Antoine Créon
@antoinecreon.bsky.social
Diving into data with causal inference tools to make kidney health crystal clear. 🧠🔍
Nephrologist | PhD student at Karolinska Institutet
🌐 https://antoine-creon.github.io/
Nephrologist | PhD student at Karolinska Institutet
🌐 https://antoine-creon.github.io/
9/9
👏 Huge congratulations to lead author Will Russel, the SCREAM and EKFC team @delanaye.bsky.social
🧾 Full open access article in @ndt-era.bsky.social shorturl.at/T5Dui
👏 Huge congratulations to lead author Will Russel, the SCREAM and EKFC team @delanaye.bsky.social
🧾 Full open access article in @ndt-era.bsky.social shorturl.at/T5Dui
Changing from the CKD-EPI to the EKFC creatinine equation to estimate glomerular filtration rate in adults in a Northern European health system
AbstractBackground. The European Kidney Function Consortium (EKFC) 2021 equation to estimate GFR performs as well or better than the CKD-EPI 2009 equation
shorturl.at
August 12, 2025 at 1:36 PM
9/9
👏 Huge congratulations to lead author Will Russel, the SCREAM and EKFC team @delanaye.bsky.social
🧾 Full open access article in @ndt-era.bsky.social shorturl.at/T5Dui
👏 Huge congratulations to lead author Will Russel, the SCREAM and EKFC team @delanaye.bsky.social
🧾 Full open access article in @ndt-era.bsky.social shorturl.at/T5Dui
8/9
🧠Why this matters
Changing eGFR equations reshapes disease prevalence, referral patterns, and treatment thresholds — if we don't update those thresholds too. Implications for clinicians, health systems, and policy makers are substantial.
🧠Why this matters
Changing eGFR equations reshapes disease prevalence, referral patterns, and treatment thresholds — if we don't update those thresholds too. Implications for clinicians, health systems, and policy makers are substantial.
August 12, 2025 at 1:36 PM
8/9
🧠Why this matters
Changing eGFR equations reshapes disease prevalence, referral patterns, and treatment thresholds — if we don't update those thresholds too. Implications for clinicians, health systems, and policy makers are substantial.
🧠Why this matters
Changing eGFR equations reshapes disease prevalence, referral patterns, and treatment thresholds — if we don't update those thresholds too. Implications for clinicians, health systems, and policy makers are substantial.
7/9
Clinical impact if EKFC adopted
· +22% nephrology referrals
· +39% eligible for SGLT2 inhibitors
· +26% more heart failure patients with spironolactone contraindication
· More dose adjustments for anticoagulants & diabetes medications
Clinical impact if EKFC adopted
· +22% nephrology referrals
· +39% eligible for SGLT2 inhibitors
· +26% more heart failure patients with spironolactone contraindication
· More dose adjustments for anticoagulants & diabetes medications
August 12, 2025 at 1:36 PM
7/9
Clinical impact if EKFC adopted
· +22% nephrology referrals
· +39% eligible for SGLT2 inhibitors
· +26% more heart failure patients with spironolactone contraindication
· More dose adjustments for anticoagulants & diabetes medications
Clinical impact if EKFC adopted
· +22% nephrology referrals
· +39% eligible for SGLT2 inhibitors
· +26% more heart failure patients with spironolactone contraindication
· More dose adjustments for anticoagulants & diabetes medications
6/9
Counter-intuitively, people reclassified to a lower eGFR by EKFC had lower KFRT risk — likely because EKFC lowers eGFR more in older adults with stable, age-related decline rather than progressive disease, plus differences in equation modeling and calibration.
Counter-intuitively, people reclassified to a lower eGFR by EKFC had lower KFRT risk — likely because EKFC lowers eGFR more in older adults with stable, age-related decline rather than progressive disease, plus differences in equation modeling and calibration.
August 12, 2025 at 1:36 PM
6/9
Counter-intuitively, people reclassified to a lower eGFR by EKFC had lower KFRT risk — likely because EKFC lowers eGFR more in older adults with stable, age-related decline rather than progressive disease, plus differences in equation modeling and calibration.
Counter-intuitively, people reclassified to a lower eGFR by EKFC had lower KFRT risk — likely because EKFC lowers eGFR more in older adults with stable, age-related decline rather than progressive disease, plus differences in equation modeling and calibration.
