Primary independent variable

Outcomes

Multivariable regression models


Linear regression model fit with ordinary least squares regression approach^{a}

Propensity score regression model fit with GLM method^{b}

Propensity score stratification model fit with GLM method^{c}


Parameter estimate

p value

Parameter estimate

p value

Parameter estimate

p value


Surgical treatment Versus nonsurgical treatment (reference variable)

ABOCOGS deband score

−0.854

0.80

−0.562

0.89

−1.06

0.76

Deband ANB angle

−2.24

0.002

−2.11

0.01

−2.40

0.001

Deband FMIA angle

0.649

0.75

−0.35

0.89

0.765

0.72

Deband IMPA angle

−3.321

0.09

−3.23

0.17

−3.50

0.08

Deband upper incisor to SN plane angle

10.564

0.001

10.03

0.01

11.53

<0.001

Deband overbite

−0.606

0.07

−0.570

0.16

−0.610

0.08

Deband overjet

0.188

0.71

0.283

0.65

0.161

0.76


^{a}In this model, the confounding effects of covariates (age at start of treatment, gender, initial discrepancy index, initial ANB angle, initial FMIA angle, initial IMPA angle, initial U1 to SN angle, initial overbite, and initial overjet) were adjusted. The linear regression models were fit using ordinary least squares regression approach

^{b}A twostaged regression approach was used. In the first stage, propensity scores (predicted probability of a patient having orthognathic surgery) were computed by using covariates (age at start of treatment, gender, initial discrepancy index, initial ANB angle, initial FMIA angle, initial IMPA angle, initial U1 to SN angle, initial overbite, and initial overjet). In the second stage, the effect of surgical versus nonsurgical treatment on outcomes was examined by GLM model in which the propensity score was used as continuous variable and was adjusted as a covariate along with all other covariates

^{c}A twostaged regression approach was used. In the first stage, propensity scores (predicted probability of a patient having orthognathic surgery) were computed by using covariates (age at start of treatment, gender, initial discrepancy index, initial ANB angle, initial FMIA angle, initial IMPA angle, initial U1 to SN angle, initial overbite, and initial overjet). In the second stage, the effect of surgical versus nonsurgical treatment on outcomes was examined by GLM model in which the propensity score was stratified into five bins (based on distribution of scores) and was adjusted as a covariate along with all other covariates