Journal of Clinical Epidemiology
Volume 63, Issue 2 , Pages 117-125, February 2010

The effect of study sponsorship on a systematically evaluated body of evidence of head-to-head trials was modest: secondary analysis of a systematic review

  • Gerald Gartlehner

      Affiliations

    • Department of Evidence-based Medicine and Clinical Epidemiology, Danube University, Krems, Austria
    • Corresponding Author InformationCorresponding author. Department of Evidence-based Medicine and Clinical Epidemiology, Danube University, Karl Dorrek-Straße 30, 3500 Krems, Austria. Tel.: +0043-2732-89327-2818; fax: +0043-2732-893-4810.
    • ,
  • Laura Morgan

      Affiliations

    • Cecil G. Sheps Center for Health Services Research, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA
    • ,
  • Patricia Thieda

      Affiliations

    • Cecil G. Sheps Center for Health Services Research, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA
    • ,
  • Anthony Fleg

      Affiliations

    • Department of Family and Community Medicine, University of New Mexico, Albuquerque, NM, USA

Accepted 29 September 2008. published online 02 November 2009.

Article Outline

Abstract 

Objective

The objective of this study was to determine the effect of industry bias in a systematically reviewed body of evidence of head-to-head trials.

Study Design and Setting

We limited our analysis to published head-to-head randomized controlled trials of selective serotonin reuptake inhibitors (SSRIs) identified in a comparative effectiveness review. Two reviewers independently determined the status of funding for each trial. We classified drugs into one of two groups: (1) drugs associated with the funding source and (2) drugs not associated with the funding source. To determine the effect of any underlying industry bias, we conducted relative-benefit meta-analyses comparing the response rates of drugs when associated with the funding source with response rates of the same drugs when not associated with the funding source.

Results

Thirteen out of 20 studies (65%) numerically favored drugs associated with the funding source over drugs used as controls. The pooled response rates of SSRIs, when associated with the funding source, are significantly greater than those of the same SSRIs when not associated with the sponsor (relative benefit=1.07; 95% confidence interval=1.02–1.11). The difference, however, is likely to be not of clinical importance.

Conclusions

The effect of industry bias in comparative effectiveness reviews might play a lesser role than in systematic reviews of placebo-controlled trials.

Keywords: Industry bias, Funding bias, Systematic review, Meta-analysis, Antidepressants, Comparative effectiveness

 

What is new?


•In 65% of studies included in a comparative effectiveness review on selective serotonin reuptake inhibitors (SSRIs), the favored drug was associated with the funding source. No industry supported trial reported a statistically significant benefit of a competitor's drug.

•Drugs, when associated with the sponsor, exhibited a statistically significantly greater efficacy than when not associated with the sponsor in a pooled analysis of response to treatment. The magnitude of the effect, however, was modest and is likely not of clinical relevance.

•Because head-to-head trials are based on a different premise than placebo controlled trials, the impact of industry bias on comparative effectiveness reviews might be different than on systematic reviews of placebo-controlled trials.

Back to Article Outline

1. Introduction 

Multiple studies have documented an association between industry funding and the reporting of positive results [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25]. The exact mechanism of this phenomenon often called “industry bias” remains unclear. Conceivable is a concurrence of omission of outcomes or entire studies and an undue focus on positive results [26]. Other proposed mechanisms include biased methods [3], [19], [27] and analyses [5], [28].

Two recent publications on placebo-controlled trials of second-generation antidepressants, submitted to regulatory agencies, highlighted a disturbing picture of publication bias in this drug class [29], [30]. One study based on trials submitted to the Food and Drug Administration (FDA) reported that only 39% (14 of 36) of trials deemed “negative” by the FDA have been published in journals [29]. Similarly, only 29% of placebo-controlled studies not showing statistically significant differences, which were submitted to the Swedish regulatory agency, were published as stand-alone publications [30]. Because several groups (sponsors, editors, authors) are involved in the publication of research, identifying who or what is exactly responsible for publication bias has not been possible.

These studies indicate that for systemic reviews of placebo-controlled trials, publication bias can have dramatic effects on the validity of results and highlights the need for extensive efforts to uncover unpublished studies.

The situation, however, is even more difficult for systematic reviews that examine head-to-head trials to determine the comparative efficacy and safety of two or more interventions. Such reviews, often called comparative effectiveness reviews, have gained increasing importance for policy makers in recent years [31]. Particularly, the interest in the comparative efficacy and safety of drugs has increased over the past years, based on initiatives, such as the Drug Effectiveness Review Project [32] the Agency for Healthcare Research and Quality's Effective Healthcare Program under the Medicare Modernization Act of 2003 [33].

Although placebo-controlled trials are based on regulatory requirements, head-to-head trials are usually conducted postapproval and thus do not have to be registered with regulatory agencies. Therefore, the detection of unpublished head-to-head evidence is substantially more difficult than for placebo-controlled trials. Protocols of comparative effectiveness reviews usually include requests to the industry for unpublished studies. In reality, however, responses to these requests are minimal, usually leaving researchers conducting comparative effectiveness reviews bound to published studies in their analyses of the evidence. As a consequence, uncertainty exists to what extent findings of comparative effectiveness reviews are distorted by underlying industry bias.

