Subchondral stress fractures of the femoral head associated with military training: Clinical features and analysis

Table 4 Multivariable logistic regression analysis of factors associated with diagnostic group

Variable	B	SE	Adjusted OR	95%CI	P value
Training intensity (per category increase)	-2.117	0.498	0.120	0.045-0.319	< 0.001a
Bilateral vs unilateral	3.053	0.630	21.188	6.168-72.784	< 0.001a
Alcohol exposure (overall)					0.019a
Alcohol exposure: Light vs none	1.181	0.749	3.259	0.751-14.137	0.115
Alcohol exposure: Moderate vs none	1.846	0.795	6.336	1.334-30.100	0.020a
Alcohol exposure: Heavy vs none	3.013	1.274	20.353	1.675-247.233	0.018a

^aP < 0.05. Outcome coding: ONFH = 1 and SSFFH = 0. Forward likelihood-ratio binary logistic regression was used (entry P < 0.05; removal P > 0.10). Training intensity was coded 1-3 and its odds ratio represents a one-category increase. Alcohol exposure used none as the reference. Corticosteroid exposure was excluded because quasi-complete separation produced unstable maximum-likelihood estimates. Overall alcohol P value is from the likelihood-ratio test; contrast P values are Wald tests. Model n = 174; omnibus likelihood-ratio χ²(5) = 124.18, P < 0.001; Hosmer-Lemeshow χ²(4) = 3.91, P = 0.418. OR: Odds ratio; SE: Standard error.