5/9
Risk associations:
Both equations were similarly predictive for:
· KFRT
· All-cause mortality
· MACE
After adjusting for age:
· Death and MACE risks stayed similar across categories
Risk associations:
Both equations were similarly predictive for:
· KFRT
· All-cause mortality
· MACE
After adjusting for age:
· Death and MACE risks stayed similar across categories
August 12, 2025 at 1:36 PM
5/9
Risk associations:
Both equations were similarly predictive for:
· KFRT
· All-cause mortality
· MACE
After adjusting for age:
· Death and MACE risks stayed similar across categories
Risk associations:
Both equations were similarly predictive for:
· KFRT
· All-cause mortality
· MACE
After adjusting for age:
· Death and MACE risks stayed similar across categories
4/9
Main finding – eGFR shift
Median eGFR was 4.9 mL/min/1.73m² lower with EKFC.
➡️ CKD G3–G5 prevalence rose from 4.5% → 6.2% (+38%).
11% of people moved to a lower CKD category; <1% moved to a higher category.
Main finding – eGFR shift
Median eGFR was 4.9 mL/min/1.73m² lower with EKFC.
➡️ CKD G3–G5 prevalence rose from 4.5% → 6.2% (+38%).
11% of people moved to a lower CKD category; <1% moved to a higher category.
August 12, 2025 at 1:36 PM
4/9
Main finding – eGFR shift
Median eGFR was 4.9 mL/min/1.73m² lower with EKFC.
➡️ CKD G3–G5 prevalence rose from 4.5% → 6.2% (+38%).
11% of people moved to a lower CKD category; <1% moved to a higher category.
Main finding – eGFR shift
Median eGFR was 4.9 mL/min/1.73m² lower with EKFC.
➡️ CKD G3–G5 prevalence rose from 4.5% → 6.2% (+38%).
11% of people moved to a lower CKD category; <1% moved to a higher category.
3/9
📍 Setting: Stockholm, Sweden (2006–21)
👥 1,784,831 adults with serum creatinine in routine care
🔍 Aim: Compare CKD-EPI vs EKFC for:
· Reclassification across GFR categories
· Risk prediction for health outcomes
· Potential changes in referrals & drug eligibility
📍 Setting: Stockholm, Sweden (2006–21)
👥 1,784,831 adults with serum creatinine in routine care
🔍 Aim: Compare CKD-EPI vs EKFC for:
· Reclassification across GFR categories
· Risk prediction for health outcomes
· Potential changes in referrals & drug eligibility
August 12, 2025 at 1:36 PM
3/9
📍 Setting: Stockholm, Sweden (2006–21)
👥 1,784,831 adults with serum creatinine in routine care
🔍 Aim: Compare CKD-EPI vs EKFC for:
· Reclassification across GFR categories
· Risk prediction for health outcomes
· Potential changes in referrals & drug eligibility
📍 Setting: Stockholm, Sweden (2006–21)
👥 1,784,831 adults with serum creatinine in routine care
🔍 Aim: Compare CKD-EPI vs EKFC for:
· Reclassification across GFR categories
· Risk prediction for health outcomes
· Potential changes in referrals & drug eligibility
2/9 Background:
· eGFR guides diagnosis & staging of CKD.
· CKD-EPI 2009 is the most widely used equation in Europe.
· EKFC 2021 was designed for broader accuracy, especially in European populations.
· But… what would a real-world switch actually do?
· eGFR guides diagnosis & staging of CKD.
· CKD-EPI 2009 is the most widely used equation in Europe.
· EKFC 2021 was designed for broader accuracy, especially in European populations.
· But… what would a real-world switch actually do?
August 12, 2025 at 1:36 PM
2/9 Background:
· eGFR guides diagnosis & staging of CKD.
· CKD-EPI 2009 is the most widely used equation in Europe.
· EKFC 2021 was designed for broader accuracy, especially in European populations.
· But… what would a real-world switch actually do?
· eGFR guides diagnosis & staging of CKD.
· CKD-EPI 2009 is the most widely used equation in Europe.
· EKFC 2021 was designed for broader accuracy, especially in European populations.
· But… what would a real-world switch actually do?