Despite their stated objectives, most head-to-head studies of drugs within the same class do not attempt to formally establish superiority or noninferiority of one drug compared with another. Most studies have mere marketing objectives, and publications of results that render similar efficacy between two drugs frequently emphasize other advantages of one drug over the other, for example, differences in adverse events.

It is unclear to date how industry biases from multiple funding sources affect an entire body of evidence in a systematic review of head-to-head studies. It is conceivable that, in a best-case scenario, biases from multiple funding sources act in opposed directions and eliminate themselves in a systematically located body of evidence. In a worst-case scenario, industry bias could be introduced by a single funding source leading to a unidirectional bias that favors the intervention associated with the sponsor. Such a situation might arise, for example, if a new drug is tested against an established medication.

Because most comparative effectiveness reviews will not be able to detect unpublished evidence, we strove to explore the magnitude of industry bias that could be introduced by drawing conclusions based primarily on published evidence. Using data from a recently conducted comparative effectiveness review as a real-world example, our study set forth to simulate the effect of unidirectional industry bias (the worst-case scenario) on a systematically located and evaluated body of evidence. Our hypothesis was that a particular drug would perform better in trials that were funded by its manufacturer than in studies where it was used as a control drug by a competitor.

Back to Article Outline

2. Methods 

The current study is based on a systematic review of the comparative efficacy and safety of second-generation antidepressants for the treatment of major depressive disorder (MDD) [34]. We chose this drug class because well-conducted systematic reviews and meta-analyses indicate that no substantial differences in efficacy exist among selective serotonin reuptake inhibitors (SSRIs), which are a subgroup of second-generation antidepressants [34], [35], [36]. However, results of individual head-to-head trials of SSRIs present considerable variations of effects. For example, point estimates of the comparative efficacy of achieving response range from a 0% relative benefit of one drug over the other in some studies, to up to a 68% higher relative benefit in others. Whether these differences are simply caused by chance fluctuations or whether underlying industry bias is favoring sponsored drugs over control drugs in some studies is unclear.

A detailed description of the methods of the underlying systematic review has been published elsewhere [34]. Briefly, to identify relevant articles, we searched MEDLINE, EMBASE, The Cochrane Library, and the International Pharmaceutical Abstracts database. We combined “major depressive disorder” with a list of 12 specific second-generation antidepressants (bupropion, citalopram, duloxetine, escitalopram, fluoxetine, fluvoxamine, mirtazapine, nefazodone, paroxetine, sertraline, trazodone, and venlafaxine) and their specific trade names; we used either Medical Subject Headings as search terms when available or key words when appropriate. We limited our searches to “human” and “English language.” Sources were searched from 1980 to 2006 (February). In addition, we manually searched reference lists of pertinent review articles and letters to the editor. Furthermore, we explored the Center for Drug Evaluation and Research database to identify unpublished research submitted to the US FDA.

Two persons independently reviewed abstracts and full-text articles and rated the internal validity of each study based on systems developed by the US Preventive Services Task Force [37] and the National Health Service Centre for Reviews and Dissemination [38]. Records were excluded if they did not meet preestablished eligibility criteria. We included all double-blinded randomized controlled trials (RCTs) lasting at least 6 weeks.

In addition to internal validity, we assessed the comparability of dosages. Because we could not find any clear definitions about equivalence of dosages among SSRIs in the published literature, we developed a roster of low, medium, and high dosages for each drug, which is outlined in Table 1. This classification, based on the interquartile dosing range, does not indicate dosing equivalence. We used this roster to detect gross inequalities in dosing that could affect the comparative efficacy. For the current study, we limited our analysis to published head-to-head RCTs of SSRIs (citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine, and sertraline).

Table 1. SSRIs with dosing classification based on lower and upper dosing range quartiles
Generic name; US trade nameDosage range (mg/d)LowMediumHighManufacturer
Citalopram; Celexa20–60<3030–50>50Forest Laboratories, Inc. and H. Lundbeck A/S
Escitalopram; Lexapro10–30<1515–25>25Forest Laboratories, Inc. and H. Lundbeck A/S
Fluoxetine; Prozac20–60<3030–50>50Eli Lilly and Company
Fluvoxamine; Luvox50–150<7575–125>125Solvay Pharmaceuticals
Paroxetine; Paxil20–60<3030–50>50GlaxoSmithKline
Sertraline; Zoloft50–150<7575–125>125Pfizer, Inc.

Abbreviation: SSRIs, selective serotonin reuptake inhibitors.

To determine the funding source or associations of authors with the funding source, two reviewers independently classified each study according to the following five criteria:

•Clearly funded by the industry: The article explicitly stated that the study was funded by the pharmaceutical industry.

•Probably funded by the industry: The article did not explicitly state the funding source but at least one author was affiliated with the pharmaceutical industry.

•Mixed funding: The article explicitly stated that the study was funded by the pharmaceutical industry and a public or an independent source.

•Clearly funded by public agencies: The article explicitly stated that the study was funded by a public or independent source.

•Funding not reported: No association regarding the funding source could be determined.

