Why did my eGFR improve by 10 points while losing 18 kg? Creatinine vs cystatin C

Eleven months, 18 kg down. My creatinine has fallen from 1.05 to 0.96 mg/dL and the reported eGFR has gone from 85 to 95. I would like to believe my kidneys have improved, and something in me does not believe it, because I have also lost visible muscle and my DEXA puts lean mass down about 4.5 kg.

My understanding is that creatinine comes from muscle. If I have less muscle, I make less creatinine, and my creatinine falls without my filtration changing at all. If that is right, then my apparently improved eGFR is an artefact of the very thing I am trying to avoid, which is a fairly perverse property for a monitoring test to have.

Questions: how large is that effect quantitatively — could 4.5 kg of lean mass really account for 10 eGFR points? Does cystatin C fix it, and what are its own failure modes? And is there a way to detect that this is happening from the numbers rather than from having had a DEXA?

You have it right and the magnitude fits. Creatinine-based eGFR is a filtration estimate that depends on a muscle-derived analyte, so in someone deliberately shrinking, it drifts optimistic. Cystatin C largely fixes this and has its own distinct failure modes.

The quantitative case

Creatinine is the spontaneous, non-enzymatic breakdown product of creatine and phosphocreatine, roughly 1.5 to 2% of the muscle creatine pool per day, and about 98% of body creatine sits in skeletal muscle. Whole-body creatinine production in an adult male is conventionally quoted around 20 to 25 mg per kilogram of body weight per day, but the more mechanistically honest statement is that it scales with skeletal muscle mass, not total mass.

Work your case. Suppose skeletal muscle mass was around 30 kg at baseline and your DEXA-measured 4.5 kg lean loss reflects roughly 3.5 to 4 kg of skeletal muscle, the balance being organ mass, connective tissue and water. That is a fall of about 12% in the creatinine-generating compartment. At steady-state filtration, plasma creatinine is proportional to production, so a 12% fall in production predicts a fall in creatinine of about 12%: 1.05 × 0.88 = 0.92 mg/dL.

Your observed value is 0.96, an 8.6% fall. So the lean-mass change alone predicts a slightly larger fall than you saw — meaning the arithmetic comfortably accounts for the whole thing, and there may in fact be a small true decline in filtration hiding underneath it.

Now push it through the equation to see how much eGFR that buys. Using CKD-EPI 2021 for a male aged 53 with Scr 0.96 mg/dL:

• Scr/κ = 0.96 ÷ 0.9 = 1.0667. Above 1, so the max term applies: ln(1.0667) = 0.06454; × (−1.200) = −0.07745; exp = 0.9255.
• 0.9938^53 = exp(53 × −0.006219) = exp(−0.32961) = 0.7192.
• eGFR = 142 × 0.9255 × 0.7192 = 142 × 0.6656 = 94.5 mL/min/1.73 m².

So yes: a 0.09 mg/dL fall in creatinine, entirely explicable by lost muscle, produces a 10-point rise in reported eGFR. Your kidneys have not been shown to have improved. They have been shown to be filtering the same or slightly less creatinine from a body that is making less of it.

Other things that move creatinine without touching GFR

Worth ruling out before attributing everything to muscle:

• Creatine monohydrate supplementation raises plasma creatinine by expanding the precursor pool. Stopping it lowers creatinine. Common in exactly the population doing resistance training to protect lean mass.
• Protein intake. High meat intake raises creatinine acutely; a large reduction in total intake lowers it. Given that appetite suppression is the point of the drug, this moves in the same direction as the muscle effect.
• Tubular secretion blockers. Trimethoprim and cimetidine raise creatinine without altering filtration, so they mask the effect you are describing.
• Amputation, prolonged immobility, neuromuscular disease, cirrhosis. All reduce creatinine generation, sometimes dramatically.

Cystatin C: what it fixes and what it does not

Cystatin C is a small protein produced at a relatively constant rate by all nucleated cells, freely filtered at the glomerulus and then almost completely reabsorbed and catabolised by the proximal tubule. It does not depend on muscle mass, diet or tubular secretion. That is exactly the property you need.

Its own confounders, which are real and different:

• Thyroid status. Raised in hyperthyroidism, lowered in hypothyroidism. Not trivial in a population where thyroid function and levothyroxine dosing are also in play.
• Corticosteroids raise it, dose-dependently.
• Systemic inflammation raises it, so it is unreliable during acute illness.
• Adiposity itself is associated with higher cystatin C at a given measured GFR, which means it may modestly overestimate impairment in obesity — and therefore may drift in the opposite direction to creatinine as you lose weight. Two markers drifting in opposite directions for non-renal reasons is awkward, and it is a reason to interpret both rather than to switch allegiance to one.
• Smoking raises it.