9/9
Congrats to first author Aurora Caldinelli and the team of co-authors from @ki.se
Congrats to first author Aurora Caldinelli and the team of co-authors from @ki.se
https://shorturl.at/0CzNq
t.co
July 14, 2025 at 12:06 PM
9/9
Congrats to first author Aurora Caldinelli and the team of co-authors from @ki.se
Congrats to first author Aurora Caldinelli and the team of co-authors from @ki.se
8/9
🎯 Take-home
Switching from traditional eGFR/ACR-based to KFRE-based referral:
✅ Sharpens referral precision
✅ Cuts unnecessary specialist visits by ~25%
✅ Preserves safety with few missed cases
✅ Aligns with KDIGO 2024 — but suggests higher, locally optimized thresholds
🎯 Take-home
Switching from traditional eGFR/ACR-based to KFRE-based referral:
✅ Sharpens referral precision
✅ Cuts unnecessary specialist visits by ~25%
✅ Preserves safety with few missed cases
✅ Aligns with KDIGO 2024 — but suggests higher, locally optimized thresholds
July 14, 2025 at 12:06 PM
8/9
🎯 Take-home
Switching from traditional eGFR/ACR-based to KFRE-based referral:
✅ Sharpens referral precision
✅ Cuts unnecessary specialist visits by ~25%
✅ Preserves safety with few missed cases
✅ Aligns with KDIGO 2024 — but suggests higher, locally optimized thresholds
🎯 Take-home
Switching from traditional eGFR/ACR-based to KFRE-based referral:
✅ Sharpens referral precision
✅ Cuts unnecessary specialist visits by ~25%
✅ Preserves safety with few missed cases
✅ Aligns with KDIGO 2024 — but suggests higher, locally optimized thresholds
7/9
🚨 Important nuance:
Even with higher thresholds, the models missed very few patients who progressed to kidney failure.
False negatives remained under 0.4%.
So we’re not sacrificing safety — just optimizing how we use specialist care.
🚨 Important nuance:
Even with higher thresholds, the models missed very few patients who progressed to kidney failure.
False negatives remained under 0.4%.
So we’re not sacrificing safety — just optimizing how we use specialist care.
July 14, 2025 at 12:06 PM
7/9
🚨 Important nuance:
Even with higher thresholds, the models missed very few patients who progressed to kidney failure.
False negatives remained under 0.4%.
So we’re not sacrificing safety — just optimizing how we use specialist care.
🚨 Important nuance:
Even with higher thresholds, the models missed very few patients who progressed to kidney failure.
False negatives remained under 0.4%.
So we’re not sacrificing safety — just optimizing how we use specialist care.
6/9
❗️But what about the risk thresholds?
The widely cited 3–5% risk cutoffs might be too low for most primary care patients.
🔹 We found that 15% (original KFRE) and 9% (Sweden-recalibrated KFRE) gave the best balance of sensitivity and specificity.
❗️But what about the risk thresholds?
The widely cited 3–5% risk cutoffs might be too low for most primary care patients.
🔹 We found that 15% (original KFRE) and 9% (Sweden-recalibrated KFRE) gave the best balance of sensitivity and specificity.
July 14, 2025 at 12:06 PM
6/9
❗️But what about the risk thresholds?
The widely cited 3–5% risk cutoffs might be too low for most primary care patients.
🔹 We found that 15% (original KFRE) and 9% (Sweden-recalibrated KFRE) gave the best balance of sensitivity and specificity.
❗️But what about the risk thresholds?
The widely cited 3–5% risk cutoffs might be too low for most primary care patients.
🔹 We found that 15% (original KFRE) and 9% (Sweden-recalibrated KFRE) gave the best balance of sensitivity and specificity.
5/9
💡 Key finding:
KFRE-based referral models had better sensitivity and specificity than traditional criteria.
They also improved:
✅ Net Reclassification Index
✅ Decision curve analysis
✅ Positive predictive value (PPV)
And they reduced unnecessary referrals by ~25%.
💡 Key finding:
KFRE-based referral models had better sensitivity and specificity than traditional criteria.
They also improved:
✅ Net Reclassification Index
✅ Decision curve analysis
✅ Positive predictive value (PPV)
And they reduced unnecessary referrals by ~25%.
July 14, 2025 at 12:06 PM
5/9
💡 Key finding:
KFRE-based referral models had better sensitivity and specificity than traditional criteria.
They also improved:
✅ Net Reclassification Index
✅ Decision curve analysis
✅ Positive predictive value (PPV)
And they reduced unnecessary referrals by ~25%.
💡 Key finding:
KFRE-based referral models had better sensitivity and specificity than traditional criteria.
They also improved:
✅ Net Reclassification Index
✅ Decision curve analysis
✅ Positive predictive value (PPV)
And they reduced unnecessary referrals by ~25%.
4/9
We compared 3 referral strategies:
1️⃣ Swedish criteria (age, eGFR, albuminuria)
2️⃣ KDIGO 2012 criteria
3️⃣ KFRE-based model, using the 4-variable equation (age, sex, eGFR, ACR)
KFRE was tested both in its original (non-North American) and recalibrated version for Sweden.
