low cystatin-c vs creatinine predicts bad outcomes

 

A new meta-analysis confirmed the rather common finding of significant discordance between estimated GFR (eGFR) by cystatin vs creatinine, finding that when eGFR was significantly lower by cystatin, there were significant increases in all-cause mortality, cardiovascular events, and kidney failure (see ckd cystatin better than creat JAMA2025 in dropbox, or doi:10.1001/jama.2025.17578)

 

Details:

-- the data source for this meta-analysis was the Participants in the Chronic Kidney Disease Prognosis Consortium (CKD-PC), focusing on participants with concurrent Cystatin-C (referred to as cystatin below) and creatinine measurements (eGFRcys and eGFRcr, respectively) and clinical outcomes, using an individual-level meta-analysis

-- 23 outpatient cohorts: total of 821327 individuals

    -- mean age 59, 48% female, 13.5% with diabetes, 40% with hypertension

    -- mean eGFRcr 87, mean eGFRcys 81, median eGFRcys-eGFRcr difference of -5.4%, mean eGFRcys-eGFRcr/eGFR < -30% in 92154 individuals (11.2%), mean eGFRcys-eGFRcr/eGFR> -30% in 31024 individuals (3.8%)

        -- the range of the eGFRdiff >30% was quite varied in the 23 studies, ranging from 0.2% to 27.9% of individuals

-- 2 inpatient cohorts: total of 39639 individuals

    -- mean age 67, 31% female, 30% with diabetes, 72% with hypertension

    -- mean eGFRcr 73, mean eGFRcys 62, median eGFRcys-eGFRcr difference of -15.4%, mean eGFRcys-eGFRcr/eGFR < -30% lower in 8965 individuals (29.6%), mean eGFRcys-eGFRcr/eGFR > -30% lower in 1908 individuals (6.3%)

        -- the range of eGFRdiff >30% was from found in 38.1% and 43.8% in the 2 cohorts, and this was higher in those with heart failure (38.0% if eGFR diff >30% and 21.1% in those with eGFRdiff of -30% to 30%)

 

-- Results below were adjusted for age; female sex; smoking status; history of hypertension, diabetes, coronary heart disease, stroke, HF, atrial fibrillation, PAD, cancer, COPD, and  liver disease; BMI (continuous, with a spline knot at 30); eGFRcr categories (≥90, 60-89, 45-59, 30-44, or <30mL/min/1.73m²); and log-uACR (urinary albumin-to-creatinine ratio)

 

-- Main outcomes and measures: the primary independent measurement was a large negative eGFR difference (eGFRdiff), defined as an eGFR calculated by cystatin (eGFRcys) that was at least 30% lower than their eGFR calculated by creatinine (eGFRcr)

    -- secondary (dependent) outcomes included all-cause and cardiovascular mortality, atherosclerotic cardiovascular disease (ASCVD), heart failure, and kidney failure with replacement therapy (KFRT)

 

-- outpatient participants had a mean followup of 11 years

  

Results:

--oupatient group followup: all-cause mortality 107584; cardiovascular deaths 25465; ASCVD events 35133; KFRT events 10060

    -- median eGFRdiff was -5.4% (-15.3% to 2.9%)

    -- outpatient participants: 11.2% had a large negative eGFRdiff (eGFRcys at least 30% lower than eGFRcr)

    -- 3.8% had a large positive eGFRdiff (eGFRcys at least 30% higher than eGFRcr)

-- inpatient group: followup

    -- 34.2% had a large negative eGFRdiff

    -- 6.4% had a large positive eGFRdiff

 

-- compared with an eGFRdiff between −30% and 30%:

    -- all-cause mortality: 28.4 vs 16.8 per 1000 person-years, a 69% increased hazard ratio [HR] 1.69 (1.57-1.82)

    -- cardiovascular mortality: (6.1 vs 3.8 per 1000 PY,  a 61% increased HR 1.61 (1.48-1.76)

    -- atherosclerotic cardiovascular disease: 13.3 vs 9.8 per 1000 PY, a 35% increased HR 1.35 (1.27-1.44)

    -- heart failure: 13.2 vs 8.6 per 1000 PY, a 54% increased HR 1.54 (1.40-1.68)

    -- kidney failure with replacement therapy (KFRT): 2.7 vs 2.1 per 1000 PY, a 29% increased HR 1.29 (1.13-1.47)

        -- all of the above were statistically significant

-- the following graph shows that the adverse outcomes associated with eGFRdiff overall increased in frequency as that difference between the eGFRcyst vs eGFRcr increased, in the fully adjusted model

