GLP-1 RAs and Kidney Cancer Risk- Methodological Considerations in Active-Compara
Table of Contents
- GLP-1 RAs and Kidney Cancer Risk: Does Study Design Drive the Signal?
- Abstract
- Study at a Glance
- Study Snapshot
- Study Facts Table
- What Researchers Actually Did
- Key Findings: Primary Outcomes
- Key Findings: Secondary Outcomes and Subgroup Analyses
- Results: Adverse Events and Safety Profile
- Statistical Approach and Rigor
- Clinical Takeaway
- Read This Paper Through Nine Different Lenses
- What is the main finding of the study regarding GLP-1 RAs and kidney cancer risk?
- How does comparator selection affect the results in studies on GLP-1 RAs and kidney cancer?
- What is target trial emulation and why was it used in this study?
- How does diabetes duration affect the relationship between GLP-1 RAs and kidney cancer risk?
- What does the divergence between HR and RR indicate in this study?
- Why was metformin used as a comparator in this study?
- What are the limitations of using an intention-to-treat (ITT) approach in this study?
- How does the study address potential biases related to treatment sequencing?
- What is the significance of the absolute risk difference (RD) in this study?
- What future research is needed based on this study’s findings?
- Read next
GLP-1 RAs and Kidney Cancer Risk: Does Study Design Drive the Signal?
What You’ll Learn
- Why three prior studies found elevated kidney cancer risk with GLP-1 RAs, and why those findings may be methodological artifacts rather than biological signals g
- How comparator selection, metformin vs. SGLT2 inhibitors vs. non-users, dramatically changes the apparent risk estimate
- What a properly emulated target trial framework shows when GLP-1 RAs are compared against a clinically matched active comparator
- Where the evidence remains genuinely uncertain and what future research must resolve
TL;DR: When GLP-1 receptor agonists are compared against SGLT2 inhibitors, a clinically appropriate active comparator, using a target trial emulation framework, the previously reported association between GLP-1 RA use and increased kidney cancer risk disappears entirely (HR 1.05, 95% CI 0.62–1.77).
Want to apply this research to your care?
CED Clinic translates emerging research into individualized clinical care. Dr. Caplan has treated 30,000+ patients.
Book a consultation →Abstract
The International Agency for Cancer Research has identified kidney cancer among 13 obesity-related cancers. Two of three recent studies reported that GLP-1 receptor agonist (GLP-1 RA) treatment was associated with lower risk of obesity-related cancers overall, yet those same studies reported elevated kidney cancer risk with GLP-1 RA use compared with either SGLT2 inhibitors (HR 1.16, 95% CI 1.00–1.35), metformin (HR 1.54, 95% CI 1.27–1.87), or non-users (HR 1.38, 95% CI 0.99–1.93). Lopez et al. conducted a retrospective cohort study using TriNetX (January 2006 to October 2024) in adults with type 2 diabetes and overweight or obesity, applying a target trial emulation framework with propensity score matching. When GLP-1 RA initiation was compared with SGLT2 inhibitor initiation, no association with kidney cancer was observed (HR 1.05, 95% CI 0.62–1.77; RD 0.0%). Comparison with metformin reproduced elevated HRs (HR 1.90, 95% CI 1.39–2.60), but this association was substantially attenuated among individuals with fewer than four years of type 2 diabetes duration (HR 1.26, 95% CI 0.77–2.05), a subgroup in which treatment groups are more clinically comparable. These findings suggest that previously reported kidney cancer signals reflect comparator-related confounding and treatment-sequencing bias rather than a causal carcinogenic effect of GLP-1 RAs.
