Estimating MEDLINE's identification of randomized control trials in pediatric dentistry

Boosting qualifies capture-recapture methods for estimating the comprehensiveness of literature searches for systematic reviews

OBJECTIVE

Capture-recapture methods were proposed to evaluate the comprehensiveness of systematic literature searches. We investigate the statistical feasibility of capture-recapture techniques with model selection for estimating the number of missing references in literature searches using two systematic reviews in gastroenterology and hematology.

STUDY DESIGN AND SETTING

First, we compared manually selected Poisson regression models that differ with respect to included interactions. Secondly, we performed selection via componentwise boosting, which provides automatic variable selection. The proposed boosting technique is a regularized, stepwise procedure allowing to distinguish between mandatory and optional variables. Results from all models were compared based on Akaike's Information Criterion and the Bayesian Information Criterion.

RESULTS

For the first example, the best manually selected model suggested a number of 82 missing articles (95% CI: 52-128), whereas the boosting technique provided 127 (95% CI: 86-186) missing articles. For the second example, 140 (95% CI: 116-168) missing articles were estimated for the manually selected and 188 (95% CI: 159-223) for the automatically selected model.

CONCLUSION

Capture-recapture analysis requires the selection of an appropriate model. Because of problems of variable selection and overfitting, manual model selection yielded large estimates, varying markedly, with broad confidence intervals. By contrast, boosting was robust against overfitting and automatically created an appropriate model for inference.

Caries management: technical solutions to biological problems or evidence-based care?

Caries-related clinical decision-making remains a centrepiece of clinical dentistry. However, the oral disease patterns are rapidly changing towards the better among major sections of the populations, most notably in the high-income countries. The caries decline is well documented among children and younger adults, and is gradually seen to trickle into middle and old age as well. Although it is tempting for the dental profession to take the credit for this development, the evidence points in a different direction. The major contribution of dentistry seems primarily related to changes in the treatment philosophies towards a less interventionist approach. This review aims to spur a further change in the diagnostic and treatment criteria used in the management of dental caries for the benefit of the oral health status of our patients. We must come to terms with the fact that our traditional core skills, our manual dexterity and technical competence, have less to offer to oral health than we have been accustomed to think. The dental schools and the professional dental organizations must carry the responsibility for promoting the necessary changes in the caries related clinical decision-making strategies to allow practicing dentists to provide appropriate oral health care to our populations.

Optimizing search strategies to identify randomized controlled trials in MEDLINE

BACKGROUND

The Cochrane Highly Sensitive Search Strategy (HSSS), which contains three phases, is widely used to identify Randomized Controlled Trials (RCTs) in MEDLINE. Lefebvre and Clarke suggest that reviewers might consider using four revisions of the HSSS. The objective of this study is to validate these four revisions: combining the free text terms volunteer, crossover, versus, and the Medical Subject Heading CROSS-OVER STUDIES with the top two phases of the HSSS, respectively.

METHODS

We replicated the subject search for 61 Cochrane reviews. The included studies of each review that were indexed in MEDLINE were pooled together by review and then combined with the subject search and each of the four proposed search strategies, the top two phases of the HSSS, and all three phases of the HSSS. These retrievals were used to calculate the sensitivity and precision of each of the six search strategies for each review.

RESULTS

Across the 61 reviews, the search term versus combined with the top two phases of the HSSS was able to find 3 more included studies than the top two phases of the HSSS alone, or in combination with any of the other proposed search terms, but at the expense of missing 56 relevant articles that would be found if all three phases of the HSSS were used. The estimated time needed to finish a review is 1086 hours for all three phases of the HSSS, 823 hours for the strategy versus, 818 hours for the first two phases of the HSSS or any of the other three proposed strategies.

CONCLUSION

This study shows that compared to the first two phases of the HSSS, adding the term versus to the top two phases of the HSSS balances the sensitivity and precision in the reviews studied here to some extent but the differences are very small. It is well known that missing relevant studies may result in bias in systematic reviews. Reviewers need to weigh the trade-offs when selecting the search strategies for identifying RCTs in MEDLINE.