Individual Patient Data (IPD) meta-analysis of diagnostic and prognostic...


Wijzig gegevens

Titel Individual Patient Data (IPD) meta-analysis of diagnostic and prognostic studies in obstetrics, gynaecology and reproductive medicine
Looptijd 06 / 2007 - onbekend
Status Afgesloten
Onderzoeknummer OND1326493
Leverancier gegevens Website ZonMw

Samenvatting (EN)

METHODOLOGICAL OBJECTIVES: 1) To derive, using raw data collected from existing studies, accurate estimates of prior probability of disease; an efficient diagnostic model with good discriminatory power which will enable us to calculate diagnostic probabilities for specific patient characteristics; and a better understanding of reasons for between study heterogeneity. 2) To initiate the execution of IPD meta-analysis in diagnostic research3) To contribute to the development of a framework for the performance of IPD meta-analysis for diagnostic and prognostic researchCLINICAL OBJECTIVESDepending on the results of the analyses, current guidelines for the management of each of the four clinical problems (Diagnosis of endometrial cancer in women with PMB, prediction of preterm delivery, diagnosis of tubal pathology in subfertile women, and prediction of ovarian response in women undergoing IVF) will be adjusted. Our ultimate goal is to introduce a probabilistic orientated approach into clinical practice to secure better patient care. The clinical objective of the project is to make these probabilities available to clinicians. We will do this with score forms on paper, website applications and software available through PDAs. General considerationsThe clinical process dictates that history, physical examination and additional investigations are carried out in that order. Research evaluating diagnostic and prognostic strategies should also take this sequence into account, but this is generally not the case. Data on the accuracy of a diagnostic or prognostic test are usually presented in isolation from history and physical examination. This means that this information on test accuracy is of limited use for clinical practice as the implicit assumption of independence between test performance and the information obtained in the diagnostic work up leading to the decision to perform the test is usually not valid. Multivariable analysis may adjust for this, but is rarely used in primary diagnostic and prognostic research.Meta-analysis aims to summarize the results of multiple studies and may therefore offer more valid data on the accuracy of tests than single studies. Several limitations of meta-analysis can be identified. First, current meta-analytic techniques for diagnostic tests requires dichotomisation of test results in positive or negative test results, and cannot take the continuous character of most medical tests into account. However, data of history and physical examination are usually not taken into account, as the primary studies do not report on the interaction between history, physical examination and additional testing. As a consequence, the integration of these issues is not taken into account in conventional meta-analyses of diagnostic tests. As the data of history and physical examination, though most times not reported, are usually collected in primary studies, meta-analysis using individual patient data (IPD) from these primary studies could potentially overcome these problems. Moreover, IPD meta-analysis allows a full investigation of the influence of patient characteristics (covariates) on the heterogeneity of estimates of test accuracy, both within and between studies. First, by using the original continuous data instead of the dichotomized reported test results, the statistical power to detect a (linear) relation between the test result and the likelihood of disease will increase. Second, IPD meta-analyses potentially allow examining the accuracy of combination of tests and the sequence of relevant tests, as current conventional meta-analyses are limited to the data presented in original reports, and these reports often only contain information on a single test. At present, IPD meta-analyses in the field of diagnosis and prognosis are lacking. MethodsWe propose to undertake IPD meta-analysis of relevant studies identified in systematic reviews previously published by our group. We plan to obtain the individual patient data from the principal investigators and to re-analyse the estimates of diagnostic or prognostic accuracy of tests in their clinical context. We will explore four clinical problems Diagnosis of endometrial cancer in women with postmenopausal bleeding Diagnosis of tubal pathology in subfertile women Prediction of preterm delivery Prediction of ovarian response in women undergoing IVFFor each clinical topic, the specific objectives are as follows: To estimate accurately prior probability of disease based on history and physical examination. To explore whether the use of continuous test results in IPD meta-analyses leads to different estimates of test accuracy as compared to conventional meta-analysis. To develop efficient diagnostic/prognostic models that will enable us to calculate probabilities for specific patient characteristics in combination with test results. To have a better understanding of the impact of heterogeneity between studies on estimates of test accuracy. In addition, we wish to promote the use of IPD meta-analysis in diagnostic research and to motivate others to experiment with it in other fields of medicine, if it were proven to be feasible. Eventually, collaboration of primary investigators might lead to new, mutually planned, prospective primary studies to corroborate or refute the model derived from the IPD meta-analyses above.We have demonstrated feasibility in a pilot IPD meta-analysis comprising raw data from four studies in postmenopausal bleeding. Even more important, we have found > 25 principal investigators of the studies relevant for the other topics, who are prepared to share data. We plan to organise a workshop to finalise the study protocol, discuss the patient characteristics and the information from the diagnostic/ prognostic tests to be analysed, the data checking procedures and the main analyses to be performed. The analytic approach will consist of thorough data checks (single variables, simple tables and plots), univariable analyses against the dependent variable, followed by regression modeling. We will develop a website to present the results of the study and make them accessible for individual clinicians.

Betrokken organisaties

Betrokken personen

Projectleider Prof.dr. B.W.J. Mol


D23220 Inwendige geneeskunde

Ga terug naar de inhoud
Ga terug naar de site navigatie