Table 2 Association of BMI with FEV₁ decline in the systematically reviewed studies

Celli (2008) [16]

TORCH

Random coefficients model. Covariates: baseline FEV₁, age, sex, smoking status, treatment, time on treatment, treatment*time, region

BMI interval, kg/m²

FEV₁ decline (SE), mL/yr

Effect of BMI on FEV₁ decline, p < 0.001

FEV₁ decline decreases with increasing BMI

<  20 (n = 719)

−51.1 (4.4)

20 to < 25 (n = 2003)

−50.5 (2.5)

25 to < 29 (n = 1424)

−42.1 (2.9)

≥29(n = 1197)

−35.1 (3.2)

Calverley (2018) [17]

SUMMIT

Random coefficients model. Covariates: baseline FEV₁, age, sex, smoking status, treatment, time, treatment*time

BMI interval, kg/m²

FEV₁ decline (SE), mL/yr

Effect of BMI on FEV₁ decline, p < 0.001

FEV₁ decline decreases with increasing BMI

< 18.5 (n = 494)

−52 (7.2)

18.5 to < 25 (n = 4562)

−50 (2.3)

25 to < 30 (n = 5362)

−40 (2)

≥30 (n = 5039)

−37 (2.1)

Tashkin (2008) [19]

UPLIFT

Random coefficients model. Covariates: treatment, subgroup, subgroup*BMI

BMI interval, kg/m²

FEV₁ decline (SE),mL/yr

ND

FEV₁ decline decreases with increasing BMI in both treatment groups

Tiotropium

<  20 (n = 242)

−53 (4)

20 to < 25 (n = 915)

−44 (2)

25 to < 30 (n = 903)

−36 (2)

≥30(n = 494)

−34 (3)

Placebo

<  20 (n = 280)

−55 (4)

20 to < 25 (n = 782)

−49 (2)

25 to < 30 (n = 853)

−37 (2)

≥30 (n = 495)

−34 (3)

Anzueto (2015) [20]

TIOSPIR

Mixed repeated measures model. Covariates: treatment, visit, treatment*visit as fixed effects, with random intercept and slope

BMI interval, kg/m²

FEV₁ decline (SE), mL/yr^b

ND

Inverse J-shape – lowest FEV₁ decline in overweight (BMI 25 to < 30 kg/m²)

< 18.5 (n = 51)

−45.2 (19.9)

18.5 to < 25 (n = 436)

−32.1 (6.6)

25 to < 30 (n = 492)

−31.8 (6.1)

≥30 (n = 391)

−35.5 (6.8)

Tkacova (2016) [15]

LHS

Multivariate linear model. Covariates: baseline FEV₁, age, sex, BMI, smoking status

No significant association