DOI: https://doi.org/10.1111/dom.70700
PubMed: Not yet indexed at time of publication
Published: 2026 | Accepted: March 12, 2026
Study at a Glance
| Parameter | Detail |
|---|---|
| Design | Retrospective cohort study with target trial emulation framework; propensity score matching 1:1 nearest-neighbour |
| Database | TriNetX global EHR database, January 1, 2006 to October 31, 2024 |
| Population | Adults with type 2 diabetes and overweight or obesity |
| N (GLP-1 RA vs. SGLT2i arm) | 63,789 eligible; 26,670 matched (13,335 per group) |
| N (GLP-1 RA vs. metformin arm) | 266,172 eligible; 102,490 matched (51,245 per group) |
| Primary Endpoint | Incident kidney cancer (ICD-10 C64) |
| Key Finding (GLP-1 RA vs. SGLT2i) | HR 1.05 (95% CI 0.62–1.77); RD 0.0%: no association |
| Key Finding (GLP-1 RA vs. Metformin) | HR 1.90 (95% CI 1.39–2.60) overall; attenuated to HR 1.26 (95% CI 0.77–2.05) among patients with T2D duration <4 years |
Study Snapshot
| Comparison | Events (GLP-1 vs. Comparator) | RD (95% CI) | RR (95% CI) | HR (95% CI) |
|---|---|---|---|---|
| GLP-1 RA vs. SGLT2i (Overall) | 30 vs. 27 | 0.0% (−0.1, 0.1) | 1.11 (0.66, 1.87) | 1.05 (0.62, 1.77) |
| GLP-1 RA vs. SGLT2i (1-year) | 12 vs. 21 | −0.1% (−0.2, 0.0) | 0.57 (0.28, 1.16) | 0.54 (0.27, 1.11) |
| GLP-1 RA vs. SGLT2i (3-year) | 27 vs. 30 | 0.0% (−0.1, 0.1) | 0.90 (0.54, 1.51) | 0.85 (0.51, 1.43) |
| GLP-1 RA vs. SGLT2i (5-year) | 28 vs. 30 | 0.0% (−0.1, 0.1) | 0.93 (0.56, 1.56) | 0.88 (0.53, 1.48) |
| GLP-1 RA vs. Metformin (Overall) | 95 vs. 80 | 0.0% (0.0, 0.1) | 1.19 (0.88, 1.59) | 1.90 (1.39, 2.60) |
| GLP-1 RA vs. Metformin (3-year) | 85 vs. 46 | 0.1% (0.0, 0.1) | 1.85 (1.29, 2.65) | 2.09 (1.45, 2.99) |
| GLP-1 RA vs. Metformin (T2D <4 years) | 32 vs. 43 | 0.0% (−0.1, 0.0) | 0.74 (0.47, 1.18) | 1.26 (0.77, 2.05) |
| GLP-1 RA vs. Metformin (1–3 years follow-up) | 41 vs. 19 | 0.0% (0.0, 0.1) | 2.16 (1.25, 3.72) | 2.86 (1.65, 4.94) |
Study Facts Table
| Full Study Profile | |
|---|---|
| Authors | Lopez DS, Abdelgadir O, Hernández-Pérez JG, Ernest D, Hong AS, DeSantis SM, Cowell LG, Almandoz JP, Messiah SE |
| Journal | Diabetes, Obesity and Metabolism |
| Year / Volume / Pages | 2026; 28:5389–5393 |
| DOI | 10.1111/dom.70700 |
| Study Design | Retrospective cohort study with target trial emulation; propensity score matched (1:1 nearest-neighbour, 0.1 calliper) |
| Database / Setting | TriNetX global EHR network, January 1, 2006 to October 31, 2024 |
| Population | Adults with type 2 diabetes and overweight or obesity |
| Intervention | Initiation of GLP-1 receptor agonist therapy |
| Comparators | SGLT2 inhibitor initiation (Arm 1); metformin initiation (Arm 2) |
| Primary Endpoint | Incident kidney cancer (ICD-10 C64), harmonized with tumour registry data |
| Key Results (GLP-1 RA vs. SGLT2i) | HR 1.05 (95% CI 0.62–1.77); RD 0.0%; RR 1.11 (95% CI 0.66–1.87), no association across all follow-up intervals |
| Key Results (GLP-1 RA vs. Metformin) | HR 1.90 (95% CI 1.39–2.60) overall; HR 1.26 (95% CI 0.77–2.05) in T2D duration <4 years subgroup |
| Adverse Events | Not a safety reporting study; oncologic endpoint only |
| Funding | National Cancer Institute (P30CA142543); Cancer Prevention and Research Institute of Texas (RP210130) |
| Conflicts of Interest | Lopez: grants from NCI and CPRIT during study conduct. Ernest: CPRIT postdoctoral fellowship. All other authors: none reported. |
What Researchers Actually Did
Lopez and colleagues designed this study as an explicit methodological challenge to three prior observational analyses that each reported an elevated kidney cancer risk among GLP-1 RA users. Using the TriNetX global EHR database spanning January 2006 through October 2024, the investigators first specified the protocol of a hypothetical randomized trial, then mapped each component to available observational data in accordance with the target trial emulation framework. Eligible participants were adults with type 2 diabetes and overweight or obesity. Treatment was defined as initiation of a GLP-1 RA versus initiation of either an SGLT2 inhibitor or metformin, with time zero anchored to the moment eligibility criteria, treatment assignment, and follow-up start all coincided. The analysis was framed as the observational analog of an intention-to-treat estimate. Propensity score matching (1:1 nearest-neighbour, 0.1 calliper) was used to adjust for baseline confounders including age, sex, race, ethnicity, socioeconomic determinants, BMI, HbA1c, CKD stage, eGFR, proteinuria, nicotine dependence, antihypertensive use, NSAID use, and others. Covariate balance was confirmed using standardized mean differences below 0.10.
