For healthcare professionals only

Analysis

Finding studies on reboxetine: a tale of hide and seek

BMJ 2010; 341 doi: https://doi.org/10.1136/bmj.c4942 (Published 12 October 2010) Cite this as: BMJ 2010;341:c4942
  1. Beate Wieseler, deputy head of department of drug assessment,
  2. Natalie McGauran, researcher,
  3. Thomas Kaiser, head of department of drug assessment
  1. 1Institute for Quality and Efficiency in Health Care, Dillenburger Strasse 27, 51105 Cologne, Germany
  1. Correspondence to: B Wieseler beate.wieseler{at}iqwig.de
  • Accepted 16 August 2010

Beate Wieseler, Natalie McGauran, and Thomas Kaiser use their experience with the assessment of reboxetine to illustrate how publication bias affects health policy decisions and offer some solutions

The antidepressant reboxetine, a selective noradrenaline (norepinephrine) reuptake inhibitor, has been approved in several European countries (including the United Kingdom and Germany) since 1997. However, approval was declined in the United States in 2001. The German Institute for Quality and Efficiency in Health Care (IQWiG) report on the benefit and harm of newer antidepressants concluded in 2009 that, overall, reboxetine was both ineffective and potentially harmful.1 2

What is IQWiG?

  • An independent non-government and non-profit foundation (www.iqwig.de)

  • Established in 2004 as part of German healthcare reform

  • Mainly funded by a levy on inpatient and outpatient services

  • Produces health technology assessments on diagnostic or therapeutic interventions and health economic evaluations for the Federal Joint Committee (G-BA)—the statutory health insurance system’s main decision making body

  • Publishes evidence based consumer health information (www.informedhealthonline.org)

  • Appraises clinical practice guidelines

Data from the IQWiG report are published in the accompanying systematic review of reboxetine versus placebo and selective serotonin reuptake inhibitors for acute treatment of major depression, which includes previously unpublished data.3 An additional analysis of published versus both published and unpublished evidence shows that published evidence overestimates the benefit of reboxetine, while underestimating harm. These typical effects of publication bias have been identified (and in part quantified) not only in other research on antidepressants4 5 but in a wide range of treatments.6 7

Biased evidence may form part of a marketing strategy. Analyses of litigation documents, which are available at the Drug Industry Document Archive (http://dida.library.ucsf.edu), have shown how trials and journal publications are used as marketing tools to promote drug use.8 9 A striking example is the promotion of the off-label use of gabapentin,9 which was heavily supported by the dissemination of literature that reported only selected outcomes.10

Problems in obtaining data for health technology assessment

To minimise the influence of publication bias and increase transparency, IQWiG requests manufacturers of drugs under assessment to sign a voluntary agreement requiring submission of a list of all sponsored published and unpublished trials; submission of CONSORT compliant documents (generally the clinical study reports) on all relevant trials selected by IQWiG; and permission for publication of all previously unpublished relevant data in the assessment report. Unlike other health technology assessment bodies, IQWiG does not accept “commercial in confidence” data.

IQWiG publishes a preliminary report on its website, which is open to comments from all interested parties.11 The retrieval of comprehensive evidence on reboxetine for the preliminary report on newer antidepressants was hampered by the fact that the manufacturer, Pfizer, although providing a list of published trials and European submission documents, did not submit a complete list of unpublished trials as requested by IQWiG.

Secondary publications suggested that there were unpublished studies. The literature search showed that reboxetine was tested in at least 16 trials including about 4600 patients. However, sufficient data were published on only about 1600 of these patients. In June 2009, IQWiG therefore issued the preliminary conclusion that because of the high risk of publication bias, no meaningful assessment of reboxetine was possible and thus no benefit of the drug could be proved.12 13 14 Pfizer immediately publicly claimed “We provided IQWiG with sufficient data: those data that from our point of view are suited for a benefit assessment of Edronax (active ingredient: reboxetine), also in comparison with other drugs” [authors’ translation].15 However, Pfizer then decided to provide most of the missing data. The subsequent assessment showed that, overall, reboxetine had no benefit.2 As a result of the IQWiG report, the Federal Joint Committee, the statutory health insurance system’s main decision making body, plans to stop reimbursement for the drug.

