [9,37
]. Anaemia develops early in the course of chronic kidney disease, with nearly one-third of the individuals with a GFR of 60–89
meeting WHO criteria for anaemia [37
]. However, these prior studies either utilized serum creatinine-based estimates of kidney function or were not representative of the general population [9,37
]. In contrast, our study used a large, nationally representative population to examine the differences in the association of anaemia with kidney dysfunction based on serum creatinine, cystatin C or both.
Anaemia commonly occurs in individuals with chronic conditions such as diabetes [38,39
] and hypertension [40
], which are also known risk factors for chronic kidney disease [41
]. Recommended clinical approaches to the workup of anaemia, however, only marginally refer to kidney disease as a possible cause for normocytic anaemia and neglect to recommend renal function assessment in anaemic individuals [42
]. In our study, we found that >15% of anaemic individuals had impaired kidney function when based on eGFRSCR
or eGFRSCR + CYSC
. The use of eGFRCYSC
to estimate kidney function led to a 9% higher prevalence estimate of kidney disease in anaemia. The mechanisms underlying these disparities among the different eGFR methods likely stem from differing effects of extrarenal factors on both biomarkers and haemoglobin. In the Multi-Ethnic Study of Atherosclerosis (MESA), eGFRSCR
inversely correlated while serum cystatin C positively correlated with several inflammatory markers in persons with chronic kidney disease [25
]. In individuals without impaired kidney function, serum cystatin C remained significantly correlated with several inflammatory markers while eGFRSCR
only correlated with tumour necrosis factor-α receptor1 (TNF-αR1) [25
]. This study did not evaluate the association between eGFRCYSC
and inflammation, which would have partially accounted for the effects of age, race and gender on serum cystatin C [25
]. A more recent study, however, of >3000 individuals supports the notion that serum cystatin C and creatinine are differentially affected by non-renal factors [44
]. After adjustment for measured GFR, age and gender were found to have greater effects on serum creatinine than on serum cystatin C. Whereas serum creatinine was 9.2% lower with each 20-year increase in age and 31.7% lower in women, serum cystatin C was 4.3% and 9.2% lower, respectively. Moreover, a higher CRP was associated with a lower serum creatinine (−3.3%) while it was associated with a higher serum cystatin C (2.3%). The association of these factors with serum creatinine noticeably diminished but had minimal effect on these associations with serum cystatin C after further adjustment for proxies of muscle mass [44
]. Our results showing an overall odd ratio for kidney disease based on eGFRSCR + CYSC
, which was in between those by eGFRSCR
, imply that the combined equation may mediate some of the differential effects of extra-renal factors on serum creatinine and cystatin C. However, this hypothesis could not be tested given our study’s lack of direct GFR measurements. Nonetheless, the use of eGFRCYSC
to assess kidney function in anaemic persons may be particularly helpful in persons aged <60 years, women and those with ongoing inflammation. Its stronger association with mortality compared with eGFRSCR
and eGFRSCR + CYSC
] may provide an added benefit of prognostication in using eGFRCYSC
for kidney function examination in anaemia.
The limitations of our study to consider include its cross-sectional design and our choice of equation to calculate eGFRCYSC
. Due to a lack of temporality inherent in cross-sectional studies, we are unable to determine if serum cystatin C predicts earlier declines in haemoglobin than serum creatinine. However, NHANES III provides a unique opportunity to examine the relationship between eGFRCYSC
and haemoglobin in an ethnically diverse, nationally representative sample. Although the eGFRCYSC
equation used in our study addressed some biases associated with age, race and gender, it may not fully account for bias as the equation was developed in a study population enriched with participants afflicted with chronic kidney disease [33
]. Our study lacks direct GFR measurements; therefore, we cannot discern the true impact of extra-renal influences on serum creatinine, serum cystatin C and anaemia. No direct comparisons have been performed between the eGFRCYSC
equation we used and those developed by other investigators; however, we believe that this equation, which was externally validated [33
], currently provides the most reliable estimate of GFR based on serum cystatin C in adults. Our prevalence estimates for microalbuminuria differed from those previously reported by Coresh and colleagues [34
]. A selection bias may have occurred in the process of selecting the NHANES III sub-sample included in our analysis. Alternatively, our prevalence estimates may have diminished accuracy compared with those by Coresh and colleagues given our smaller study sample size [34
]. Despite these limitations, our study provides a thorough comparison of the association of anaemia with kidney disease based on the serum creatinine- and cystatin C-based estimates of kidney function.
In conclusion, the prevalence of kidney disease among anaemic persons was greater with eGFRCYSC than with eGFRSCR and eGFRSCR + CYSC. Anaemia may be more strongly associated with eGFRCYSC rather than eGFRSCR and eGFRSCR + CYSC in persons aged <60 years, women and those with ongoing inflammation. This observation may be due to disparities in the effect of non-renal factors on both renal biomarkers. Further studies with measured GFR are needed to examine the differential effects of non-renal factors on the three GFR-estimating equations. Our study suggests that an assessment of kidney function, regardless of which GFR-estimating method is used, may need to be incorporated in to the routine workup of anaemia given the high prevalence of renal disease among anaemic individuals.