PREDICTING IMMEDIATE RESPONDENTS OF CERVICAL MANIPULATION FOR PATIENTS WITH NECK PAIN. Tzyy-Jiuan W. Wang*1; Yuh-Liang Tseng1; Wen-Yin Wang1; Tsun-Jen Hou2; Tzu-Ching Chen3; Fu-Kang Lau3 1. Physical therapy, National Yang-Ming University, Taipei, Taiwan; 2. Rehabilitation Medicine, Tungs Taichung MetroHarbor Hospital, Taichung, Taiwan; 3. Rehabilitation Medicine, Cheng-Hsin Rehabilitation Medical Center, Taipei, Taiwan PURPOSE: To identify the predictors to predict immediate respondents to cervical manipulation for neck pain. BACKGROUNDS/SIGNIFICANCE: Cervical manipulation has been considered an effective treatment for managing nonspecific neck pain. However, clinical observation showed that cervical manipulation may not be effective for every patient with neck pain. Development of clinical prediction rules for identifying patients who are likely to respond to the cervical manipulation treatment may improve clinical decision making and the treatment success rate. SUBJECTS: One hundred patients with neck pain (34 males and 66 females, mean age = 46, SD=11 years) participated in the study. Patients who were contraindicated to cervical manipulation were excluded. METHODS AND MATERIALS: This study was a prospective cohort trial. A standardized interview and a series of physical examinations (cervical range of motion tests, forward head assessment, cervical compression tests, cervical distraction tests, cervical lateral glide tests, atlantoaxial rotation test, first rib elevation test, and cervical short flexor endurance test) were administered to collect the baseline data. A single cervical manipulation session was provided to all participants by two experienced manual therapists. The post-manipulation assessment was executed immediately following the cervical manipulation treatment. Immediate response to treatment was determined by improvement seen in one of the three outcome variables including pain intensity (pain reduction greater than or equal to 50%), general perceived improvement (rated as 'much improvement' or more), and the patient satisfaction level (rated as 'very satisfied'). ANALYSES: Univariate analyses were used to assess the relationships between predictors (the baseline data and the test result of each physical examination) and the immediate treatment response. Stepwise multiple logistic regression analyses were introduced to further identify the significantly important predictors from those potential factors associating with the treatment outcome. The statistical significance level for a variable to enter into the final model was set at p < 0.05 and the statistical level for removing the variable from the model was set at p > 0.10. Accuracy analyses including sensitivity, specificity, and positive likelihood ratio (PLR) were calculated for each potential predictor and for the final prediction models. Positive likelihood ratios, pre-test and post-test probability statistics were used to express the strength of the prediction model. RESULTS: According to the criteria of judging the successful treatment, sixty patients (60%) were classified as respondents to the cervical manipulation, and forty (40%) were nonrespondents. Univariate analyses of all variables between the respondent and non-respondent groups revealed some potential predictors positively associated with successful cervical manipulation. The positive likelihood ratios of these predictors ranged from 1.12 to 3.11. Multivariate logistic regression analysis demonstrated that the 'initial scores on Neck Disability Index < 11.50', 'having bilateral involvement pattern', 'not performing sedentary work > 5 hr/day', 'feeling better while moving the neck', 'without feeling worse while extending the neck', and 'the diagnosis of spondylosis without radiculopathy' were the final six significant predictors of the immediate respondents to cervical manipulation. If 2 predictors were present, the probability of successful manipulation was 67%. If 3 predictors were present (PLR = 3.00), the probability could jump to 82%. Finally, using 4 or more of the 6 identified predictors from the initial evaluation (PLR = 6.22), the probability of successful manipulation would move up to 90%. Therefore, presence of at least three of these six predictors could increase the probability of success with cervical manipulation from 60% to 82% or higher. CONCLUSIONS: Six important predictors for immediate respondents to cervical manipulation were identified. These results implied that when a patient having three or more of the predictors identified in this prediction model, cervical manipulation should be attempted because the patient would have a chance of over 80% to respond immediately to the treatment. FUNDING SOURCE: None. KEYWORDS: cervical manipulation, clinical prediction rule, neck pain, predictor Copyright 2009 by the American Physical Therapy Association. Requests for reprints should be directed to the corresponding author of the article. Educators, students, and other academic customers may receive permission to reprint copyrighted material from Physical Therapy (ISSN 1538-6724) by contacting the Copyright Clearance Center Inc, 222 Rosewood Dr, Danvers, MA 01923. Other types of customers who want permission to reprint should contact the APTA Editorial Office, Attn: Physical Therapy.