Background: The grading of radiological severity in clinical trials in tuberculosis (TB) remains unstandardised. The aim of this study was to generate and validate a numerical score for grading chest x-ray (CXR) severity and predicting response to treatment in adults with smear-positive pulmonary TB. Methods: At a TB clinic in Papua, Indonesia, serial CXRs were performed at diagnosis, 2 and 6 months in 115 adults with smear-positive pulmonary TB. Radiographic findings predictive of 2-month sputum microscopy status were used to generate a score. The validity of the score was then assessed in a second data set of 139 comparable adults with TB, recruited 4 years later at the same site. Relationships between the CXR score and other measures of TB severity were examined. Results: The estimated proportion of lung affected and presence of cavitation, but not cavity size or other radiological findings, significantly predicted outcome and were combined to derive a score given by percentage of lung affected plus 40 if cavitation was present. As well as predicting 2-month outcome, scores were significantly associated with sputum smear grade at diagnosis (p<0.001), body mass index, lung function, haemoglobin, exercise tolerance and quality of life (p<0.02 for each). In the validation data set, baseline CXR score predicted 2-month smear status significantly more accurately than did the proportion of lung affected alone. In both data sets, CXR scores decreased over time (p<0.001). Conclusion: This simple, validated method for grading CXR severity in adults with smear-positive pulmonary TB correlates with baseline clinical and microbiological severity and response to treatment, and is suitable for use in clinical trials.