Quality of logistic regression reporting in studies of tooth survival after periodontal treatment

Abstract

Aim: To evaluate the quality of reporting of logistic regression models used to assess risk factors for tooth loss in patients who have received periodontal treatment.

Materials/methods: The PubMed, EMBASE, BIOSIS Citation Index, CINAHL, Web of Science, and LILACS electronic databases were searched up to 01 March 2014 to identify interventional longitudinal studies assessing risk factors for tooth loss after periodontal treatment. The reference lists of included studies were searched manually. No language restriction was applied to the search. Quality of reporting of logistic regression models was assessed using analytical and documentation criteria with a 15-item checklist. Criteria were judged as met (adequately reported) or not met (not reported). All searches, selection, data extraction, and quality assessment were performed independently and in duplicate.

Results: Of 621 records initially retrieved, 24 articles were included in the analysis. Less than 30% of all 360 datapoints were met. "Coding of independent variables" was reported most frequently [n = 22 (83%) articles]. Criteria such as "internal and external validation of the model" were not met in any study assessed.

Conclusion: The reporting of logistic regression models in studies assessing risk factors for tooth loss in patients who have received periodontal treatment is not optimal.

Keywords: logistic regression; methodological quality; periodontal treatment; prognostic; risk factors; tooth survival.