eGFR: not such a great estimate of renal function

 An important article was just published finding that on an individual level, the calculated eGFR can vary pretty dramatically from the measured one, potentially affecting issues ranging from the validity of eGFR for renal dosing of medications, referrals to renal clinic, anxiety in patients (and providers), and perhaps invasive procedures (such as AV fistula surgery, kidney biopsies, etc), see ckd eGFR not so accurate AIM2022​ in dropbox, or doi:10.7326/M22-06

 

Details: 

-- a cross-sectional study of 4 US community-based epidemiologic cohort studies assessing both measured GFR (mGFR) and estimated GFR (eGFR), including 3223 participants

    -- the studies included GENOA (Genetic Epidemiology Network of Arteriopathy), ECAC (Epidemiology of Coronary Artery Calcification), ALTOLD (Assessing Long Term Outcomes in Living Kidney Donors), and CRIC (Chronic Renal Insufficiencies Cohort). The first three of these had participants who were not selected based on known chronic kidney disease (CKD)

    -- mGFR was measured directly using urinary clearance of iothalamate (some studies used radiolabeled, others not), or plasma clearance of iohexol

    -- all studies had measurements of serum creatinine and cystatin-C; eGFR was calculated using the Chronic Kidney Disease Epidemiology (CKD-EPI) race-free equation including calculation based on creatinine, cystatin-C and the combination equation including both creatinine and serum cystatin-C, and the European Kidney Function Consortium) equation 

-- mean age 59, 58% white/32% Black, 55% women 

--  mean mGFR was 68 (9-208), with mGFR breakdown: 

    -- >120: 89 individuals (3%) 

    -- 90-119: 675 (21%)

    -- 60-89: 1160 (36%) 

    -- 45-59: 537 (17%) 

    -- 30- 45:477 (15%) 

    -- 15 to 29:267 (8%) 

    -- <15: 18 (1%) 

-- mean eGFR [note: the parentheses for all of these GFR values include minimum to max, not confidence intervals]:  

    -- CKD-EPI_Creatinine : 69 (11-134)

    -- CKD-EPICreatinine-Cystatin C  :74 (12-164)

    -- CKD-EPI _{Cystatin C} 75 (13-184) 

-- comorbidities: CVD 16%, diabetes 31%, hypertension 70%, current smokers 11%, BMI 31, mean microalbumin 51 mg/g, mean serum creatinine 1.3 (0.4-4.8), mean serum cystatin-C 1.1 (0.2-3.8) 

-- Primary analysis: comparing the individual-level differences between mGFR and eGFR_CR (estimated GFR based on creatinine)

-- secondary analysis: comparing the individual-level differences between mGFR and both eGFR_CR-CS (eGFR based on creatinine and cystatin-C) and eGFRCS

 

Results:

-- population-based differences between mGFR and eGFR_CR were small: -0.6 (-1.2 to -0.2)

    -- mean eGFR calculations were very similar whether based on creatinine or cystatin-C, or the combo

 

-- individual level differences between eGFR_CR and mGFR:

    -- if eGFR_CR = 60:

        -- mGFR:

            -- 50% ranged from 52-67

            -- 80% from 45-76

            -- 95% from 36-87

    -- if eGFR_CR = 30:

        -- mGFR:

            -- 50% ranged from 27-38

            -- 80% from 23-44

            -- 95% from 17-54

 

This graph reveals the broad mGFR spread by eGFR: people with eGFR within the normal range (>90) could have mGFR at the level of those in the stage 3a group (45-60); those with eGFR stage 3a could overlap those with mGFR stage 4 (15-29).  Really big individual differences…

 

    -- If eGFR_CR = 45-59 (ie stage 3a CKD):

        -- 36% had mGFR>60 (stage 1 or2 ) and 20% <45 (stage 3b)

        -- 57% had mGFR outside the 40-50 range!!

    -- If eGFR_CR =  15 to 29 (ie stage 3a CKD):

        -- 30% had mGFR>30 (stage 3b or higher) and 20% <15 (stage 3b)

 

This graph shows that 50% of people with pretty much each CKD stage by eGFR  (and somewhat more so in those with lower eGFR) had at least a 15% probability of the mGFR exceeding the stage thresholds of 15 (stage 4), 30 (stage 3b), 45 (stage 3a), 60 (stage 2) and 90 (stage 1)

 

-- eGFR based on cystatin-C  did not provide substantial improvement: there was a somewhat narrower range of prediction intervals with  eGFR_CS than with either of the other 2 eGFRs, but no substantial differences

-- no substantial difference using the European Kidney Function Consortium eGFR equation

 

Commentary: 

-- the big issue for us clinicians is that we only get eGFR calculations, which are based on a few inputs: for the CKD-EPI formula, these include serum creatinine, cystatin-C if available, age, sex, and  the option to adjust for body surface area (see https://www.kidney.org/professionals/kdoqi/gfr_calculator which generates eGFR results by creatinine, cystatin-C, and the combo. So, not lots of variables used in the calculations. And it is important for us to realize that the gold standard is the measured GFR which involves injecting a filtration marker and measuring plasma or urinary clearance by serial blood and urine sampling.