Conflicts in classifications were resolved by consensus. We contacted the lead authors of studies where the funding source was not reported. Table 1 summarizes the included drugs and their manufacturers.

Citalopram and escitalopram are both produced by the same manufacturer (H. Lundbeck A/S, Copenhagen, Denmark). Citalopram, however, is also available as a generic drug, whereas escitalopram is still patent protected. In studies comparing citalopram with escitalopram that were funded by Lundbeck A/S, we viewed citalopram as the control drug and escitalopram as the drug associated with the sponsor.

In a next step, we assigned drugs of studies termed to be clearly funded by the industry or probably funded by the industry to one of two groups:

1.Drugs associated with the funding source

2.Drugs not associated with the funding source

Consequently, because of multiple funding sources, each drug was represented at least once in each group.

To simulate a situation where all studies have been funded by the same source causing unidirectional bias, we conducted relative-benefit meta-analyses comparing response rates of drugs termed as associated with the funding source with those of drugs termed as not associated with the funding source. The relative benefit reflects the ratio of benefits in one treatment group compared with that in another. Because we were not interested in the comparative response rates of specific drugs on specific rating scales, we included studies with different rating scales (e.g., Montgomery-Asberg Depression Rating Scale [MADRS], Hamilton Depression Rating Scale [HAM-D]). We used the term “response” as defined by the authors of the individual studies. Because we did not have enough data to compare the response rates of individual drugs, we examined all SSRIs as a class.

For each meta-analysis, we tested for heterogeneity of treatment effects using the I2 statistic. The I2 statistic describes the percentage of total variation across studies that is because of heterogeneity rather than chance [39]. Although no heterogeneity was present in any of our analyses (I2=0%), we report results of the more conservative random-effect models. Results of random-effect models, however, yielded almost identical findings as fixed-effect models.

To detect possible publication bias introduced by the tendency of published studies to be positive, we used funnel plots and the Egger regression approach. All statistical analyses were conducted using StatsDirect Statistical Software program, version 2.5.7 (StatsDirect Ltd, Cheshire, United Kingdom 2006), and STATA 9.1 (StataCorp, College Station, TX, USA 2005).

Back to Article Outline

3. Results 

Our searches identified 2,099 citations. Twenty-nine studies met our eligibility criteria and compared one SSRI with another [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68]. We did not detect any gross dosing inequalities in these studies. Twenty studies were clearly funded by the pharmaceutical industry [40], [41], [43], [44], [46], [47], [50], [51], [52], [54], [56], [57], [58], [59], [60], [61], [63], [65], [67], [68]. Six studies were probably funded by the industry [42], [45], [48], [55], [62], [66]. One study reported a mixed funding source of public and industry monies [49]. In two studies, we could not determine the funding source after contacting the corresponding authors multiple times [53], [64].

Most trials were funded by the following: H. Lundbeck A/S, the manufacturer of citalopram and escitalopram (seven studies, 2,691 patients) [43], [46], [54], [55], [58], [59], [62]; Pfizer, Inc., New York, NY, USA, the manufacturer of sertraline (six studies, 1,520 patients) [40], [41], [42], [61], [66], [68]; and Glaxo Smithkline, London, United Kingdom, the manufacturer of paroxetine (six studies, 959 patients) [44], [45], [48], [50], [65], [67]. Eli Lilly and Company, Indianapolis, IN, USA, the producer of fluoxetine [51], [52], [57], and Solvay Pharmaceuticals, Brussels, Belgium, the producer of fluvoxamine [47], [56], [60], [63], financially supported three and four RCTs each (1,169 and 439 patients, respectively). Table 2 summarizes the included studies.

