Two RCTs and 11 observational studies (n=528, range eight to 94) met the inclusion criteria. The quality of observational studies ranged between 3 and 8 (low to moderate quality). There were no drop-outs in the two RCTs, but allocation concealment was unclear and neither study was blinded or used ITT analysis. Follow-up ranged between four weeks and one year in the non-surgical studies and between one and two years in the surgical studies (one study reported follow-up up to 13 years).
For non-surgical interventions, weight loss did not statistically significantly change glomerular filtration rate or creatinine clearance (two RCTs, three observational studies); there was some evidence of statistical heterogeneity among the three observational studies for glomerular filtration rate (I2=57%).
Weight loss interventions statistically significantly reduced proteinuria (WMD -1.65g/24hours, 95% CI -3.21 to -0.08, I2=0%; two RCTs and WMD -1.05g/24hours, 95% CI -2.08 to -0.01, I2=67%; two observational studies).
Compared with controls, non-surgical weight loss interventions reduced BMI significantly, but this involved the pooling of two RCTs and three observational studies and there was evidence of significant statistical heterogeneity (I2=90%).
Systolic blood pressure and total cholesterol were also reduced with non-surgical interventions. Triglycerides and HDL cholesterol levels did not change significantly. Findings for other secondary outcomes were reported in the review.
For surgical interventions, weight loss normalised glomerular filtration rate in patients with glomerular hyperfiltration (WMD -25.56mL/min, 95% CI -36.23 to -14.89; three observational studies) without evidence of statistical heterogeneity (I2=0%). BMI was statistically significantly reduced with weight loss (WMD -16.53kg/m2, 95% CI -19.59 to -13.48; three observational studies), but with evidence of statistical heterogeneity (I2=78%). Systolic blood pressure was reduced (results reported in the review), but with statistical heterogeneity (I2=86%).
Subgroup and sensitivity analyses did not significantly alter the results.