Statistical analysis:
Baseline comparisons of the three surgical groups for women’s
characteristics used ANOVA tests for continuous and Chi-square tests for
categorical variables.
We used a propensity score matching approach with inverse probability of
treatment weighting to balance the baseline differences between the
surgical groups and limit indication
bias.11Robins JM, Hernan
MA, Brumback B. Marginal structural models and causal inference in
epidemiology. Epidemiology 2000;11:550-60,22Lunceford
JK, Davidian M. Stratification and weighting via the propensity score
in estimation of causal treatment effects: a comparative study. Stat
Med 2004;23:2937-60. A multinomial logistic regression was
constructed to estimate each women’s probability of receiving one of the
three types of surgeries given their baseline covariates (i.e., the
propensity score). Variables of the propensity score model were
prespecified before outcome analyses and included age, body mass index,
smoking, diabetes, surgical history (hysterectomy, or surgery for stress
urinary incontinence or pelvic organ prolapse), physical status score
(ASA), menopausal status, and anatomical defect. Stabilized weights were
used to estimate the average treatment effect in the entire population,
and the extreme weights were
truncated.33Austin PC. The
performance of different propensity-score methods for estimating
differences in proportions (risk differences or absolute risk
reductions) in observational studies. Stat Med 2010;29:2137-48.
Balance between treatment populations was evaluated by standardized
differences of all baseline covariates, with a threshold of 0.1 used to
indicate imbalance.16
Survival curves were obtained with the Kaplan-Meier estimator. In the
absence of earlier events, we censored events as of December 10, 2019.
Two weighted frailty models — one for complications and one for
recurrence/reoperations — were used to compare the three surgical
groups. The models included a non-parametric estimation of the baseline
hazard and a gamma frailty term for the centre effect.44Duchateau
L, Janssen P, Lindsey P, Legrand C, Nguti R, Sylvester R. The shared
frailty model and the power for heterogeneity tests in multicenter
trials. Computational Statistics & Data Analysis 2002;40:603-20.,55Gutierrez
RG. Parametric frailty and shared frailty survival models. The Stata
Journal 2002;2:22-44.
All statistical tests were two-sided, a p-value <.05 was
considered significant. A multiple imputation (R mice package) strategy
was used to deal with the missing data. All statistical analyses were
performed with the R statistical package version 3.6.1 or later (The R
Foundation for Statistical Computing,https://www.R-project.org/).
Patients were not involved in the development of the VIGI-MESH registry.
No core outcome sets were used.