The two treatment comparisons were analyzed separately because they address different methodological questions. The GLP-1 RA versus SGLT2i arm (63,789 eligible, 26,670 matched) was designed to replicate and interrogate the Xie et al. finding using a rigorous trial emulation structure. The GLP-1 RA versus metformin arm (266,172 eligible, 102,490 matched) was designed to replicate and interrogate the Wang et al. finding, with additional stratification by diabetes duration (fewer than 4 years versus 4 or more years) to probe whether channelling bias driven by treatment sequencing explains the observed signal. Reporting adhered to both STROBE and TARGET guidelines.
Key Findings: Primary Outcomes
- GLP-1 RA vs. SGLT2i, no association: In the fully matched cohort of 26,670 adults, 30 kidney cancer events occurred in the GLP-1 RA group versus 27 in the SGLT2i group. The overall HR was 1.05 (95% CI 0.62–1.77), the RR was 1.11 (95% CI 0.66–1.87), and the absolute risk difference was 0.0% (95% CI −0.1 to 0.1). No association was detected at 1-, 3-, or 5-year follow-up intervals.
- GLP-1 RA vs. Metformin, elevated HR overall, attenuated in early T2D subgroup: In the matched cohort of 102,490 adults, the overall HR was 1.90 (95% CI 1.39–2.60). The RR was 1.19 (95% CI 0.88–1.59), substantially lower than the HR, indicating that the hazard signal is temporally front-loaded and that cumulative incidence differences are modest. Absolute risk difference remained 0.0% (95% CI 0.0–0.1), reflecting the low baseline incidence of kidney cancer.
- Attenuation in T2D duration <4 years subgroup: Among the 52,104 matched individuals with fewer than four years of type 2 diabetes at treatment initiation, the HR fell to 1.26 (95% CI 0.77–2.05), no longer reaching statistical significance. The RR in this subgroup was 0.74 (95% CI 0.47–1.18). This subgroup represents the population in which GLP-1 RA and metformin initiators are most clinically comparable.
- 1–3 year follow-up window, GLP-1 RA vs. Metformin: HR was 2.86 (95% CI 1.65–4.94), with RR 2.16 (95% CI 1.25–3.72), suggesting the excess hazard is concentrated early in the follow-up, a pattern more consistent with detection bias or pre-existing subclinical disease than with a causal carcinogenic effect.
Key Findings: Secondary Outcomes and Subgroup Analyses
- GLP-1 RA vs. SGLT2i at 1 year: HR 0.54 (95% CI 0.27–1.11), RR 0.57 (95% CI 0.28–1.16), a numerically lower point estimate for the GLP-1 RA group, though confidence intervals cross 1.0.
- GLP-1 RA vs. Metformin, T2D duration 4 or more years: HR 1.68 (95% CI 0.66–4.23), based on 11 versus 10 events (n=7,535 per group). Wide confidence intervals preclude meaningful interpretation; this reflects low statistical power from sparse events.
- Stratified analyses, GLP-1 RA vs. SGLT2i by T2D duration: Models did not converge due to insufficient event counts in both the <4-year and 4-or-more-year subgroups. Results were not reported for these strata.
- Divergence between HR and RR in the metformin comparison: The overall HR of 1.90 versus an RR of 1.19 indicates temporal non-proportionality. Hazard excess is concentrated early in follow-up, and cumulative cancer burden over the study period is modest, a pattern the authors interpret as consistent with detection or surveillance differences rather than sustained carcinogenic exposure.
Results: Adverse Events and Safety Profile
This study was designed exclusively to evaluate kidney cancer incidence as an oncologic safety signal, not to characterize the broader adverse event profile of GLP-1 receptor agonists. No additional safety or tolerability outcomes were reported. The outcome of interest, incident kidney cancer (ICD-10 C64), was identified using TriNetX data harmonized with tumour registry data to enhance diagnostic validity. Residual misclassification was expected to be non-differential and to bias results toward the null.