-- As mentioned above, the value of the actual GFR is really critical for clinical care, especially for the administration of medications that need to be renally-dosed, prognostication of the deterioration of renal function and need for renal replacement therapy (dialysis, transplant), and preparing the patient emotionally/socially for their likely clinical course

    -- the study above found both underestimations and overestimations of eGFR: the former might lead to patients being denied important meds (eg, not giving a clinically useful med to patients where there is a contraindication in those having an eGFR <30 for example), the latter leading to perhaps toxic doses of prescribing of meds such as chemotherapy dosed by eGFR

-- there was a large divergence between the population-based differences between eGFR and mGFR and the individual-level ones: the former were negligible, suggesting that eGFR was a useful marker of mGFR for large groups. But, on an individual level, there was a huge difference in the correlation between mGFR and the eGFR.  This study found that for the CKD stages of mGFR, there was very wide variability for the eGFR_CR levels, often crossing 1 or more CKD stages.

-- also, there is the issue of race-based eGFR. Several articles have devised methodology for eliminating race-based calculations (see http://gmodestmedblogs.blogspot.com/2021/10/eliminating-race-in-calculating-renal.html ).  Of note, in the above study, they found that the population-level racial  differences were really small (<5 in GFR), as compared to the individual-level differences (95% prediction interval of GFR of  >50). This reinforces the importance of developing accurate actual measurements of GFR in order to eliminate the large variance of estimated GFR as now done, measurements which do not include race, age, socioeconomic factors, muscle mass, etc (serum creatinine is influenced by muscle mass, cooked meat intake, fasting status; cystatin-C by obesity)

-- I do have some concerns about the fact that cystatin-C measured eGFRs were no better than creatinine ones in this study. I suspect this has to do with differences in their populations versus what I am seeing (many of my patients being older than in this study, and with more severe CKD to begin with; the population in this study had a  mean age of 59, and the mGFR was largely mild renal impairment). I have found substantial differences between cystatin-C and creatinine, especially in two groups: elderly without much muscle mass, with cases of creatinine-based eGFR calculations in the stage 3a range who have stage 5 by cystatin-C; and alternatively some with very developed muscles out who have creatinine-based stage 3a but normal renal function by cystatin-C  (see prior blog http://gmodestmedblogs.blogspot.com/2022/06/chronic-kidney-disease-review.html )

-- at this point we do not know the stability of the eGFRs over time: if we measure an eGFR on a patient (which we now know may not be so accurate), and we know that the eGFR (by creatinine, cystatin-C, or the combo) is stable relative to mGFR and reproducible over time, then at least we can be comfortable about knowing the progression of disease. This information, I think, would require a baseline mGFR, the array of calculated eGFRs, and followup over time to assess the correlation. And, would be useful to also check changes in the items that affect eGFR, such as muscle mass, cooked meats, fasting for creatinine and obesity for cystatin-C

 

Limitations:

-- these calculations and measurements were based on a single set of values, without short-term confirmation or correlation over time

-- it is hard to draw conclusions for those with mGFR<30 given the small numbers of patients

-- this study combined 3 studies with very different patient populations and GFR distributions (eg most patients in the GENOA and ECAC had eGFR 60-89 (stage 2); most in ALTOLD had 90-119 (stage 1)); most in CRIC had GFR 30-44 (stage 3b). so, mathematically combining them is a bit fraught

-- in assessing differences in CKD stages between the mGFR and eGFR, those people close to the cutpoints of the GFR stage differences are much more likely to cross a CKD stage threshold with minor GFR changes than those mid-range in the CKD categories; clumping them all together may lead to many bumping into another stage with potentially only minimal GFR change

 

So, this is pretty profound. I am not exactly sure how to interpret these findings in light of the fact that we cannot get mGFR in clinical practice:

-- we should not treat the eGFR calculation (a very exact sounding number) as if it were more than an approximation of the mGFR. This is particularly a problem with prescribing renally-dosed meds, patient/provider anxiety about the state of the patient’s kidney function, preparation for renal replacement therapy, and there may well be unnecessary evaluations/testing for renal disease (including potentially invasive renal biopsies)

-- it probably is reasonable to assume for now that changes in eGFR over time are likely stable and predict pretty accurately the changes in renal function (though would be important to know this for sure, as per above), and proceed clinically based on the eGFR changes. though this might be problematic if the patient changes significantly (eg weight, cooked meat consumption)

-- it would be great to have a way of assessing mGFR in the community: there have been advances in developing a simple and feasible outpatient measurement of GFR using nonisotopic/nonradiolabeled methods (see https://academic.oup.com/ckj/article/13/3/397/5537357 ). this study involved injection of iohexol and subsequently measuring plasma levels 1-2 hours later. Perhaps further research into this and novel approaches will lead to a simpler, accurate, and generalizable method. And perhaps doing this mGFR a single time, and assessing its correlation with creatinine or cystatin-C,  would allow us to continue using creatinine or cystatin-C to assess changes in GFR?? More development and studies are needed....

geoff

 

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

 

to get access to all of the blogs (2 options):

1. go to http://gmodestmedblogs.blogspot.com/ to see them in reverse chronological order

2. click on 3 parallel lines top left, if you want to see blogs by category, then click on "labels" and choose a category​

3. 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

 

or: go to https://www.bucommunitymedicine.org/ , a website from the Community Medicine section at Boston Medical Center.  This site does have a very searchable and accessible list of my blogs (though there have been a few that did not upload over the last year or two). but overall it is much easier to view blogs and displays more at a time.

 

 

please feel free to circulate this to others. also, if you send me their emails, i can add them to the list