high relative relapse rate (RRR) of 6-11% between 1995 and 2009, exceeding the acceptable norm (<5%). [3][4][5] The present study was designed to measure the actual relapse rate and to identify associated risk factors.
METHODS
Study designA prospective nested case-control study was conducted to identify risk factors for relapse among TB patients. At the end of the follow-up period (28 February 2010), all those who relapsed within 12 months after completion of treatment were identifi ed from the tuberculosis registry. For each confi rmed relapse case, four control subjects were randomly selected from among patients who had successfully completed the initial phase of treatment and had not relapsed. Sampling was performed using the Statistical Package for the Social Sciences sampling procedure (Base 7.4 for Windows, SPSS Inc, Chicago, IL, USA).
Study area and populationThe study was conducted at health centres with TB units throughout the 10 governorates of Yemen: Metropolitan, Amran, Aden, Taiz, Al-Hodeida, Hajjah, Ibb, Dhamar, Hadramout and Mareb. The study population was a cohort of smear-positive pulmonary TB (PTB) patients registered for DOTS-based treatment between July 2007 and June 2008. Patients who had completed treatment, were considered cured and were aged ⩾15 years were included as subjects, while all patients with smearnegative PTB and extra-pulmonary TB (EPTB), children aged <15 years and adult patients receiving other treatment regimens were excluded.
SamplingThe calculation of the sample size for determining factors contributing to relapse was based on a balanced design for simple logistic regression with a binary covariate (X ). 6 In this method, the proportion (B) of the sample (X = 1) was assumed to be 0.5. The event rates P 1 and P 2 are 0.05 (X = 0) and 0.1 (X = 1), respectively. The power of the study was set at 80%, while the two-tailed type I error was assumed to be 5%. The sample size required to detect a change in probability (Y = 1) from a baseline value of 0.05 to 0.1 was 862 patients. Adjusting for the variance infl ation factor, a total sample size of 880 was needed for multiple logistic regression. Based on the required sample size and the number of patients available according to the previous years' statistics, it was decided that recruitment of all patients within the timeframe July 2007 to June 2008 would provide an adequate sample size. Results: A relapse rate of 5.7% was found. Multivariate logistic regression analysis showed that unemployment, smoking, presence of cavitations, weight gain, weight loss, non-adherence during the continuation phase and diabetes were significantly associated with relapse (P < 0.05).
Interna onal Union Against Tuberculosis and Lung Disease
Conclusion:Relapse rates can be reduced by ensuring that patients take their treatment regularly and are counselled effectively to stop smoking. Reinforcing the implementation of the DOTS strategy and strengthening the anti-smoking campaigns are important actions. Action to help unemployed patients, including free...