The decision this page serves. Is obesity a big-rock lever for preventing kidney disease in a reasonably-healthy person, the way it is for cardiovascular disease, cancer, and fatty liver? The prior evidence was confounded because it mixed people who already had CKD at baseline into the “incidence” estimate, so the onset question stayed open. Garofalo’s meta-analysis isolates the prevention question: general-population adults with normal renal function at baseline, followed for new-onset kidney disease. (Garofalo et al., 2017)
This is the incidence / prevention cell only — obesity as a modifiable risk factor for CKD onset. Management of established CKD (dialysis, agent selection, RAS-inhibitor titration) is the prescriber zone and lives elsewhere -> Semaglutide and Kidney Outcomes in Chronic Kidney Disease. Dietary protein and kidney function is a different exposure -> Protein Intake and Kidney Function.
The pooled estimates
39 cohorts, 630,677 general-population adults with normal baseline renal function, mean follow-up 6.8 years (range 2.0-18.5); search Jan 2000 - Aug 2016; random-effects pooling; Newcastle-Ottawa quality high. (Garofalo et al., 2017)
| Exposure | Outcome | RR (95% CI) | I2 | Read |
|---|---|---|---|---|
| Obesity | low eGFR (<60) | 1.28 (1.07-1.54) | 95.0% | significant; +28% |
| Obesity | albuminuria | 1.51 (1.36-1.67) | 62.7% | significant; +51% |
| Obesity | CKD (combined, 3 studies) | 1.36 (1.18-1.56) | - | significant |
| Overweight | low eGFR | 1.06 (0.94-1.21) | 50.0% | NOT significant |
| Overweight | combined renal | 1.09 (0.98-1.21) | - | NOT significant |
| BMI +1 kg/m2 | low eGFR | 1.02 (1.01-1.03) | 24.3% | significant, monotone |
| BMI +1 kg/m2 | albuminuria | 1.02 (1.00-1.04) | 0% | significant |
Absolute anchor: across the 3 obesity cohorts that reported eGFR events separately, new-onset low eGFR occurred in 12.3% of obese vs 4.2% of nonobese over ~5.2 years — so the relative 1.28 sits on a low-single-digit annual baseline, and the absolute excess is modest per-person but large at population scale given obesity’s prevalence. (Garofalo et al., 2017) The author’s conclusion: high BMI «predicts onset of albuminuria without kidney failure (CKD stages 1-2) as well as CKD stages 3 and higher, the effect being significant only in obese individuals.» (Garofalo et al., 2017)
Robustness of the eGFR estimate. The 1.28 carries very high heterogeneity (I2 95%), driven almost entirely by one cohort (Weycker); with it removed, RR 1.18 (1.09-1.28) and I2 falls to 57.9% — so the central estimate is a tighter, still-significant 1.18-1.28 band, not a fragile one. Publication bias was not significant across pools. (Garofalo et al., 2017)
Mediation vs confounding — how to read the RR (the load-bearing subtlety)
Obesity causes diabetes and hypertension, and both are on the causal path to CKD. So how the pooled estimate handles diabetes/BP adjustment decides what the RR actually measures. Garofalo took the «most adjusted model in each study» (Garofalo et al., 2017), and the per-study adjustment sets (Table 1) routinely include diabetes, systolic/diastolic BP, fasting glucose, and glycemic status. The paper frames this as rigor — «estimates of risk excess were adjusted for main determinants of new-onset CKD.» (Garofalo et al., 2017)
But diabetes and hypertension are mediators, not just confounders, of the obesity->CKD path. Adjusting a mediator away removes the indirect (mediated) portion and leaves at most the direct (non-mediated) effect — so the pooled RR under-estimates obesity’s TOTAL causal effect on CKD onset. The paper’s own subgroup analysis points the same way: studies that adjusted for glycemic status pooled to RR 1.15 (1.00-1.32) vs 1.32 (1.03-1.68) for those that did not — the identical split for BP adjustment (1.15 vs 1.32) — i.e. adjusting the mediator attenuates the estimate toward null, exactly as over-adjustment predicts. The interaction test is not significant (P=0.325), so this is a direction, not a proven effect-modification. (Garofalo et al., 2017)
The countervailing evidence — that a direct renal pathway exists independent of metabolic syndrome — is the healthy-vs-unhealthy-obese split: CKD risk was raised in metabolically healthy obese (RR 1.30, 1.12-1.52) as well as unhealthy obese (1.63, 1.32-2.03), with no significant difference (P=0.183). (Garofalo et al., 2017) So obesity raises CKD risk even absent the metabolic-syndrome mediators — a direct effect is real.
Decision reading: for the weight-management-as-CKD-prevention decision the TOTAL effect is the relevant quantity (losing weight removes both the direct renal-hemodynamic harm and the diabetes/HTN it would otherwise cause). Because the pooled RR is partly mediator-adjusted, the true prevention benefit of weight loss is plausibly larger than a naive read of 1.28-1.51 — the estimate errs conservative for this decision, not liberal. This is the net-effect, watch-for- compensation discipline running in the favourable direction. Do NOT present the RR as a clean causal total effect: it is observational, residually confounded, and partly mediator-adjusted.