Table 2. Study characteristics, status of funding, and quality ratings of included studies
StudyStatus of fundingDrug associated with funding source and dose (mg/d)Control drug and dose (mg/d)Sample size and duration (wk)Response (%) and significance levelQuality rating
Aberg-Wistedt et al., 2000 [40]Clearly industrySertraline, 50–150Paroxetine, 20–40353; 2469 vs. 72; P=NR (NS)Fair
Bennie et al., 1995 [41]Clearly industrySertraline, 50–100Fluoxetine, 20–40286; 651 vs. 59; P=NRFair
Boyer et al., 1998 [42]Likely industrySertraline, 20–60Fluoxetine, 50–150242; 2642.6 vs. 47.4; P=NRFair
Burke et al., 2002 [43]Clearly industryEscitalopram, 20Citalopram, 40491; 845.6 vs. 51.2; P=NR (NS)Fair
Chouinard et al., 1999 [45]Likely industryParoxetine, 20–50Fluoxetine, 20–80203; 1288.4 vs. 85.7; P=NR (NS)Fair
Colonna et al., 2005 [46]Clearly industryEscitalopram, 10Citalopram, 20357; 2478 vs. 80; P=NR (NS)Fair
Dalery and Honig, 2003 [47]Clearly industryFluvoxamine, 100Fluoxetine, 20184; 660 vs. 60; P=NR (NS)Fair
Cassano et al., 2002 [44]Clearly industryParoxetine, 20–40Fluoxetine, 20–60242; 52NRFair
De Wilde et al., 1993 [48]Likely industryParoxetine, 20–40Fluoxetine, 20–60100; 662 vs. 67; P=NRFair
Ekselius et al, 1997 [49]Mixed public and industrySertraline, 50–150Citalopram, 20–60400; 2481 vs. 75.5; P=NR (NS)Good
Fava et al., 2002 [52]Clearly industryFluoxetine, 20–60Paroxetine, 20–60; Sertraline, 50–200284; 10–1664.8 vs. 68.8 vs. 72.9; P=NRFair
Fava et al., 2000 [51]Clearly industryFluoxetine, 20–60Sertraline, 50–200; Paroxetine, 20–60284; 26–32NRFair
Fava et al., 1998 [50]Clearly industryParoxetine, 20–50Fluoxetine, 20–80128; 12NRFair
Gagiano, 1993 [53]NRParoxetine, 20–40Fluoxetine, 20–6090; 663 vs. 70; P=NRFair
Haffmans et al., 1996 [54]Clearly industryCitalopram, 20–40Fluvoxamine, 100–200217; 630.5 vs. 28.4; P=NRFair
Kasper et al., 2005 [55]Likely industryEscitalopram, 10Fluoxetine, 20518; 846 vs. 37; P=NR (NS)Fair
Kiev and Feiger, 1997 [56]Clearly industryFluvoxamine, 50–150Paroxetine, 20–5060; 7Data not reported; P=NR (NS)Fair
Kroenke et al. 2001 [57]Clearly industryFluoxetine, 20Paroxetine, 50; Sertraline, 20601; 36NRFair
Lepola et al., 2003 [58]Clearly industryEscitalopram, 10–20Citalopram, 20–40471; 852.6 vs. 63.7; P=0.021Fair
Moore et al., 2005 [59]Clearly industryEscitalopram, 20Citalopram, 40280; 861.3 vs. 76.1; P=0.008Fair
Nemeroff et al., 1995 [60]Clearly industryFluvoxamine, 50–150Sertraline, 50–20095; 7Data not reported; P=NR (NS)Fair
Newhouse et al., 2000 [61]Clearly industrySertraline, 50–100Fluoxetine, 20–40236; 1271 vs. 73; P=NRFair
Patris et al., 1996 [62]Likely industryCitalopram, 20Fluoxetine, 20357; 878 vs. 76; P=NR (NS)Fair
Rapaport et al., 1996 [63]Clearly industryFluvoxamine, 100–150Fluoxetine, 20–80100; 7NRFair
Rossini et al., 2005 [64]NRSertraline, 200Fluvoxamine, 15093; 7Data not reported; P=NR (NS)Fair
Schöne and Ludwig, 1993 [65]Clearly industryParoxetine, 20–40Fluoxetine, 20–60108; 6Data not reported; P=0.03Fair
Sechter et al., 1999 [66]Likely industrySertraline, 50–150Fluoxetine, 20–60234; 2464 vs. 74; P=NRFair
Tignol, 1993 [67]Clearly industryParoxetine, 20Fluoxetine, 20178; 678 vs. 75; P=NR (NS)Fair
Van Moffaert et al., 1995 [68]Clearly industrySertraline, 50Fluoxetine, 20165; 859 vs. 59; P=NR (NS)Fair

Abbreviations: NR, not reported; NS, not significant.

All six SSRIs were associated with the funding source in at least one trial. Likewise, all drugs were used as control interventions in at least one trial. Table 3 outlines the number of comparisons of individual drugs and their association with the funding source. In the following paragraphs, we will refer to drugs not associated with the funding source as the “control drugs.”

Table 3. Number of studies comparing a drug associated with the funding source with individual control drugs
Drug associated with the funding sourceControl drug
CitalopramEscitalopramFluoxetineFluvoxamineParoxetineSertraline
CitalopramN/A01100
Escitalopram4N/A1000
Fluoxetine00N/A021
Fluvoxamine002N/A11
Paroxetine0060N/A0
Sertraline00501N/A

Abbreviation: N/A, not applicable.

Most studies received a fair quality rating. Only one study was rated good [49]. Limitations of internal validity included high attrition or unclear methods of randomization and allocation concealment. Some of these issues, however, are more likely to be attributed to insufficient reporting in the published manuscripts than to methodological flaws of the study designs.

All studies used either the HAM-D or the MADRS as instruments to assess the comparative efficacy of SSRIs. In all studies, response was defined as a 50% improvement of points on either of these scales. Populations were mostly younger than 60 years and suffered from acute-phase MDD. Twenty studies provided sufficient data for relative-benefit meta-analyses of response to treatment (Fig. 1) [40], [41], [42], [43], [45], [46], [47], [48], [50], [51], [52], [54], [55], [58], [59], [61], [62], [66], [67], [68]. Overall, these trials included data on 4,706 patients. No heterogeneity could be detected in any of our analyses (I2 statistic: 0%).

  • View full-size image.
  • Fig. 1 

    Relative-benefit meta-analysis (random effects) comparing response rates of selective serotonin reuptake inhibitors (SSRIs) when associated with the funding source with response rates of SSRIs when not associated with the funding source.