The practical answer is not "use cystatin C instead". It is to use both and compute the combined estimate. The creatinine-cystatin C equation is more accurate than either single-marker equation against measured GFR, and current guidance recommends confirming a creatinine-based estimate with cystatin C whenever there is reason to think creatinine is unreliable — which describes your situation precisely.

How to detect this from the numbers alone

The tell is discordance between the two estimates. If eGFR from creatinine substantially exceeds eGFR from cystatin C, the usual explanation is low muscle mass. If it falls substantially below, suspect high muscle mass, creatine supplementation, or a secretion blocker.

A gap of more than roughly 15 to 20% is generally treated as meaningful, and the direction is the diagnostic information. Such discordance has itself been associated with adverse outcomes in cohort studies, independently of the eGFR value — unsurprising, since a large gap usually marks sarcopenia or chronic illness.

Two other cheap tells that do not require cystatin C:

• Urea and creatinine moving apart. Falling creatinine with stable or rising urea is suspicious for reduced creatinine generation rather than improved filtration.
• Albuminuria. If the albumin:creatinine ratio is stable while eGFR "improves", that is at least consistent. If albuminuria is rising while creatinine-based eGFR improves, believe the albuminuria.

None of this is medical advice, and the genuinely useful action here is to bring the DEXA result and both eGFR estimates to a clinician together, because the interpretation depends on them being read against each other.

edited 31 Oct 2025 by Dr_Aoife_Brennan — tightened the wording; no substantive change

Worth adding what "eGFR" is estimating, because the word estimate is doing heavy lifting and the precision limits bound everything above.

Measured GFR requires an exogenous filtration marker — inulin classically, now iohexol, iothalamate or a chromium- or technetium-labelled agent — with timed plasma sampling. It is the reference method, it takes hours, and nobody does it for monitoring.

Estimating equations are regressions of measured GFR on an endogenous marker plus demographics. Their accuracy is conventionally summarised as P30: the proportion of estimates falling within 30% of measured GFR. For CKD-EPI 2021 with creatinine, P30 is roughly 85 to 90% in the development and validation populations. That sounds good until you translate it: one estimate in eight is more than 30% away from the truth, and at an eGFR of 85 a 30% error spans 60 to 110.

Combining creatinine and cystatin C improves P30 by several percentage points, which is the strongest single argument for the combined equation.

Three consequences that matter for how you read your own numbers:

• A single eGFR value is a wide interval, not a number. Treating 85 versus 95 as a meaningful difference on two draws ignores that either could be 20 units from the truth.
• Serial eGFR in one person is far better than the P30 suggests, because much of the individual error is a stable person-specific offset that cancels when you take differences. This is the same structure as the glycation gap in HbA1c. Your own trend is informative; the absolute value is fuzzy.
• The equations were developed in populations that under-represent extremes of body composition. Very high or very low muscle mass, and large recent changes in it, are exactly where they perform worst. That is your situation.

Practical implication: report and track the trend, know that the level is approximate, and stop treating crossings of the 90 or 60 boundaries as events. They are boundaries on an estimate whose uncertainty is wider than the distance between them.

One further note on the 2021 equations: the race coefficient present in the earlier CKD-EPI equations was removed, which shifted reported values for Black patients downwards relative to the old equation by a few units. If you have historical results spanning that change, part of any step you see is the equation, not you. Ask which equation each result used.

A short practical note on ordering cystatin C, since availability is the usual obstacle.

It is a routine immunoassay — usually particle-enhanced immunoturbidimetric or immunonephelometric — running on the same chemistry analysers as everything else, and it is standardised against an international certified reference material, so results are reasonably comparable between labs. It does not need a special tube, a fasting state or a specific time of day. Where it is not offered, that is a formulary and cost decision rather than a technical limitation, and the cost is typically a small multiple of a creatinine.

Two things to specify when requesting it:

• Ask for the combined creatinine-cystatin C eGFR to be reported, not just the raw cystatin C concentration. Many labs will report the concentration and leave the estimate uncalculated, which puts you in the position of needing the equation. It is available, but you also need to know which version the lab's own creatinine value is calibrated for.
• Ask for both markers on the same sample. Drawn weeks apart, the comparison is contaminated by whatever changed in between, which defeats the point of the discordance analysis.

On indications, guidance in this area has moved towards using cystatin C confirmation whenever a creatinine-based estimate is likely to be inaccurate or is being used for a consequential decision. The situations where that applies most clearly include extremes of muscle mass, rapid change in body composition, amputation, chronic liver disease, and drug dosing decisions where the estimate directly sets a dose. Substantial ongoing weight loss with documented lean-mass change is a defensible reason to request it, and framing the request that way is more likely to succeed than asking for it as an addition to a general panel.

If it is genuinely unavailable, the fallback is to track albuminuria and urea alongside creatinine and to interpret any creatinine improvement during weight loss as unproven rather than reassuring. That is not as good, and it is much better than reading the 95 at face value.