How this instrument works
The 4-variable Kidney Failure Risk Equation (KFRE) is a Cox proportional-hazards model that predicts the probability of progressing to kidney failure, defined as needing dialysis or a transplant, within 2 or 5 years, from four inputs: age, sex, estimated glomerular filtration rate (eGFR), and urine albumin-to-creatinine ratio (ACR). It was derived by Tangri and colleagues from two large Canadian chronic kidney disease cohorts and published in JAMA in 2011, then recalibrated for use outside North America in a companion JAMA paper in 2016, which is where this calculator's Population toggle comes from.
This calculator only accepts eGFR values under 60 mL/min/1.73m^2, corresponding to CKD stages G3 through G5, because that is the population the KFRE was derived and validated on: the underlying cohorts enrolled patients already referred to nephrology with reduced kidney function, not a general population sample. eGFR should ideally come from the CKD-EPI 2009 creatinine equation, the estimate both source papers used; that's a scoping detail describing what the model was built on, not a hard requirement the calculator enforces. Many labs today instead report eGFR from the newer, race-free CKD-EPI 2021 equation, which runs a few mL/min/1.73m^2 higher on average than the 2009 version; a 2021-based eGFR is a reasonable substitute but not an exact match to the population the coefficients were fit on. ACR is entered in mg/g, the US laboratory convention; if a lab instead reports ACR in mg/mmol, multiply that value by 8.8401 to convert to mg/g before entering it here.
Population and Risk horizon both change which baseline survival constant, S0(t), gets plugged into the model; they do not change the four coefficients on age, sex, eGFR, or ACR. The 2016 multinational validation found that predicted risk using the original North American baseline systematically overestimated observed risk in cohorts outside North America, so it recalibrated S0(t) downward, about 32.9% lower at 2 years and 16.5% lower at 5 years, rather than re-deriving new coefficients. That is why this calculator carries four separate baseline-survival constants, North American and non-North American, each at 2 and 5 years, layered on top of one shared linear predictor.
This is a risk-prediction tool, not a treatment-effect calculator, and the 2016 multinational validation paper says so directly: it reports that there is no evidence using the equation improves patient outcomes, only that it predicts them accurately. A predicted risk percentage is meant to inform, not replace, a clinician's judgment about monitoring intensity, timing of vascular access planning, or a nephrology referral; it does not diagnose kidney failure, cannot see factors the model was never given such as the trend in eGFR decline over time, and should not be the sole basis for a treatment decision.
- Enter Age in years and set Sex is male to Yes or No; both feed the Cox linear predictor's age and sex terms.
- Enter eGFR in mL/min/1.73m^2 from a CKD-EPI 2009 creatinine estimate; only values under 60 are accepted, the CKD G3-G5 range the KFRE was built on.
- Enter Urine albumin-to-creatinine ratio, ACR, in mg/g; if your lab reports mg/mmol, multiply by 8.8401 first to convert.
- Choose Population, North American or non-North American, which selects the recalibrated baseline survival constant, not different coefficients.
- Choose Risk horizon, 2-year or 5-year, a second selector for that same baseline-survival constant.
- Read the Cox linear predictor, baseline survival, and the final predicted kidney failure risk percentage.
Worked example: 50-year-old man, CKD stage G4
A 50-year-old man (male = Yes) has eGFR 25 mL/min/1.73m^2 (CKD stage G4) and urine ACR 265.2 mg/g (about 30 mg/mmol), assessed with the North American, 2-year settings. The age term is −0.2201×(50/10−7.036) = 0.448, the sex term is 0.2467×(1−0.5642) = 0.108, the eGFR term is −0.5567×(25/5−7.222) = 1.237, and the ACR term is 0.4510×(ln(265.2)−5.137) = 0.200. Summing all four gives a linear predictor L of about 1.99.
With L about 1.99, exp(L) is about 7.33. The North American 2-year baseline survival S0 is 0.9750, so S0 raised to exp(L) works out to about 0.831, and the predicted kidney failure risk is (1 − 0.831) × 100, about 16.9%, matching this calculator's own computed value for these exact inputs. That figure lines up with McCudden et al. (PLoS ONE, 2018), a published KFRE variability study using the same age, sex, eGFR, and ACR, which reports a predicted risk of approximately 17%.