The overweight null and the shape question
Overweight (BMI 25-30 Western; 23-25 Asian-Pacific) did NOT significantly predict low eGFR (1.06, 0.94-1.21) or the combined renal outcome (1.09, 0.98-1.21), yet continuous BMI was monotonically positive (1.02 per kg/m2, low heterogeneity). Garofalo reads this as non-linearity — a threshold where «the risk significantly increased only in the presence of increments to higher BMI values (obesity range)», supported by a meta-regression in which effect size rose with mean baseline BMI. (Garofalo et al., 2017)
The wiki’s reading is more cautious. A categorical null in the overweight band is weak evidence of a true knee: a monotone per-unit effect of 1.02/kg/m2 across a ~5-unit overweight band predicts a categorical contrast of only ~RR 1.10 — indistinguishable from the observed null at that power. The significant, low-heterogeneity continuous slope is more consistent with a monotone curve whose overweight-vs-normal contrast is simply too small to clear significance than with a hard threshold at BMI 30. Per the dose-response discipline, a measured null does not locate a knee, and a guideline-style threshold is first suspected to mark the edge of the evidence, not a curve feature -> The U-Shaped Association Artifact (here the concern is a manufactured plateau below obesity, the mirror image). The decision default is unchanged either way: every reduction into and out of the obese range pays.
Mechanism
The direct renal pathway is hemodynamic. Obesity raises GFR and renal blood flow via afferent arteriolar dilation (from proximal salt reabsorption) coupled with efferent vasoconstriction driven by elevated angiotensin II, producing hyperfiltration, glomerular hypertrophy, focal glomerulosclerosis, and proteinuria; visceral fat also compresses the renal hilum and synthesizes renin-angiotensin-system proteins, and RAS is abnormally activated in obesity. (Garofalo et al., 2017) This is a human-corroborated physiological mechanism (informs direction, not magnitude), and it explains the healthy-obese signal: the RAS/hyperfiltration channel does not require diabetes or hypertension to fire. It also ties obesity->CKD to the visceral/ectopic rather than the total-mass depot -> Ectopic Fat and Depot-Specific Risk (a prior MA found waist circumference predicted CKD; Garofalo used BMI because only 3 studies had waist data). (Garofalo et al., 2017)
Onset is in scope; progression is the prescriber boundary
The prevention case rests on a scope asymmetry the paper states directly: weight-loss intervention in obese subjects «could be therefore considered as a main strategy to limit the CKD burden», and the importance of this preventive approach «grows when considering that once CKD has developed, the role of abnormal BMI on pro- gression of CKD becomes less evident.» (Garofalo et al., 2017) So obesity’s leverage is largest before CKD onset — which is exactly the fabric-zone (prevention) side. Once CKD is established, the lever weakens and the decision becomes therapeutic (out of scope).
Limits and transportability
- Observational, residually confounded. The authors note «although analyses were adjusted for multiple determinants of CKD de novo, the effect of residual confounders could not be excluded», with the very high eGFR heterogeneity plausibly reflecting unmeasured genetic/environmental factors; no individual-patient data. (Garofalo et al., 2017)
- Survival bias / competing risk. Most studies did not report mortality, so obese people dying before CKD onset could bias the estimate; direction unknown (likely toward the null). (Garofalo et al., 2017)
- Single outcome assessment may overestimate incidence, though the chronicity subgroup argued against pure acute-kidney-injury misclassification. (Garofalo et al., 2017)
- Population transportability. 29 of 39 cohorts (512,230 of 630,677 participants) were Asian-Pacific, where “obesity” is defined at BMI >=25, not >=30 — so the pooled “obesity” exposure is often a lower absolute BMI than Western obesity, and the Western-obesity estimate rests on only 10 cohorts. (Garofalo et al., 2017)
- BMI, not adiposity. BMI is the exposure of convenience; the causally-relevant quantity is visceral fat -> Ectopic Fat and Depot-Specific Risk.
Decision relevance (Layer 1)
CKD is prevalent and carries high cardiovascular and quality-of-life burden, and obesity is a big-rock, non-substitutable lever here: no single drug replicates weight loss’s pleiotropy across the renal-hemodynamic, glycemic, and blood-pressure channels at once, so this adds kidney to the list of organ systems (CVD, cancer, liver) on which the same weight-management action pays -> Body Fatness and Cancer Risk, Fatty Liver MASLD and Weight Loss, BMI and All-Cause Mortality. The absolute per-person renal benefit is modest (low baseline incidence), so for an individual this ranks below the smoking/BP/glycemia big rocks unless baseline CKD risk is already elevated (route (a): absolute benefit scales with baseline risk). The GLP-1 landscape partly overlaps the obesity-CKD channel for the treatment side, but for primary prevention in the reasonably-healthy the lever is weight itself -> Semaglutide and Kidney Outcomes in Chronic Kidney Disease.