Elsevier

The Lancet

Volume 356, Issue 9230, 19 August 2000, Pages 635-638
The Lancet

Articles
The uncertainty principle and industry-sponsored research

https://doi.org/10.1016/S0140-6736(00)02605-2Get rights and content

Summary

Background

Reporting of pharmaceutical-industry-sponsored randomised clinical trials often result in biased findings, either due to selective reporting of studies with non-equivalent arms or publication of low-quality papers, wherein unfavourable results are incompletely described. A randomised trial should be conducted only if there is substantial uncertainty about the relative value of one treatment versus another. Studies in which intervention and control are thought to be non-equivalent violates the uncertainty principle.

Methods

We examined the quality of 136 published randomised trials that focused on one disease category (multiple myeloma) and adherence to the uncertainty principle. To evaluate whether the uncertainty principle was upheld, we compared the number of studies favouring experimental treatments over standard ones. We analysed data according to the source of funding.

Findings

Trials funded solely or in part by 35 profit-making organisations had a trend toward higher quality scores (mean 2·94 [SD 1·3]; median 3) than randomised trials supported by 95 governmental or other non-profit organisations (2·4 [0·8]; 2; p=0·06). Overall, the uncertainty principle was upheld, with 44% of randomised trials favouring standard treatments and 56% innovative treatments (p=0·17); mean and median preference evaluation scores were 3·7 (1·0) and 4. However, when the analysis was done according to the source of funding, studies funded by non-profit organisations maintained equipoise favouring new therapies over standard ones (47% vs 53%; p=0·608) to a greater extent than randomised trials supported solely or in part by profit-making organisations (74% vs 26%; p=0·004).

Interpretation

The reported bias in research sponsored by the pharmaceutical industry may be a consequence of violations of the uncertainty principle. Sponsors of clinical trials should be encouraged to report all results and to choose appropriate comparative controls.

Introduction

There is concern that clinical trials sponsored by the pharmaceutical industry result in biased findings.1, 2 Two reasons have been suggested. First, the quality of research sponsored by profit-making firms may be poor. For example, low-quality trials overestimate therapeutic benefit by an average of 34%.3 Second, it could be that industry will sponsor only those research projects that are likely to be positive, thus violating the uncertainty principle—a fundamental scientific and ethical principle for running randomised clinical trials.4, 5 The uncertainty principle, or equipoise, states that the patient should be enrolled in a randomised controlled trial only if there is substantial uncertainty ("equal bet”)4, 5, 6, 7, 8, 9 about which of the trial treatments would benefit a patient most. To distinguish between these possibilities we assessed the influence of the source of funding on the quality of randomised trials and whether the uncertainty principle was adhered to in their conduct.

Section snippets

Trials

We reviewed therapeutic studies in patients with multiple myeloma. By focusing on one disease, existing knowledge that plays a key role in the planning and design of the study can be better ascertained. Using the Cochrane search strategy, we identified all randomised trials for multiple myeloma from the period from 1996 to 1998.10 113 articles on 136 trials were identified. The quality of each trial was assessed with the scale of Jadad and colleagues.3, 11 This scale has a score from 0 to 5 and

Results

The overall quality of the randomised trials was poor (mean Jadad score 2·5 [SD 1]). 53 (39%) and 45 (33%) of the trials had scores of 2 and 3, respectively, accounting for 98 (72%) of the trials. Only seven trials (5%) had the maximum score of 5.

Figure 1, Figure 2 show the rating scores for all trials, with 56% of the studies reporting success for innovative treatments and 44% favouring standard therapy (p=0·170). The mean evaluation score was 3·7 (SD 1) with a median rating of 4, slightly

Discussion

Chalmers17 previously asked “what is the probability that new treatments will be superior to standard ones?” Gilbert and colleagues14 reviewed 46 randomised trials that compared surgical and anaesthesia treatments. 49% of innovations were successful when compared with standard treatment. Colditz and colleagues15 examined 36 randomised trials and found a 61% chance that a patient who received the new therapy would fare better than a patient receiving standard therapy. Machin and colleagues18

References (31)

  • A Bradford Hill

    Medical ethics and controlled trials

    BMJ

    (1963)
  • A Bradford Hill

    Clinical trials and the acceptance of uncertainty

    BMJ

    (1987)
  • K Dickersin et al.

    Identifying relevant studies for systemic reviews

    BMJ

    (1994)
  • RJ Lilford et al.

    Equipoise and the ethics of randomization

    J R Soc Med

    (1995)
  • DL Sackett

    Why randomized controlled trials fail but needn't: 1. Failure to gain “coal-face” commitment and to use the uncertainty principle

    CMAJ

    (2000)
  • Cited by (424)

    • Identification of threshold for large (dramatic) effects that would obviate randomized trials is not possible

      2022, Journal of Clinical Epidemiology
      Citation Excerpt :

      If such uncertainty does not exist, as in case when there is a high likelihood (say, greater than 80-90% [10,11]) that one treatment is predictably better than the other, scientifically no meaningful results would be generated, it would be unethical to enroll people in such a trial, [9,5] ethical committees would likely not approve it [10,11] and well-informed patients would likely refuse to participate in the studies where the results are predictable in advance[5,9]. However, this unpredictability of the results at the level of individual trial led to hypothesis that there is a predictable relationship between the uncertainty requirement (the moral principle) on which trials are based and the distribution of treatments effects at the group level of cohort(s) of clinical trials [5,9,12–14]. That is, it can be predicted that when overall distribution of all treatment effects is evaluated, the probability of finding that a new treatment is better than a standard treatment is about 50% [5,9,12-14].

    View all citing articles on Scopus
    View full text