Questions
What does the 4-variable KFRE actually predict?
It predicts the probability that someone with reduced kidney function will progress to kidney failure, defined as starting dialysis or receiving a transplant, within either 2 or 5 years, calculated from age, sex, eGFR, and urine albumin-to-creatinine ratio (ACR) using a Cox proportional-hazards model published by Tangri and colleagues in JAMA in 2011 and recalibrated for international use in 2016.
Why does this calculator refuse eGFR values of 60 or above?
Because the KFRE was derived and validated on patients with CKD stages G3 through G5, eGFR under 60 mL/min/1.73m^2, who had already been referred to nephrology care. The source papers do not report how the model performs above that threshold, so extending it to eGFR 60 or above would be extrapolating beyond the population it was built on; this calculator's input check reflects that limitation rather than an arbitrary restriction.
My lab reports ACR in mg/mmol, how do I convert it?
Multiply the mg/mmol value by 8.8401 to get mg/g, the unit this calculator expects. For example, an ACR of 30 mg/mmol converts to 30 × 8.8401, about 265.2 mg/g, the exact ACR value used in this page's own worked example. Getting the unit right matters: entering a mg/mmol value directly into a mg/g field would understate the albuminuria term by roughly a factor of nine.
Does eGFR have to be calculated with the CKD-EPI 2009 equation?
Both the 2011 derivation and the 2016 recalibration used the CKD-EPI 2009 creatinine equation to compute eGFR, so that's the estimate this calculator's scope describes. It is a scoping detail rather than a value this calculator can enforce; the field only takes a number in mL/min/1.73m^2. Using a different eGFR estimate, such as an older MDRD-based value or the newer, race-free CKD-EPI 2021 equation most labs use today, will still run the equation, but introduces a mismatch with the population the coefficients were fit on.
What does the Population toggle change, and does it change the formula?
It changes only the baseline survival constant, S0(t), not the four coefficients on age, sex, eGFR, and ACR. The 2016 multinational validation found the original North American baseline overestimated risk in cohorts outside North America, so it recalibrated S0(t) downward for non-North American use, about 32.9% lower at 2 years and 16.5% lower at 5 years, rather than refitting the underlying model.
What's the difference between the 2-year and 5-year risk horizon?
Each horizon has its own baseline survival constant, S0(t), calibrated separately: 0.9750 (2-year) and 0.9240 (5-year) for North American populations, or 0.9832 and 0.9365 for non-North American populations. The same linear predictor is reused for both; only the baseline survival it's raised to a power against changes, so 5-year risk is always higher than 2-year risk for the same person, reflecting the longer follow-up window.
Does a high predicted risk mean I should start dialysis or see a nephrologist?
Not on its own. This calculator performs risk prediction only, and the 2016 multinational validation paper is explicit that it found no evidence that using the equation improves patient outcomes, only that it accurately predicts them. A high percentage is a signal worth discussing with a clinician about monitoring intensity or a nephrology referral, not a standalone trigger for dialysis, transplant workup, or any other treatment decision.
How was the 16.9% figure in the worked example checked?
The worked example's inputs, a 50-year-old man with eGFR 25, ACR 265.2 mg/g, North American population, 2-year horizon, run through the published KFRE formula and this calculator's engine to the same result, about 16.9%. That figure was independently cross-checked against McCudden et al. (PLoS ONE, 2018), a published KFRE variability study using the same age, sex, eGFR, and ACR, which reports a predicted risk of approximately 17%, consistent with normal rounding.
References
- Tangri N, et al. A predictive model for progression of CKD to kidney failure, JAMA. 2011;305(15):1553-1559
- Tangri N, et al. Multinational assessment of equations predicting kidney failure risk, JAMA. 2016;315(2):164-174
- Levey AS, et al. CKD-EPI 2009 creatinine equation for estimating GFR, Ann Intern Med. 2009;150(9):604-612
- McCudden C, et al. Individual patient variability with the KFRE in advanced CKD, PLoS ONE. 2018;13(6):e0198456
Read this first: This instrument computes a screening figure from population formulas — it is not a diagnosis, and it cannot see the whole picture a clinician can. Use it to inform a conversation, not to replace one.