Influence of drug company authorship and sponsorship on drug trial outcomes

Studies of drug treatments are more likely to report favourableoutcomes when they are funded by the pharmaceutical industry.We compared drug trials reported in three major psychiatricjournals to investigate these influences. Independent studieswere more likely to report negative findings than industry-fundedstudies. However, the involvement of a drug company employee had a much greater effect on study outcome than financial sponsorship alone.

INTRODUCTION

TOP

INTRODUCTION

It is known that studies of drug treatments are more likelyto report favourable outcomes when they are funded by drugcompanies (Bekelman et al, 2003; Lexchin et al, 2003). Thereis also concern over the conflict of interest created by authors' personal financial links to companies (Bodenheimer, 2000; Komsaroff & Kerridge, 2002). Most studies of these influencesare based upon randomised controlled trials in internal medicine.The study reported here concerns a broad range of drug trialsin psychiatry. We explore the difference between having anauthor who is an `employee' of a drug company (defined hereas holding a consultancy, being an employee or being a shareholder)and receiving financial support from a drug company, and howthese influence study outcome in comparison with independentstudies.

METHOD

TOP

METHOD

The British Journal of Psychiatry, American Journal of Psychiatry and Archives of General Psychiatry were selected as being widelyread journals. They were surveyed for original data-based papersconcerning psychiatric drug treatment, published between January2000 and December 2004 inclusive. All methodologies were included(e.g. randomised controlled trials (RCTs), studies of druglevels in breast milk). Journals were searched manually andinformation was collected from full-text versions.

Outcomes were rated by T.T. He was aware of funding, as thiswas apparent in the papers. Studies were classified as reportingpositive findings if they clearly stated that use of the indexdrug led to a better clinical outcome or was better toleratedthan another treatment. Studies were classified as reportingnegative findings if they clearly stated that use of a comparison treatment led to a better outcome or was better tolerated thanthe index drug or that there was no difference in clinicaloutcome or tolerability. Where the conclusions in the fulltext and abstract were equivocal, T.T. made a judgement asto whether the balance of findings was positive or negative.

Papers were included from all psychiatric sub-specialties. Outcomestudies were included that compared an index drug with placebo,another drug or a psychological therapy. Studies were excludedif they concerned an index drug that was long established (e.g.tricyclic antidepressants, lithium, older antipsychotics) unlessit was being used for a novel indication (e.g. testosteronefor resistant depression). Short reports, letters to the editor, editorials, review articles and meta-analyses were excluded.

The authors' relationship with the drug company was determinedfrom declared affiliations and conflicts of interest, or fromacknowledgements. Studies were classified as industry-fundedif the study was wholly or partly funded by a drug company,including funding in kind (provision of drugs and placebos,or an author who was an employee). Authors were regarded as employees if they worked full time for the company, or declaredconsultancy positions or shareholdings. Studies were regardedas independently funded if sufficient information was providedto exclude any of these relationships.

RESULTS

TOP

RESULTS

Of the 198 studies that met the inclusion criteria, 8 (4%) lacked sufficient information on funding and were excluded. The remainingstudies fell into three groups:

studies funded independently of the drug industry (`independent');
studies with one or more authors employed by a drug company (`industry-authored');
studies funded by industry but withoutan employee author (`industry-sponsored').

Of these 190 studies, 33 (17%) were published in the BritishJournal of Psychiatry, 98 (52%) in the American Journal ofPsychiatry and 59 (31%) in the Archives of General Psychiatry.Most studies (157) concerned adults; the remainder concernedelderly people, children, or mothers and babies. Of the 132studies that were randomised controlled trials, 112 (85%) wereindustry-funded. In 75% of studies the index drug was an antipsychoticor an antidepressant (Table 1).

View this table:
[in this window]
[in a new window]

Table 1 Frequency distribution

There was a significant difference between journals in reportingof negative results, the British Journal of Psychiatry beingmore likely to report negative findings than the other two( ${chi}$ ²=7.99, d.f.=2, P=0.0184).

Financial relationship with the drug industry
Forty-four studies (23%) were independent. Of the 146 that were industry-funded, 58 (40%) also received funding from a non-industrysource. Six pharmaceutical companies funded nearly half ofall the studies surveyed. There were 76 industry-authored studies(40%); of these, 64 (84%) had authors who were employees orshareholders. Seventy studies (37%) were industry-sponsored.

Outcomes
Positive findings were reported in 152 (80%) studies, whereas38 (20%) reported negative findings. Independent studies weremore likely to report negative findings than industry-fundedstudies. Sixteen (36%) of the 44 independent studies reportednegative findings compared with 22 (15%) of the industry-fundedstudies. The difference was statistically significant (Yates' corrected ${chi}$ ²=8.3, d.f.=1, P=0.004). Only two (3%) of the 76industry-authored studies reported negative findings. The difference between this group and the independent studies was highly statistically significant (Yates' corrected ${chi}$ ²=22.29, d.f.=1, P<0.0001).A similar statistically significant difference was observedin the reporting of negative findings between industry-authoredand industry-sponsored studies (Yates' corrected ${chi}$ ²=17.18, d.f.=1, P<0.0001). There was no significant difference between independent and industry-sponsored studies in reporting of positive ornegative findings ( ${chi}$ ²=0.44, d.f.=1, P=0.51).

DISCUSSION

TOP

DISCUSSION

The involvement of a drug company employee seems to exert apowerful effect on study outcome, whereas merely acceptingindustry sponsorship appears to have little or no effect. Thisfinding is both novel and counter-intuitive. One might expectthat the difference between the two forms of industry funding would be subtle. In fact, the difference is highly statisticallysignificant, in contrast to the lack of difference betweenstudies with financial sponsorship only and fully independentstudies.

There are some factors that might have confounded our findings.There were more RCTs among the industry-funded studies. Unlikeother investigators, we included all methodologies becausethe number of independent RCTs in psychiatry is small. It mightbe that RCTs are intrinsically more likely to produce positivefindings. Equally, they might be particularly vulnerable to being abandoned when preliminary findings are not promising (Henry et al, 2005). We did not assess the scientific qualityof different studies. It is possible that independent studiestend to be statistically underpowered and that this leads tooverreporting of negative findings (Djulbergovic et al, 2000;Procyshyn et al, 2004).

Our findings are unlikely to be solely due to these factors.All previous studies comparing industry-funded RCTs with independentones have shown that the former are more likely to report positivefindings. If industry-funded studies are less likely to beunderpowered or methodologically flawed, then one would expectthat the reporting of negative findings would be similar in the industry-authored and industry-sponsored groups, whereasactually the sponsored and independent studies were similar.We seem to have found an `all or nothing' effect related tothe involvement of a drug company employee.

In conclusion, we have confirmed previous findings that industry-funded studies are less likely to report negative findings. Our novelfinding is that this effect appears to be largely or exclusivelydue to the presence of a company employee among the authorship.This finding requires replication with attention to differencesin studies' methodological rigour and statistical power, inorder to exclude these as confounding variables.

REFERENCES

TOP