Statistical Approach and Rigor
The analytic strategy was appropriate for its stated purpose. Propensity score matching on a broad set of clinically relevant confounders, including markers of renal risk (CKD stages 1–3, proteinuria, eGFR) and metabolic severity (HbA1c, BMI), was a methodological strength. Standardized mean differences below 0.10 confirmed acceptable post-match balance. Reporting HR alongside RR for each comparison was a deliberate and informative choice: divergence between these metrics is diagnostic of front-loaded risk, which the authors appropriately flag as a marker of possible detection bias. Stratification by diabetes duration was prespecified and directly addressed the channelling bias hypothesis. Conditional Cox proportional hazards models accounted for within-match clustering. Limitations of the analytic approach include the ITT framing, which estimates initiation effects and may attenuate associations dependent on sustained exposure; the lack of cumulative dose or persistence data; and the inherently limited power to detect associations for a rare outcome in subgroup strata, as evidenced by model non-convergence in two SGLT2i subgroups and wide confidence intervals throughout.
Clinical Takeaway
Three prior studies raised a signal that GLP-1 receptor agonists may increase kidney cancer risk, a biologically counterintuitive finding given that obesity is a primary driver of renal cell carcinoma and GLP-1 RAs are first-line therapies for obesity management. This target trial emulation, the most methodologically rigorous of the four studies now in this literature, shows that the signal dissolves when the comparator group is clinically appropriate. Comparing GLP-1 RAs to SGLT2 inhibitors, drugs prescribed for overlapping indications at similar points in the treatment pathway, produced no kidney cancer association across any follow-up interval. The elevated HRs seen against metformin are substantially explained by channelling bias from treatment sequencing: GLP-1 RA initiators in real-world data tend to have longer disease duration and greater comorbidity burden than metformin initiators, and when that gap
Read This Paper Through Nine Different Lenses
The same evidence can produce very different conclusions depending on the question being asked. Explore this study through multiple physician-guided interpretive frameworks.
Overview
This study challenges prior findings of elevated kidney cancer risk with GLP-1 receptor agonists by using a target trial emulation framework. It found no increased risk when comparing GLP-1 RAs to SGLT2 inhibitors, highlighting the impact of comparator selection on risk estimates.
The research underscores the importance of methodological rigor in observational studies and suggests that previous signals may have been influenced by biases rather than a causal effect.
- No increased kidney cancer risk with GLP-1 RAs compared to SGLT2 inhibitors.
- Comparator selection significantly impacts risk estimates.
- Target trial emulation provides a robust framework for comparing treatments in observational studies.
Patient Takeaway
For patients, this study suggests that using GLP-1 receptor agonists may not increase the risk of kidney cancer when compared to SGLT2 inhibitors. This information can help in making informed decisions about diabetes management options.
The findings highlight the importance of considering methodological rigor in studies assessing treatment risks and benefits.
- No increased kidney cancer risk with GLP-1 RAs vs. SGLT2 inhibitors.
- Comparator selection affects risk estimates.
- Target trial emulation provides a robust framework for comparing treatments.
Clinician’s POV
Clinicians can use this study’s findings to better understand the risk profile of GLP-1 receptor agonists compared to SGLT2 inhibitors. The research emphasizes the importance of methodological rigor in assessing treatment risks and supports informed decision-making.
These insights can help clinicians provide more accurate information to patients about potential risks and benefits of different diabetes treatments.
- No increased kidney cancer risk with GLP-1 RAs vs. SGLT2 inhibitors.
- Comparator selection affects risk estimates.
- Target trial emulation provides a robust framework for comparing treatments.
A Skeptical Read
This study challenges previous findings of increased kidney cancer risk with GLP-1 receptor agonists by using a more rigorous methodological approach. The results suggest that prior signals may have been influenced by biases rather than a causal effect.
For skeptics, this research underscores the importance of critical analysis and robust methodologies in observational studies to ensure accurate risk assessments.
- No increased kidney cancer risk with GLP-1 RAs vs. SGLT2 inhibitors.
- Comparator selection affects risk estimates.
- Target trial emulation provides a robust framework for comparing treatments.
