Optimisation of statistical procedures to assess the diagnostic accuracy of cervical cancer screening tests

Worldwide, cervical cancer is the fourth most common cancer in females affecting more than half a million women worldwide, half of whom die. Eighty five percent of cervical cancer cases occur in low and middle-income countries. It is the leading cause of death in women with cancer in sub-Saharan Africa. Cervical cancer is caused by high-risk types of the human papillomavirus (HPV). A Pap smear is a screening test used to detect cervical precancer. Out of every 100 screened women, 90 to 98 will have a negative result i.e. normal cervical cells. They are recommended to wait for three to five years before the next screening. Approximately 3 to 20 per thousand women screened would have high-grade abnormal cells meaning that they would be at high risk of developing cervical cancer. They would be referred immediately for further investigation and treatment. Two to ten percent of women will have inconclusive or low-grade Pap smear results. Cervical precancerous lesions, in particular those with low-grade or undetermined atypical findings, clear without treatment. Hence, it is very important to identify women at an increased risk of developing cervical (pre-) cancer and need further testing and treatment. This avoids anxiety and discomfort related to diagnostic and/or therapeutic work-up. Furthermore, it would reduce the financial burden on the women and the health-care system as a whole. The accuracy of different tests triaging women with minor Pap smear results (equivocal or low-grade) has been assessed in many diagnostic studies. These tests detect DNA or RNA of high-risk HPV types. Other triage tests used for triage of minor lesions include repeat Pap smear and protein markers indicative of a transforming HPV infection. The results from all the studies can be synthesized in a metaanalysis. This is done in a systematic review where evidence fitting a pre-specified eligibility criterion is gathered to answer a specific research question. Meta-analyses offer a comprehensive method to gather information for clinical decision making. By conducting a meta-analysis, precision of the accuracy estimates can be improved compared to separate studies. Moreover, controversies arising from apparently conflicting studies can be explained. In this thesis, statistical methods to perform meta-analysis of diagnostic accuracy studies are developed. The methods are specific to binomial and proportions data which is typical of diagnostic studies. The thesis addresses the breakdown of normal approximation procedures encountered when the estimated proportions are equal to zero or one, or when there are few number of studies. A conventional diagnostic accuracy meta-analysis compares two tests only. This thesis presents also a comprehensive and unified inference framework called network meta-analyses that utilizes and synthesizes available data on all different diagnostic tests for the same disease simultaneously. Network meta-analyses improve the estimation process by sharing information between studies yielding more precise estimates, especially for diagnostic tests evaluated in a small number of studies. Moreover, it ensures more efficient use of data and decreases the chance of finding spurious significant effects. Furthermore, the procedure yields all comparisons between any two tests. Such comparisons are of greater relevance to different stakeholder (clinicians, policy makers, epidemiologists) in making decisions on which tests to use in practice.

Jaar van publicatie:2017

Toegankelijkheid:Open