BACKGROUND: Low-dose macrolides (LDM) are anti-inflammatory agents with antineutrophilic activity, but patient selection for LDM therapy in treating chronic rhinosinusitis (CRS) is controversial. This study aimed to assess factors which predict LDM responders. METHODOLOGY: A prospective cohort study was performed. Patients with CRS received roxithromycin (150 mg) once daily for 12 weeks. Nasal secretions and serology were collected. Nine predictors for LDM response were assessed: nasal secretion IgE, nasal secretion IL-5, serum IgE, serum eosinophils, serum neutrophils, nasal polyps, asthma, allergy, and aspirin hypersensitivity, using receiver-operating curve analysis and multivariable logistic regression. Macrolide responders were those with sino-nasal outcome test-22 improvement, symptoms visual analogue scale decreased to ≤5, and no rescue medication. RESULTS: One hundred CRS patients (mean age 47.4±14.1 years, 45% male) were enrolled. Univariable logistic regression showed local total IgE less than 5.21; and serum eosinophils less than 2.2% associated with macrolide response. Multivariate models showed local total IgE maintained an independent association with macrolide response, with an ability to discriminate between responders and non-responders of 63%. Serum total IgE, nasal secretion IL-5, serum neutrophil, nasal polyp, asthma, allergy, and aspirin hypersensitivity showed no association with LDM response. CONCLUSIONS: Low total IgE level in the nasal secretion but not in the serum, predict LDM response.