Table 5.

Baseline Predictors of Improvement in Gastroparesis Symptoms after 48 Weeks using Logistic Regression Analyses in Patients with Gastroparesis

Baseline Predictors Selected by Minimum AIC	Symptoms Improved Vs Not Improved (N=260; 73 Improved)
	OR*	CI	P†
Demographics:
Sex (male vs female)	2.30	0.93–5.67	.07
Age (≥ 50 vs <50 years)	3.35	1.62–6.91	.001
Education: college degree or higher (yes vs < no)	0.58	0.27–1.22	.15
Gastroparesis Symptoms:
Overall GCSI score	2.87	1.57–5.23	.001
Post-prandial fullness severity subscale	1.53	0.92–2.57	.10
Abdominal pain moderate/severe‡ (vs not)	0.40	0.17–0.96	.04
GERD severity subscale	0.66	0.50–0.89	.006
Gastric Emptying Scintigraphy:
Retention > 20% at 4 hours (vs ≤ 20%)	2.22	1.12–4.40	.02
Clinical factors:
Overweight or obese (BMI ≥25 vs < 25 kg/m2)	0.43	0.22–0.84	.01
Smoked (ever vs never regularly)	0.46	0.22–0.98	.04
Initial infectious prodrome (vs not)	2.22	1.01–4.87	.05
Prokinetics use (vs none)	0.57	0.29–1.10	.09
Antidepressants use (vs none)	2.27	1.13–4.58	.02
Anti-anxiety (anxiolytics) use (vs none)	0.28	0.10–0.79	.02
Pain modulator medication use (vs none)	0.34	0.14–0.81	.01
Psychological:
Depression‡: Moderate to severe (BDI > 20 vs ≤ 20)	0.45	0.22–0.93	.03
Intercept	0.02	0.00–0.11	<.001
AUROC	0.77§	0.71–0.84

^*Odds of improvement; improvement defined as GCSI score at 48 weeks compared to the baseline score is 1 point or more lower (number patients with improvement=73, without improvement=187, with missing data for a predictor=2).

^†

P values were determined from a multiple logistic regression analyses of improvement using the 35 baseline predictors in the candidate set listed in Table 4, including baseline GCSI score. To account for over-dispersion of the model, standard errors were scaled using square root of the deviance-based dispersion. Nausea, post-prandial, and GERD severity subscales and the SF-36 v2 physical and mental health components were analyzed as continuous rather than categorical variables. The bloating subscale was excluded from the candidate set due to collinearity with other baseline predictors. The model presented was the model selected having the minimum Akaike Information Criterion (AIC) amongst all of the possible subsets of predictors.

• Goodness of fit: Hosmer-Lemeshow χ² (d.f.=6)=10.0, P=.12
• Cross-validated AUROC using jackknifed regression method is presented in the table.
• AUROC without cross-validation is: 0.84 (0.78–0.89).

^‡Moderate/severe abdominal pain defined as either upper abdominal pain or discomfort PAGI-SYM symptom score≥3; Beck Depression Index (BDI) score>20 indicates a moderate to severe level of depression; Trait Anxiety score ≥50 indicates severe anxiety

^§Prediction equation (b(SE)) for probability(improvement) = odds/(1+odds), where odds=EXP[−3.981(0.932) + 0.834(0.480) if male + 1.209(0.385) if age≥50 (yrs) −0.548(0.398)^*if college degree or higher + 1.053(0.321)^*GCSI total score + 0.428(0.273)^*post-prandial subscale score −0.907(0.461) if abdominal pain moderate/severe −0.409(0.156)^*GERD subscale score + 0.795(0.365) if 4 hr GES retention >20% −0.842(0.356) if BMI ≥25 (kg/m²) −0.770(0.400) if smoked regularly + 0.798(0.418) if initial infectious prodrome −0.564(0.350) if use prokinetics + 0.820(0.374) if use anti-depressants −1.264(0.545) if use anxiolytics −1.086(0.463) if use pain modulators −0.788(0.381) if BDI>20]