Thirteen out of 20 studies (65%) numerically favored drugs associated with the funding source over drugs used as controls. In two of these studies, differences reached statistical significance [58], [59]. None of the trials found a statistically significant difference in favor of a comparator drug.

The pooled relative benefit of achieving response at study endpoint for drugs clearly or probably funded by the pharmaceutical industry was 1.07 (95% confidence interval [CI]=1.02–1.11; Fig. 1) compared with the same drugs when not associated with the funding source. Although this result reached statistical significance, the difference is not likely to be clinically relevant.

Limiting the analysis to studies that explicitly reported industry support did not change the results (relative benefit=1.07; 95% CI=1.01–1.13). Including the study with mixed industry and public funding [49] in our analyses reduced the point estimate of the effect slightly (relative benefit=1.06; 95% CI=1.02–1.11). The funnel plot (presented in Fig. 2) and the Egger bias test do not suggest any publication bias (P=0.854).

Back to Article Outline

4. Discussion 

Our study found that the pooled response rate of SSRIs when associated with the funding source is statistically significantly greater than that of the same SSRIs when not associated with the sponsor. Pooled results indicated a relative benefit of 1.07 (95% CI=1.02–1.11) favoring drugs associated with the funding source. The magnitude of the effect of this estimate, however, is small and not likely to be clinically significant. Overall, 65% of studies favored drugs associated with the funding source. In two studies, differences reached statistical significance. None of the trials found a statistically significant difference in favor of the comparator drug.

Results of our study fit well into the mixed picture regarding the effects of industry bias. A comparison of numbers needed to treat (NNTs) derived from pooled estimates of industry- and nonindustry-funded studies of migraine drugs and acute pain relievers did not detect any statistically significant differences in NNTs [26]. Similarly, a meta-analysis of industry-sponsored statin trials [69] yielded the same risk reduction of coronary events as the large, independently funded Heart Protection Study [70].

Nevertheless, most studies have found that industry funding leads to more optimistic outcomes or interpretations of outcomes than public funding [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24]. Differences of findings might be attributable to different statistical approaches. The large majority of studies assessing industry bias determined statistical associations between funding sources and conclusions, rather than differences in effect sizes. Furthermore, the influence of industry bias might vary substantially among conditions and comparisons of interventions. Our study focused on head-to-head trials using a real-world example of a comparative effectiveness review on second-generation antidepressants.

One of the great strengths of systematic reviews is their ability to provide a comprehensive picture of the entire body of the available evidence. Individual articles often overemphasize positive results, differences in adverse events, or chance findings simply for marketing reasons. Drug representatives use such articles to aggressively promote their drug of interest to physicians. In such situations, when individual studies are taken as a basis for decision making, industry bias can lead to wrongful conclusions. Positive results can occur by chance, and overemphasizing such findings can be highly misleading. In contrast, comparative effectiveness reviews aggregate data from multiple studies with several different funding sources leading to an entirely different, more balanced perspective of the evidence.

The small effect of industry bias in our study cannot be extrapolated to other drug classes and certainly not to a body of placebo-controlled evidence. The effect of industry bias can be substantially greater in comparisons of drugs vs. placebo or first-generation drugs vs. second-generation drugs. For example, Montgomery et al. found that industry-sponsored trials consistently favored second-generation antipsychotics over first-generation antipsychotics [3]. The Clinical Antipsychotic Trials of Intervention Effectiveness investigation, a recently published, large, publicly funded RCT, however, detected no differences between first- and second-generation antipsychotics [71].

Our study has several limitations. First, the premise of our study was based on the assumption that no significant differences in efficacy exist among SSRIs. Although such a notion has been confirmed by three systematic reviews [34], [35], [36], an exact equivalence of treatment effects is unlikely and would be impossible to prove. Therefore, grouping interventions into two classes, one associated and the other not associated with the sponsor, could yield differences by chance alone. Such a random error, however, could introduce bias in either direction.

Second, our assumption was that industry and publication bias in head-to-head trials follow a similar pattern as was reported from placebo-controlled trials of this drug class. Because placebo-controlled trials are required for regulatory purposes and head-to-head trials are mainly conducted postapproval, this assumption might not be entirely correct. It is conceivable that the proportion of published head-to-head trials is greater than that of placebo-controlled trials. After all, “negative” head-to-head trials can still be used to promote a drug by switching the focus on other differences. The consequence would be that industry bias might actually be less prominent in head-to-head trials of the same drug class than in placebo-controlled trials, which could partly explain the small magnitude of our findings.

Third, as discussed earlier, our results are not generalizable to other drug classes or designs beyond head-to-head trials within this drug class. Our results should not imply that industry bias, in general, is modest. Moreover, our findings are limited to a systematically located and evaluated body of evidence. Finally, although two authors assessed data independently, they were not blinded to the funding status; therefore, author bias cannot be ruled out.

Taken as a whole, it appears that bias may be introduced by industry funding of research, but the evidence on the effect of such a bias is not conclusive. Whether or not industry bias causes erroneous conclusions may vary depending on the condition, intervention, and outcome. The registration of trials before patient enrollment, as required since 2005 by the International Committee of Medical Journal Editors [72], is an important step forward and will, to some extent, increase transparency and reduce industry bias.

