OP136 Full Texts Versus Conference Abstract Data: How Does The Message Change?

Abstract

Introduction:

High quality evidence for test accuracy can be scarce. We assessed the test accuracy of two tests (Actim Partus and PartoSure) for the prediction of preterm birth. Twenty published full-text papers were included whilst conference abstracts were excluded. Since systematic reviews of diagnostic tests on other topics may need to rely on data from conference abstracts, we test whether the findings of our review would change with conference abstracts included.

Methods:

Conference citations previously excluded (n=108) were re-screened for inclusion using the following criteria: i) the diagnostic test was Actim Partus or PartoSure ii) test accuracy data of preterm delivery within seven days was reported iii) the population was women with signs/symptoms of preterm labor with intact membranes. Relevant test accuracy data were extracted and used to calculate sensitivity and specificity. Pooled sensitivity and specificity for each test were run using data from full-text papers and conference abstracts combined. These values were compared with the pooled sensitivities and specificities produced for the systematic review using full-text papers only.

Results:

Preliminary pooled sensitivities of the sixteen full-text Actim Partus studies and sixteen full-texts and two abstracts were 0.77 (95% confidence interval (CI) 0.68, 0.83) and 0.76 (95% CI 0.69, 0.83) respectively whilst pooled specificities were 0.81 (95% CI 0.76, 0.85).and 0.80 (95% CI 0.75, 0.84) respectively. Preliminary, pooled sensitivities of the four full-text PartoSure studies and four full-texts and three abstracts were 0.83 (95% CI 0.61, 0.94) and 0.82 (95% CI 0.65, 0.92), respectively, whilst pooled specificities were 0.95 (95% CI 0.89, 0.98) and 0.96 (95% CI 0.94, 0.97), respectively.

Conclusions:

Our findings suggest that the test accuracy results would not alter substantially with the inclusion of conference abstracts. However, work is ongoing to investigate how the assessment of heterogeneity and risk of bias across studies would alter given the difficulties associated with limited methodological reporting from conference abstracts.