Addressing reporting bias in systematic reviews

Tags: Oral
Jørgensen L1, Gøtzsche PC1, Jefferson T1
1Nordic Cochrane Centre

Background:

Unabridged access to drug industry and regulatory trial registers and data reduces reporting bias in systematic reviews and may provide a complete index of a drug's clinical study programme.

Objective:

To index the human papillomavirus (HPV) vaccine study programmes as a basis for a systematic review of clinical study reports.

Methods:

By cross-verification via enquiries to the HPV vaccine manufacturers and regulators and searches of regulatory reviews, common technical documents, drug approval packages, European Public Assessment Reports, 45 trial registers (including ClinicalTrials.gov), CENTRAL, Google Scholar, PubMed and WikiLeaks, we indexed clinical interventional prospective comparative HPV vaccine studies in humans.

Results:

We indexed 206 studies. Most were cross-verified from two or more sources (160/206, 78%) and listed on regulatory or industry trial registers or journal publication databases (195/206, 95%) - in particular, on ClinicalTrials.gov (176/195, 90%). However, study results were only posted for 48% (71/147) of the completed studies on ClinicalTrials.gov. Journal publications were available in journal publication databases for 62% (92/149) of the completed studies.

Conclusions:

Our index showed serious deficiencies in the availability of HPV vaccine studies and data. Only one of four manufacturers provided information for our index and a fifth of the index could not be cross-verified. However, we indexed larger study programmes than those listed by major regulators (EMA and FDA, which based their HPV vaccine assessments on half of the available trials). Compared to electronic database searches, indexing is superior at addressing reporting bias. We advocate registering and making publicly available all studies.

Healthcare involvement:

To reduce reporting bias and the risk of biased reviews, systematic review authors must stop producing reviews based solely on journal publications (e.g. the publication for one study (HPV-008) was 14 pages long, while its publicly available corresponding clinical study report was more than 7000 pages). Looking for regulatory data and compiling an index may give authors a very good idea of what they are missing and what the limits of their reviews are. Authors of systematic reviews may recognise and reduce reporting bias if they adopt our index process for study programmes.