Given the increasing occurrence of industry-funded studies in peer-reviewed medical literature, more attention should be paid to the magnitude of potential biases rather than the simple presence or absence of statistical associations between funding source and positive results.

Back to Article Outline

Acknowledgments 

We wish to thank our colleagues from the University of North Carolina at Chapel Hill: Tim Carey and Andy Hansen. We also express our appreciation to Kathy Lohr and Linda Lux of RTI International.

Back to Article Outline

References 

  1. Shah RV, Albert TJ, Bruegel-Sanchez V, Vaccaro AR, Hilibrand AS, Grauer JN. Industry support and correlation to study outcome for papers published in Spine. Spine. 2005;30:1099–1104
  2. Lesser LI, Ebbeling CB, Goozner M, Wypij D, Ludwig DS. Relationship between funding source and conclusion among nutrition-related scientific articles. PLoS Med. 2007;4:e5
  3. Montgomery JH, Byerly M, Carmody T, Li B, Miller DR, Varghese F, et al. An analysis of the effect of funding source in randomized clinical trials of second generation antipsychotics for the treatment of schizophrenia. Control Clin Trials. 2004;25:598–612
  4. Brown A, Kraft D, Schmitz SM, Sharpless V, Martin C, Shah R, et al. Association of industry sponsorship to published outcomes in gastrointestinal clinical research. Clin Gastroenterol Hepatol. 2006;4:1445–1451
  5. Als-Nielsen B, Chen W, Gluud C, Kjaergard LL. Association of funding and conclusions in randomized drug trials: a reflection of treatment effect or adverse events?. JAMA. 2003;290:921–928
  6. Baker CB, Johnsrud MT, Crismon ML, Rosenheck RA, Woods SW. Quantitative analysis of sponsorship bias in economic studies of antidepressants. Br J Psychiatry. 2003;183:498–506
  7. Barnes DE, Bero LA. Why review articles on the health effects of passive smoking reach different conclusions. JAMA. 1998;279:1566–1570
  8. Bhandari M, Busse JW, Jackowski D, Montori VM, Schunemann H, Sprague S, et al. Association between industry funding and statistically significant pro-industry findings in medical and surgical randomized trials. CMAJ. 2004;170:477–480
  9. Chard JA, Tallon D, Dieppe PA. Epidemiology of research into interventions for the treatment of osteoarthritis of the knee joint. Ann Rheum Dis. 2000;59:414–418
  10. Cho MK, Bero LA. The quality of drug studies published in symposium proceedings. Ann Intern Med. 1996;124:485–489
  11. Dieppe P, Chard J, Tallon D, Egger M. Funding clinical research. Lancet. 1999;353:1626
  12. Finucane TE, Boult CE. Association of funding and findings of pharmaceutical research at a meeting of a medical professional society. Am J Med. 2004;117:842–845
  13. Friedman LS, Richter ED. Relationship between conflicts of interest and research results. J Gen Intern Med. 2004;19:51–56
  14. Hartmann M, Knoth H, Schulz D, Knoth S. Industry-sponsored economic studies in oncology vs studies sponsored by nonprofit organisations. Br J Cancer. 2003;89:1405–1408
  15. Kemmeren JM, Algra A, Grobbee DE. Third generation oral contraceptives and risk of venous thrombosis: meta-analysis. BMJ. 2001;323:131–134
  16. Kjaergard LL, Als-Nielsen B. Association between competing interests and authors' conclusions: epidemiological study of randomised clinical trials published in the BMJ. BMJ. 2002;325:249
  17. Lexchin J, Bero LA, Djulbegovic B, Clark O. Pharmaceutical industry sponsorship and research outcome and quality: systematic review. BMJ. 2003;326:1167–1170
  18. Procyshyn RM, Chau A, Fortin P, Jenkins W. Prevalence and outcomes of pharmaceutical industry-sponsored clinical trials involving clozapine, risperidone, or olanzapine. Can J Psychiatry. 2004;49:601–606
  19. Rochon PA, Gurwitz JH, Simms RW, Fortin PR, Felson DT, Minaker KL, et al. A study of manufacturer-supported trials of nonsteroidal anti-inflammatory drugs in the treatment of arthritis. Arch Intern Med. 1994;154:157–163
  20. Stelfox HT, Chua G, O'Rourke K, Detsky AS. Conflict of interest in the debate over calcium-channel antagonists. N Engl J Med. 1998;338:101–106
  21. Thomas PS, Tan KS, Yates DH. Sponsorship, authorship, and accountability. Lancet. 2002;359:351
  22. Turner C, Spilich GJ. Research into smoking or nicotine and human cognitive performance: does the source of funding make a difference?. Addiction. 1997;92:1423–1426
  23. Wahlbeck K, Adams C. Beyond conflict of interest. Sponsored drug trials show more-favourable outcomes. BMJ. 1999;318:465
  24. Yaphe J, Edman R, Knishkowy B, Herman J. The association between funding by commercial interests and study outcome in randomized controlled drug trials. Fam Pract. 2001;18:565–568