Study Critic
Critics can evaluate this study’s use of target trial emulation to assess the risk of kidney cancer with GLP-1 receptor agonists. The research provides a methodologically rigorous approach that challenges previous findings and highlights the importance of comparator selection in observational studies.
This critique emphasizes the need for robust methodologies in assessing treatment risks and benefits.
- No increased kidney cancer risk with GLP-1 RAs vs. SGLT2 inhibitors.
- Comparator selection affects risk estimates.
- Target trial emulation provides a robust framework for comparing treatments.
Compared to Past Research
This study builds on and challenges previous findings of increased kidney cancer risk with GLP-1 receptor agonists. It highlights the evolution of methodological approaches in observational studies, particularly the use of target trial emulation to address biases.
The research underscores the importance of continuous evaluation and improvement in study design to ensure accurate risk assessments.
- No increased kidney cancer risk with GLP-1 RAs vs. SGLT2 inhibitors.
- Comparator selection affects risk estimates.
- Target trial emulation provides a robust framework for comparing treatments.
Practical Considerations
Practically, this study’s findings can inform clinical decision-making regarding the use of GLP-1 receptor agonists versus SGLT2 inhibitors. The research emphasizes the importance of considering methodological rigor in assessing treatment risks and supports informed patient care.
Clinicians can use these insights to provide more accurate information about potential risks and benefits to patients.
- No increased kidney cancer risk with GLP-1 RAs vs. SGLT2 inhibitors.
- Comparator selection affects risk estimates.
- Target trial emulation provides a robust framework for comparing treatments.
Future Directions
Future research should focus on understanding the mechanisms behind early detection bias and exploring long-term effects with more detailed exposure data. This study’s findings highlight the need for continued evaluation of treatment risks using robust methodologies.
Further studies can provide deeper insights into the safety profile of GLP-1 receptor agonists in diabetes management.
- No increased kidney cancer risk with GLP-1 RAs vs. SGLT2 inhibitors.
- Comparator selection affects risk estimates.
- Target trial emulation provides a robust framework for comparing treatments.
Misreadings & Bad-Faith Takes
Common misinterpretations of this study may include assuming a causal effect without considering methodological nuances. The research emphasizes the importance of comparator selection and robust methodologies in assessing treatment risks.
Critics should be aware that the findings challenge previous signals of increased kidney cancer risk, suggesting these were likely influenced by biases rather than a causal effect.
- No increased kidney cancer risk with GLP-1 RAs vs. SGLT2 inhibitors.
- Comparator selection affects risk estimates.
- Target trial emulation provides a robust framework for comparing treatments.
Have thoughts on this? Share it:
What is the main finding of the study regarding GLP-1 RAs and kidney cancer risk?
The study found no increased risk of kidney cancer with GLP-1 receptor agonists when compared to SGLT2 inhibitors using a target trial emulation framework.
How does comparator selection affect the results in studies on GLP-1 RAs and kidney cancer?
The choice of comparator (e.g., metformin vs. SGLT2 inhibitors) significantly impacts the risk estimates, with different comparators yielding varying hazard ratios.
What is target trial emulation and why was it used in this study?
Target trial emulation is a method that maps observational data to the design of an ideal randomized controlled trial, allowing for more accurate risk comparisons.
How does diabetes duration affect the relationship between GLP-1 RAs and kidney cancer risk?
The study found that the hazard ratio was attenuated in patients with fewer than four years of type 2 diabetes, suggesting channelling bias may influence results.
What does the divergence between HR and RR indicate in this study?
The divergence suggests that any excess risk is concentrated early in follow-up, which could be due to detection or surveillance differences rather than a sustained carcinogenic effect.
Why was metformin used as a comparator in this study?
Metformin was used to replicate and interrogate the findings of previous studies, particularly those that reported elevated kidney cancer risk with GLP-1 RAs.
What are the limitations of using an intention-to-treat (ITT) approach in this study?
The ITT approach may attenuate associations dependent on sustained exposure and lacks cumulative dose or persistence data, which could affect risk estimates.
The study stratified analyses by diabetes duration to probe whether channelling bias driven by treatment sequencing explains the observed signal.
What is the significance of the absolute risk difference (RD) in this study?
The RD remained 0.0% across all comparisons, indicating that the cumulative incidence differences are modest despite elevated hazard ratios.
What future research is needed based on this study’s findings?
Future research should focus on understanding the mechanisms behind early detection bias and exploring long-term effects with more detailed exposure data.


