Access to clinical trial dataBMJ 2011; 342 doi: https://doi.org/10.1136/bmj.d80 (Published 12 January 2011) Cite this as: BMJ 2011;342:d80
All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
Studying the whole population is the most reliable way to describe
epidemiological phenomena. However, it usually comes with the penalty of
higher costs and other difficulties. Efficiency is the main reason why we
sample the population. Sampling itself brings some other problems, such as
selection bias, statistical complexity, etc.
Health systems with universal coverage, on the other hand, provide
researchers with the potential to access the medical administrative
database of the whole covered population. These secondary data are an
invaluable tool for secondary analysis and researchers can easily study
public health issues from the complete data.
Take the Taiwanese claim database as an example. Taiwan's national
health insurance (NHI) system has been graded as one of the most efficient
health systems in the world.1 However, this system only opens a small
window for researchers to access to the data, a window sized one-twenty-
third of its total dimension. The Taiwanese population is about 23
million,2 but the NHI research database (NHIRD) only allows at maximum a
random sample of one million persons' data for research per application.
The two main official reasons are computational limitations of handling
large amount of data, and for privacy protection.3
However, if computer power is indeed an issue, according to Moore's
law in computer science, computer processing power roughly doubles every
18 months.4 The NHIRD first provided sampled data of 200 thousand people
in year 2000 and of 1 million people in year 2005. Since 2005, computer
processing power should have doubled four times. We reasonably postulate
that at least sixteen million persons' data should be able to be provided
by now, and the whole population data can be accessed by the end of year
2012, as shown in figure 1.
The data, encrypted with personal identities, is obviously scrambled
and linkage to other external dataset is not possible therefore,
confidentiality should not be a concern, either.
However impeccable the sampling method, sampling itself is merely a
proxy of direct measurement but not a replacement. Sampling can never be
better than the whole. Full access to data in healthcare research is vital
for patient safety and quality of care not only in clinical trials, but
also in studies using secondary databases. There are countries where full
access to the administrative database is available, such as the English
HES database,5 or in the Scandinavian countries. The Taiwanese NHIRD is a
national treasure, and data holders should try to lower the barriers to
apply for the complete data for research purposes.
Chen-Kun Liaw, Head of Department 1
Tai-Yin Wu, Attending Physician 23
Azeem Majeed, Professor of Primary Care and Public Health 4
1Department of Orthopaedics, Taoyuan General Hospital, Taoyuan, Taiwan 33004
2Department of Family Medicine, Taipei City Hospital, Renai Branch, Taipei City, Taiwan 10629
3Institute of Preventive Medicine, National Taiwan University, Taipei City, Taiwan 10055
4Department of Primary Care and Public Health, Imperial College London, Charing Cross Campus, London W6 8RP
1. Reinhardt UE. Humbled in Taiwan. BMJ 2008;336(7635):72.
2. Ministry of Interior, Department of Satistics, Taiwan. Monthly
Bulletin of Interior Statistics:
3. National Health Research Institutes, Taiwan. National Health
Insurance Research Database: http://w3.nhri.org.tw/nhird/en/index.htm.
4. Moore GE. Cramming more components onto integrated circuits,
Reprinted from Electronics, volume 38, number 8, April 19, 1965,
pp.114 ff. Solid-State Circuits Newsletter, IEEE 2006;20(3):33-35.
5. Department of Health, United Kingdom. HES online:
Figure 1. According to Moore’s law, computer power doubles every 18 months. If the computer could handle 0.2 million data in year 2000 (blue line), it should be able to handle 32.3 million data by 2011. If the computer could handle 1.0 million data in 2005 (red line), it should be able to handle 16.0 million data by 2011.
Competing interests: No competing interests
Data monitoring committees could play a role in reducing risk of selective outcome reporting during clinical trials
Selective outcome reporting leading to biased treatment and adverse
effect estimates could well be discerned by independent re-evaluation of
full outcome data that are made publicly available after publication of a
clinical trial.1 The requirement for post-publication release of
anonymised complete datasets for all efficacy and adverse outcomes
specified in the study protocol could dissuade selective outcome reporting
by authors when manuscripts are submitted for publication,1 reducing the
risk of biased estimates related to understating serious adverse events2
or the omission of outcomes that do not reach statistical significance.3
However, limited funds, time and staff not being familiar with the
trial protocol and required data analyses restrict feasibility of
independent post-publication reanalysis of raw data. 1 Published trial
findings are viewed as credible until selective outcome reporting is
detected by re-appraisal of publicly accessible datasets. Delay to
discovery of omitted or revised outcomes in journal articles could be
considerable with changed clinical practice based on biased evidence
becoming well established in the meantime. Article corrections and
retractions take time and may only be partly effective at rebutting and
correcting original claims.
With in-depth understanding of the study objectives and protocol as
well as access to fully disclosed results as they accumulate, data
monitoring committees (DMC's) are well placed to gauge compliance with
data collection and use for potentially all prespecified outcomes from
trial commencement to completion. 4 This could improve the chance of
unselected outcome reporting during manuscript preparation. As DMC's are
independent and widely perceived to be extremely competent, 4, 5
completeness of outcomes reported that have also been reviewed and
approved by DMC's prior to publication possess enhanced credibility and
are less likely to be biased. The DAMOCLES Study Group recommends that
DMC's "ensure that trial results are published in an unbiased, correct and
timely manner" and that the DMC has input during discussions of final data
and its interpretation with study investigators. 5 The ability to
effectively monitor for selective data collection and use whilst the study
is being conducted improves the validity and credibility of submitted
manuscripts, eliminates need for resource intensive post-publication re-
evaluation and averts disseminating biased estimates to journal audiences
in the first place.
1. Chan AW. Access to clinical trial data. BMJ 2011; 342: 117-8.
2. Jefferson T, Doshi P, Thompson M, Henegan C on behalf of the Cochrane
Acute Respiratory Group. Ensuring safe and effective drugs: who can do
what it takes? BMJ 2011; 342: 148-51.
3. Smyth MRD, Kirkham JJ, Jacoby A, Altman DG, Gamble C, Williamson PR.
Frequency and reasons for outcome reporting bias in clinical trials:
interviews with trialists. BMJ 2011; 342: c7153.
4. DAMOCLES Study Group. A proposed charter for clinical trial data
monitoring committees: helping them do their job well. Lancet 2005; 365:
5. Sydes MR, Spiegelhalter DJ, Altman DG, Babiker AB, Parmar MKB and the
DAMOCLES Group. Systematic qualitative review of the literature on data
monitoring committees for randomized in clinical trials. Clinical Trials
2004; 1: 60-79.
Competing interests: No competing interests