  25. Kelly RE, Cohen LJ, Semple RJ, Bialer P, Lau A, Bodenheimer A, et al. Relationship between drug company funding and outcomes of clinical psychiatric research. Psychol Med. 2006;36:1647–1656
  26. Barden J, Derry S, McQuay HJ, Moore RA. Bias from industry trial funding? A framework, a suggested approach, and a negative result. Pain. 2006;121:207–218
  27. Davidson RA. Source of funding and outcome of clinical trials. J Gen Intern Med. 1986;1:155–158
  28. Montaner JS, O'Shaughnessy MV, Schechter MT. Industry-sponsored clinical research: a double-edged sword. Lancet. 2001;358:1893–1895
  29. Turner EH, Matthews AM, Linardatos E, Tell RA, Rosenthal R. Selective publication of antidepressant trials and its influence on apparent efficacy. N Engl J Med. 2008;358:252–260
  30. Melander H, Ahlqvist-Rastad J, Meijer G, Beermann B. Evidence b(i)ased medicine—selective reporting from studies sponsored by pharmaceutical industry: review of studies in new drug applications. BMJ. 2003;326:1171–1173
  31. Teutsch SM, Berger ML, Weinstein MC. Comparative effectiveness: asking the right questions, choosing the right method. Health Aff (Millwood). 2005;24:128–132
  32. Oregon Health and Science University (OHSU). Drug Effectiveness Review Project. Available at http://www.ohsu.edu/drugeffectiveness/description/index.htm, Accessed November 15, 2007.
  33. Agency for Healthcare Research and Quality (AHRQ). Available at: http://effectivehealthcare.ahrq.gov/repFiles/Antidepressants_Final_Report.pdf. Accessed August 26, 2009.
  34. Gartlehner G, Hansen RA, Thieda P, DeVeaugh-Geiss AM, Gaynes BN, Krebs EE, et al Comparative effectiveness of second-generation antidepressants in the pharmacologic treatment of adult depression. Comparative effectiveness review no. 7. (Prepared by RTI-UNC under contract no. 290-02-0016.) Rockville, MD: Agency for Healthcare Research and Quality. Available at: http://effectivehealthcare.ahrq.gov/repFiles/Antidepressants_Final_Report.pdf. Accessed August 26, 2009.
  35. Hansen RA, Gartlehner G, Lohr KN, Gaynes BN, Carey TS. Efficacy and safety of second-generation antidepressants in the treatment of major depressive disorder. Ann Intern Med. 2005;143:415–426
  36. Williams JW, Mulrow CD, Chiquette E, Noel PH, Augilar C, Cornell J. A systematic review of newer pharmacotherapies for depression in adults: evidence report summary. Ann Intern Med. 2000;132:743–7562000
  37. Harris RP, Helfand M, Woolf SH, Lohr KN, Mulrow CD, Teutsch SM, et al. Current methods of the US Preventive Services Task Force: a review of the process. Am J Prev Med. 2001;20(3 Suppl):21–35
  38. Centre for Reviews and Dissemination. Undertaking systematic reviews of research on effectiveness: CRD's guidance for those carrying out or commissioning reviews. CRD Report Number 4 (2nd edition). York, UK: York Publishing Services; 2001;
  39. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003;327:557–560
  40. Aberg-Wistedt A, Agren H, Ekselius L, Bengtsson F, Akerblad AC. Sertraline versus paroxetine in major depression: clinical outcome after six months of continuous therapy. J Clin Psychopharmacol. 2000;20:645–652
  41. Bennie EH, Mullin JM, Martindale JJ. A double-blind multicenter trial comparing sertraline and fluoxetine in outpatients with major depression. J Clin Psychiatry. 1995;56:229–237
  42. Boyer P, Danion JM, Bisserbe JC, Hotton JM, Troy S. Clinical and economic comparison of sertraline and fluoxetine in the treatment of depression. A 6-month double-blind study in a primary-care setting in France. Pharmacoeconomics. 1998;13(1 Pt 2):157–169
  43. Burke WJ, Gergel I, Bose A. Fixed-dose trial of the single isomer SSRI escitalopram in depressed outpatients. J Clin Psychiatry. 2002;63:331–336
  44. Cassano GB, Puca F, Scapicchio PL, Trabucchi M. Paroxetine and fluoxetine effects on mood and cognitive functions in depressed nondemented elderly patients. J Clin Psychiatry. 2002;63:396–402
  45. Chouinard G, Saxena B, Belanger MC, Ravindran A, Bakish D, Beauclair L, et al. A Canadian multicenter, double-blind study of paroxetine and fluoxetine in major depressive disorder. J Affect Disord. 1999;54:39–48
  46. Colonna L, Andersen HF, Reines EH. A randomized, double-blind, 24-week study of escitalopram (10mg/day) versus citalopram (20mg/day) in primary care patients with major depressive disorder. Curr Med Res Opin. 2005;21:1659–1668
  47. Dalery J, Honig A. Fluvoxamine versus fluoxetine in major depressive episode: a double-blind randomised comparison. Hum Psychopharmacol. 2003;18:379–384