We compared 3 referral strategies:
1️⃣ Swedish criteria (age, eGFR, albuminuria)
2️⃣ KDIGO 2012 criteria
3️⃣ KFRE-based model, using the 4-variable equation (age, sex, eGFR, ACR)
KFRE was tested both in its original (non-North American) and recalibrated version for Sweden.
July 14, 2025 at 12:06 PM
4/9
We compared 3 referral strategies:
1️⃣ Swedish criteria (age, eGFR, albuminuria)
2️⃣ KDIGO 2012 criteria
3️⃣ KFRE-based model, using the 4-variable equation (age, sex, eGFR, ACR)
KFRE was tested both in its original (non-North American) and recalibrated version for Sweden.
We compared 3 referral strategies:
1️⃣ Swedish criteria (age, eGFR, albuminuria)
2️⃣ KDIGO 2012 criteria
3️⃣ KFRE-based model, using the 4-variable equation (age, sex, eGFR, ACR)
KFRE was tested both in its original (non-North American) and recalibrated version for Sweden.
3/9
📊 Data:
192,964 adults with eGFR <60 ml/min/1.73m² and an albuminuria test, from 2006–2021.
🔁 887,388 repeated observations
👴 Median age: 76
📉 Median eGFR: 54
🎯 Outcome: initiation of kidney replacement therapy (KRT) within 5 years
📊 Data:
192,964 adults with eGFR <60 ml/min/1.73m² and an albuminuria test, from 2006–2021.
🔁 887,388 repeated observations
👴 Median age: 76
📉 Median eGFR: 54
🎯 Outcome: initiation of kidney replacement therapy (KRT) within 5 years
July 14, 2025 at 12:06 PM
3/9
📊 Data:
192,964 adults with eGFR <60 ml/min/1.73m² and an albuminuria test, from 2006–2021.
🔁 887,388 repeated observations
👴 Median age: 76
📉 Median eGFR: 54
🎯 Outcome: initiation of kidney replacement therapy (KRT) within 5 years
📊 Data:
192,964 adults with eGFR <60 ml/min/1.73m² and an albuminuria test, from 2006–2021.
🔁 887,388 repeated observations
👴 Median age: 76
📉 Median eGFR: 54
🎯 Outcome: initiation of kidney replacement therapy (KRT) within 5 years
2/9
The 2024 @kdigo.org guidelines recommend using the KFRE risk equation to guide referrals when the 5-year risk of kidney failure exceeds 3–5%.
But:
✅ Is that threshold right?
✅ Does KFRE actually outperform traditional rules?
We tested this in >190,000 patients from Stockholm.
The 2024 @kdigo.org guidelines recommend using the KFRE risk equation to guide referrals when the 5-year risk of kidney failure exceeds 3–5%.
But:
✅ Is that threshold right?
✅ Does KFRE actually outperform traditional rules?
We tested this in >190,000 patients from Stockholm.
July 14, 2025 at 12:06 PM
2/9
The 2024 @kdigo.org guidelines recommend using the KFRE risk equation to guide referrals when the 5-year risk of kidney failure exceeds 3–5%.
But:
✅ Is that threshold right?
✅ Does KFRE actually outperform traditional rules?
We tested this in >190,000 patients from Stockholm.
The 2024 @kdigo.org guidelines recommend using the KFRE risk equation to guide referrals when the 5-year risk of kidney failure exceeds 3–5%.
But:
✅ Is that threshold right?
✅ Does KFRE actually outperform traditional rules?
We tested this in >190,000 patients from Stockholm.
6/6
👏 Huge credit to first author Will Russel for completing this work, and to the dream team led by JJ Carrero
Proud to share this effort!
🔗 tinyurl.com/yx3m67yk
#Nephrology #eGFR #CKD #ClinicalPharmacology #BMI
👏 Huge credit to first author Will Russel for completing this work, and to the dream team led by JJ Carrero
Proud to share this effort!
🔗 tinyurl.com/yx3m67yk
#Nephrology #eGFR #CKD #ClinicalPharmacology #BMI
Obesity, Underweight, and Accuracy of eGFR Using Cystatin C ... : Journal of the American Society of Nephrology
,503 repeated observations) referred for measured GFR (mGFR) in Stockholm, Sweden. We calculated indexed eGFR (in mL/min/1.73 m2) and non-indexed eGFR (in mL/min) with validated equations that use cre...
tinyurl.com
July 4, 2025 at 2:24 PM
6/6
👏 Huge credit to first author Will Russel for completing this work, and to the dream team led by JJ Carrero
Proud to share this effort!
🔗 tinyurl.com/yx3m67yk
#Nephrology #eGFR #CKD #ClinicalPharmacology #BMI
👏 Huge credit to first author Will Russel for completing this work, and to the dream team led by JJ Carrero
Proud to share this effort!