Disparate standards for approval and post-approval decisions

It is generally agreed that regulatory authorities need to have access to the complete data on drugs, and extensive regulation guarantees this access. In contrast, there is insufficient regulation on the evidence required to make health policy decisions after approval, even though these decisions have a considerable effect on the treatment of patients. Furthermore, although regulatory authorities hold the relevant trial data, they often cite confidentiality laws as a reason for denying policy makers and other parties access. Beyond current legislation making trial data publicly available in results databases (see below), there is no legal obligation for manufacturers to provide data for post-approval evaluations by health technology assessment bodies. Consequently, health policy decision makers often have to rely mainly on published information.

Role of regulatory agencies

Reboxetine was approved in several European countries in 1997. In 1999, the Food and Drug Administration granted preliminary approval on the basis of trials primarily conducted outside the US; however, after the results of later US and Canadian trials had been submitted,16 the FDA declined final approval in 2001 because of “a lack of compelling evidence of efficacy.”17 The new trial results and the FDA’s decision had no effect on the European approval status. Post-approval regulatory reviews of drugs have generally focused on safety data; evaluation of new efficacy data (for example, after different approval decisions in other markets) has not been the primary focus.

Authorising bodies will inevitably make decisions at different times, and often new evidence will be available to the body that makes the decision later. If different approval decisions are made, there is no system in place that triggers a review by the body that made the first decision. Of course such a re-evaluation would increase the regulatory workload, and a structured approach would be required to ensure the efficient use of resources—for example, the quantification of the effect of new evidence and an analysis of the reasons for differences in approval decisions. (Discrepant decisions may not necessarily be caused by new evidence; other reasons, such as the availability of alternatives in the respective market may also have a role.) Some regulatory agencies such as the FDA and European Medicines Agency (EMA) are already cooperating closely with each other—for example, the FDA opened an office in Brussels in 2009 and has staff members located at EMA in London.18 Although extended cooperation and additional regulatory resources would be required, the example of reboxetine shows that re-evaluations may improve the quality of treatment for patients.

Legal changes

The US FDA Amendments Act of 2007 solves part of the problem surrounding publication bias in drug research.19 The act requires standardised data on protocol and results to be publicly posted at clinicaltrials.gov for clinical trials (except phase 1) of drugs subject to FDA regulation.20 In general, results must be posted within one year after completion of study. Non-compliance with the law results, among other things, in penalties of up to $10 000 (£6500; €7900) a day.

However, the act has loopholes: it does not cover trials completed before 27 September 2007 or trials of drugs that were never approved.21 Thus although comprehensive information will in future be available on newer drugs, published information on older drugs will remain biased, resulting in an unfair comparison of old and new treatment options. It also means that negative findings on drugs never approved may not be made public, even if the drug is approved elsewhere, as in the case of reboxetine.

In addition, posting of results for trials of approved drugs tested in new, unapproved indications can be delayed for two years (if approval is not granted in the new indication, results must still be posted).21 Finally, existing regulations are insufficient to tackle the problem of reporting selected outcomes; this requires registration of full study protocols, including any amendments,10 22 and of plans for statistical analysis.

Similar legislation to the FDA Amendments Act is under way in Europe with the mandatory public disclosure of data from the clinical trials database EudraCT.23 24 These laws should be expedited and also close the loopholes that the FDA experience shows can be critical.

Recommendations

Current regulations on the publication of trial results are clearly insufficient. The reboxetine case shows that in order to provide patients, clinicians, and health policy makers with unbiased and verified evidence on which to base decisions the following measures are required:

  • Extension of the FDA Amendments Act to include drugs for which approval was declined and worldwide implementation of this type of legislation

  • Public access to regulatory databases containing trials of older drugs not covered by current law21 25

  • Greater data sharing between regulatory authorities, as well as re-evaluation of a drug if approval is declined elsewhere

  • Legal obligation for manufacturers to provide all requested data to health technology assessment bodies without commercial restrictions to publication.

Implementation of all four measures, which should cover trials of both drug and non-drug interventions, would close the information gaps for post-approval decisions and enable adequate decision making in health care.

Notes

Cite this as: BMJ 2010;341:c4942

Footnotes

  • Research, doi:10.1136/bmj.c4737
  • We thank Hilda Bastian for her helpful comments on the first draft of the manuscript.

  • Contributors and sources: BW and NMcG conceived the paper and drew up the first draft. TK critically reviewed and revised the first draft. All the authors reviewed and revised the following drafts and approved the final version. BW is the guarantor.

  • Competing interests: All authors have completed the unified competing interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare no support from any company for the submitted work; the authors are employed by IQWiG and support the mandatory worldwide establishment of trial registries and study results databases.

  • Provenance and peer review: Not commissioned; externally peer reviewed.

References

View Abstract