    -- of note, the clinical benefit on measuring cystatin and creatinine and calculating eGFR difference starts to reveal increased adverse events as soon as the difference was 5-10%  and then continued to increase as the difference increased to -59% (the left side of the graph)

    -- the converse was that those with lower eGFRcr than eGFRcyst revealed benefit in terms of the clinical endpoints, with difference being smaller and plateauing at 30%

 

Commentary:

-- I have personally been checking both eGFRcys and eGFRcr on my patients for a couple of decades, finding that it was quite common that there was a substantial difference between these 2 measurements. This was especially true for patients who had either very increased or a very decreased muscle mass. However, I also found this discordance pretty often in other patients as well

-- a prior blog of mine based on a large Stockholm study (158601 participants in the SCREAM study, which was included in the current meta-analysis) who had both creatinine and cystatin levels measured found that those with lower eGFRcys vs eGFRcr had higher risk of all-cause deaths, ASCVD, heart failure, acute kidney injury, and kidney failure with replacement therapy: https://gmodestmedblogs.blogspot.com/2023/12/cystatin-c-better-predictor-of-bad.html . this study evaluated the absolute eGFRdiff, not the % change as was done in the above study. Of note, in the SCREAM study they assessed 2 levels of discordance, one with an eGFRdiff of -83 to -28 and the other at -27 to -10, finding evidence of increased adverse events in both groups though more so in those with the more severe discordances of -83 to -28

    -- so, the SCREAM study found that increasingly lower eGFR by cystatin versus creatinine was associated with increasing risk of these bad outcomes

    -- those with lower eGFR by creatinine had fewer events: they did not find that the eGFRcr added anything regarding clinical events

 

Of noted in the above blog on the Swedish study:

-- Sweden has been in the forefront of testing for cystatin C in the world, after the Swedish researcher Anders Grubb identified cystatin C as a filtration marker in 1985

-- in 2012, KDIGO (Kidney Disease: Improving Global Outcomes) recommended measuring cystatin C for either confirmatory testing of abnormal creatinine levels and in situations when creatinine is not accurate enough for clinical decision-making

-- creatinine levels increase (and estimated GFR decrease) with increased muscle mass, as well as with increased protein intake (which may increase muscle mass), exercise but also with the acute ingestion of cooked meats (https://diabetesjournals.org/care/article/37/2/483/29467/ ), fluid status, diurnal variation, and:

    --there are several meds reported in the medical literature that can increase serum creatinine levels though do not reflect renal injury: cimetidine, trimethoprim/sulfamethoxazole, fenofibrate, IV cephalosporins, salicylates, pyrimethamine, corticosteroids, flucytosine, vitamin D metabolites, and I suspect others….

-- cystatin C is a low-molecular weight protein that is filtered and not reabsorbed, and can be falsely elevated by chronic inflammation, obesity, smoking, glucocorticoid use, and hyper/hypothyroidism. Higher levels are also associated with risk of developing diabetes

-- despite the KDIGO recommendation and recommendations by the American Society of Nephrology in 2021 for the “increased, routine, and timely use of cystatin C”, cystatin C is underutilized in the US

-- the current meta-analysis found that the higher the percent that the eGFRcys was lower than the eGFRcr, the more adverse clinical events measured. And, as per their figure 2 printed above, this difference seems to have begun with a discrepancy of only 5-10%

-- the researchers also adjusted their results for the eGFRcr-cyst (not mentioned above), a calculation that includes both cystatin and creatinine levels and more accurately corresponds to the actually measured GFR, finding a significant but somewhat attenuated cardiovascular risk for the cardiovascular and all-cause mortality outcomes and an actual reversal of KFRT

    -- the researchers posit that this KFRT reversal "may be explained by usual clinical practice wherein dialysis initiation is largely determined by eGFRcr; therefore, adjustment for eGFRcr-cys becomes an artificial construct, with eGFRdiff serving as a proxy for eGFRcr in the model"

-- the equation for the eGFR measured by the combination of creatinine and cystatin (eGFRcr-cyst) requires just the values of creatinine and cystatin, the age, and the sex of the patient: for calculator see  https://www.kidney.org/professionals/gfr_calculator from the National Kidney Foundation

 

-- another study, though using creatinine-based eGFR, found that those with CKD stage 3-5 who had no history of hospitalization for heart failure but had routine BNP measurements (which reflects heart failure) found that BNP increases over time were associated with decreased KFRT by 56%, acute kidney injury by 64%, the composite of death or KFRT by 77%, and  subsequent heart failure hospitalizations by 63%: https://gmodestmedblogs.blogspot.com/2023/12/routine-bnp-assessment-helpful-for.html . This suggests that it may well be an advantage to measuring routine BNP levels in people with more advanced kidney failure and treating these elevated BNP levels appropriately. 