🔗 tinyurl.com/yx3m67yk
#Nephrology #eGFR #CKD #ClinicalPharmacology #BMI
5/6
📣 Take-home:
✅ Use eGFRcr-cys when estimating GFR in patients with underweight or obesity
✅ Consider non-indexed eGFR for precise dosing decisions
✅ Rethink reliance on creatinine alone in extremes of BMI
📣 Take-home:
✅ Use eGFRcr-cys when estimating GFR in patients with underweight or obesity
✅ Consider non-indexed eGFR for precise dosing decisions
✅ Rethink reliance on creatinine alone in extremes of BMI
July 4, 2025 at 2:24 PM
5/6
📣 Take-home:
✅ Use eGFRcr-cys when estimating GFR in patients with underweight or obesity
✅ Consider non-indexed eGFR for precise dosing decisions
✅ Rethink reliance on creatinine alone in extremes of BMI
📣 Take-home:
✅ Use eGFRcr-cys when estimating GFR in patients with underweight or obesity
✅ Consider non-indexed eGFR for precise dosing decisions
✅ Rethink reliance on creatinine alone in extremes of BMI
4/6
💊 Clinical impact:
Using eGFRcr-cys improved correct dosing and eligibility for drugs like carboplatin, apixaban, and SGLT2 inhibitors—especially in underweight and obese groups.
Non-indexed values offered small added gains, particularly for chemo dosing.
💊 Clinical impact:
Using eGFRcr-cys improved correct dosing and eligibility for drugs like carboplatin, apixaban, and SGLT2 inhibitors—especially in underweight and obese groups.
Non-indexed values offered small added gains, particularly for chemo dosing.
July 4, 2025 at 2:24 PM
4/6
💊 Clinical impact:
Using eGFRcr-cys improved correct dosing and eligibility for drugs like carboplatin, apixaban, and SGLT2 inhibitors—especially in underweight and obese groups.
Non-indexed values offered small added gains, particularly for chemo dosing.
💊 Clinical impact:
Using eGFRcr-cys improved correct dosing and eligibility for drugs like carboplatin, apixaban, and SGLT2 inhibitors—especially in underweight and obese groups.
Non-indexed values offered small added gains, particularly for chemo dosing.
3/6
🔍 What we found:
– eGFR based on creatinine alone overestimated kidney function at both BMI extremes
– eGFR based on cystatin C alone underestimated GFR in obesity
– The combo (eGFRcr-cys) had least bias and highest accuracy across all BMIs
🔍 What we found:
– eGFR based on creatinine alone overestimated kidney function at both BMI extremes
– eGFR based on cystatin C alone underestimated GFR in obesity
– The combo (eGFRcr-cys) had least bias and highest accuracy across all BMIs
July 4, 2025 at 2:24 PM
3/6
🔍 What we found:
– eGFR based on creatinine alone overestimated kidney function at both BMI extremes
– eGFR based on cystatin C alone underestimated GFR in obesity
– The combo (eGFRcr-cys) had least bias and highest accuracy across all BMIs
🔍 What we found:
– eGFR based on creatinine alone overestimated kidney function at both BMI extremes
– eGFR based on cystatin C alone underestimated GFR in obesity
– The combo (eGFRcr-cys) had least bias and highest accuracy across all BMIs
2/6
⚙️ What we did:
We analyzed 7,503 measured GFR (mGFR) tests from over 4,700 adults in Sweden. We compared multiple eGFR equations (creatinine, cystatin C, both) across BMI from <18 to >40 kg/m².
Both indexed and non-indexed formulas were evaluated.
⚙️ What we did:
We analyzed 7,503 measured GFR (mGFR) tests from over 4,700 adults in Sweden. We compared multiple eGFR equations (creatinine, cystatin C, both) across BMI from <18 to >40 kg/m².
Both indexed and non-indexed formulas were evaluated.
July 4, 2025 at 2:24 PM
2/6
⚙️ What we did:
We analyzed 7,503 measured GFR (mGFR) tests from over 4,700 adults in Sweden. We compared multiple eGFR equations (creatinine, cystatin C, both) across BMI from <18 to >40 kg/m².
Both indexed and non-indexed formulas were evaluated.
⚙️ What we did:
We analyzed 7,503 measured GFR (mGFR) tests from over 4,700 adults in Sweden. We compared multiple eGFR equations (creatinine, cystatin C, both) across BMI from <18 to >40 kg/m².
Both indexed and non-indexed formulas were evaluated.