 

Limitations

-- as a meta-analysis, the researchers are combining many different studies into a composite, but each study is actually different from the others by their study designs, included/excluded individuals, outcomes measured (the researchers only included outcomes in the studies that measured them, and some of the results had many fewer studies for their data), demographics and medical/social comorbidities, etc.

    -- since major differences in creatinine and cystatin results, as noted above, are related to different underlying conditions, their lack of information on muscle mass, thyroid disorders, medications taken, exercise, diet, as well as more detailed information about those who were hospitalized also makes it difficult to be certain of the accuracy of their results

-- many of the studies in the meta-analysis used diagnosis codes, which may not be accurate

-- we cannot rule-out a selection bias: did those patients having cystatin measurement, for example, get that measurement for a specific reason, perhaps some comorbidities, and that reason might bias the results (the only way to eliminated selection bias is through randomized controlled studies)

-- there was also the elephants-in-the-room: of the 23 outpatient studies, 821327 participants were involved. the SCREAM trial had 151502 (18% of the total) and the UK Biobank had 467963 participants (57%). so, 75% of the total patients were from these 2 studies. and this high percentage will dwarf the results in the meta-analysis of the smaller and maybe even better conducted studies in the remaining 21 of them (25% of the total). Also, the huge UK Biobank study had a predominantly white population (>90%), and this could reflect very different social conditions from other groups, which could affect the summation of their results of the creatinine/cystatin measurements

-- this study relied on the US National Kidney Foundation eGFR calculator, which might have different results than the European Kidney Function Consortium equation, perhaps decreasing the generalizability of the results. And the labs measuring the creatinine and cystatin levels may have different results by using different methodologies

-- the researchers included the range of eGFRdiff from -30% to 30%, though we know from the SCREAM study above as well as their Figure 2 above that when eGFRcr is worse than eGFRcys (ie an eGFRdiff with lower creatinine calculation), there were fewer bad outcomes (the SCREAM study found no effect of this situation on outcomes: the lower the eGFRcys, the worse). it would have been more useful to assess and document results of the intervals individually, eg eGFRdiff -60 to -30, -30 to 0, 0 to 30, and 30-60

-- although I think that the evidence is quite clear that cystatin is a much better predictor of adverse clinical events than creatinine for most patients, one concern is that a lot of the clinical data we have, such as renal-dosing of medications, is based on creatinine. We really do need to have a more accurate and consistent approach to measuring renal function than one based solely on creatinine

So,

-- a broad meta-analysis documenting that the difference in bad outomes (mortality, cardiovascular, renal) is related to increasing percent differences in eGFR measured by cystatin and creatinine:

    -- these differences in these bad outcomes became manifest when the eGFR by cystatin is only 5-10% lower than that measured by creatinine, and the bad outcomes increasing with increasing eGFRdiff

        -- by the way, the mean eGFRcr was 87 and eGFRcys was 81. so, on average the eGFRcys was actually 7% lower than eGFRcr

    -- and this difference potentially has important implications for primary care clinicians (and others):

        -- we should routinely be testing creatinine and cystatin, and calculating the % difference in these calculated eGFRs (or perhaps just assessing the absolute difference, per the SCREAM study)

        -- if the eGFRcys is significantly lower than the eGFRcr, in the 5-10% or more range, it seems important to modify therapy some, such as:

            -- following the patient more closely for renal dysfunction, with perhaps assessing the cystatin and urinary albumin-to-creatinine ratio more often (perhaps every 6 months?)

            -- considering CKD (including high urinary albumin-to-creatinine ratio) as ASCVD risk factors and being more aggressive in addressing lipid therapy, smoking cessation, weight loss when indicated, diet, exercise, etc

            -- considering checking BNP for heart failure on a regular basis (perhaps every 3 months, as in the cited study above) in those with CKD stage 3 or worse

    -- and there really needs to be more research:

        -- we need good RCTs to see if these interventions mentioned here above are appropriate and associated with improved outcomes

        -- the drug companies should include cystatin in their "renal-dosing" for meds, if the meds might vary by renal function

        -- it would also be useful to know whether the percent change in eGFRcys vs eGFRcr most accurately predicts adverse clinical outcomes, or the absolute difference between these two, or the use of the creatinine/cystatin calculator (eGFRcr-cyst).

geoff

-----------------------------------

If you would like to be on the regular email list for upcoming blogs, please contact me at gmodest@bidmc.harvard.edu

to get access to all of the blogs:  go to http://gmodestmedblogs.blogspot.com/ to see the blogs in reverse chronological order

or you can just click on the magnifying glass on top right, then type in a name in the search box and get all the blogs with that name in them