  48. De Wilde J, Spiers R, Mertens C, Bartholome F, Schotte G, Leyman S. A double-blind, comparative, multicentre study comparing paroxetine with fluoxetine in depressed patients. Acta Psychiatr Scand. 1993;87:141–145
  49. Ekselius L, von Knorring L, Eberhard G. A double-blind multicenter trial comparing sertraline and citalopram in patients with major depression treated in general practice. Int Clin Psychopharmacol. 1997;12:323–331
  50. Fava M, Amsterdam JD, Deltito JA, Salzman C, Schwaller M, Dunner DL. A double-blind study of paroxetine, fluoxetine, and placebo in outpatients with major depression. Ann Clin Psychiatry. 1998;10:145–150
  51. Fava M, Judge R, Hoog SL, Nilsson ME, Koke SC. Fluoxetine versus sertraline and paroxetine in major depressive disorder: changes in weight with long-term treatment. J Clin Psychiatry. 2000;61:863–867
  52. Fava M, Hoog SL, Judge RA, Kopp JB, Nilsson ME, Gonzales JS. Acute efficacy of fluoxetine versus sertraline and paroxetine in major depressive disorder including effects of baseline insomnia. J Clin Psychopharmacol. 2002;22:137–147
  53. Gagiano CA. A double blind comparison of paroxetine and fluoxetine in patients with major depression. Br J Clin Res. 1993;4:145–152
  54. Haffmans PM, Timmerman L, Hoogduin CA. Efficacy and tolerability of citalopram in comparison with fluvoxamine in depressed outpatients: a double-blind, multicentre study. The LUCIFER Group. Int Clin Psychopharmacol. 1996;11:157–164
  55. Kasper S, de Swart H, Friis Andersen H. Escitalopram in the treatment of depressed elderly patients. Am J Geriatr Psychiatry. 2005;13:884–891
  56. Kiev A, Feiger A. A double-blind comparison of fluvoxamine and paroxetine in the treatment of depressed outpatients. J Clin Psychiatry. 1997;58:146–152
  57. Kroenke K, West SL, Swindle R, Gilsenan A, Eckert GJ, Dolor R, et al. Similar effectiveness of paroxetine, fluoxetine, and sertraline in primary care: a randomized trial. JAMA. 2001;286:2947–2955
  58. Lepola UM, Loft H, Reines EH. Escitalopram (10-20mg/day) is effective and well tolerated in a placebo-controlled study in depression in primary care. Int Clin Psychopharmacol. 2003;18:211–217
  59. Moore N, Verdoux H, Fantino B. Prospective, multicentre, randomized, double-blind study of the efficacy of escitalopram versus citalopram in outpatient treatment of major depressive disorder. Int Clin Psychopharmacol. 2005;20:131–137
  60. Nemeroff CB, Ninan PT, Ballenger J, Lydiard RB, Feighner J, Patterson WM, et al. Double-blind multicenter comparison of fluvoxamine versus sertraline in the treatment of depressed outpatients. Depression. 1995;3:163–169
  61. Newhouse PA, Krishnan KR, Doraiswamy PM, Richter EM, Batzar ED, Clary CM. A double-blind comparison of sertraline and fluoxetine in depressed elderly outpatients. J Clin Psychiatry. 2000;61:559–568
  62. Patris M, Bouchard JM, Bougerol T, Charbonnier JF, Chevalier JF, Clerc G, et al. Citalopram versus fluoxetine: a double-blind, controlled, multicentre, phase III trial in patients with unipolar major depression treated in general practice. Int Clin Psychopharmacol. 1996;11:129–136
  63. Rapaport M, Coccaro E, Sheline Y, Perse T, Holland P, Fabre L, et al. A comparison of fluvoxamine and fluoxetine in the treatment of major depression. J Clin Psychopharmacol. 1996;16:373–378
  64. Rossini D, Serretti A, Franchini L, Mandelli L, Smeraldi E, De Ronchi D, et al. Sertraline versus fluvoxamine in the treatment of elderly patients with major depression: a double-blind, randomized trial. J Clin Psychopharmacol. 2005;25:471–475
  65. Schone W, Ludwig M. A double-blind study of paroxetine compared with fluoxetine in geriatric patients with major depression. J Clin Psychopharmacol. 1993;13(6 Suppl 2):34S–39S
  66. Sechter D, Troy S, Paternetti S, Boyer P. A double-blind comparison of sertraline and fluoxetine in the treatment of major depressive episode in outpatients. Eur Psychiatry. 1999;14:41–48
  67. Tignol J. A double-blind, randomized, fluoxetine-controlled, multicenter study of paroxetine in the treatment of depression. J Clin Psychopharmacol. 1993;13(6 Suppl 2):18S–22S
  68. Van Moffaert M, Bartholome F, Cosyns P, De Nayer AR. A controlled comparison of sertraline and fluoxetine in acute and continuation treatment of major depression. Human Psychopharmacol. 1995;10:393–405
  69. LaRosa JC, He J, Vupputuri S. Effect of statins on risk of coronary disease: a meta-analysis of randomized controlled trials. JAMA. 1999;282:2340–2346
  70. MRC/BHF Heart Protection Study of cholesterol lowering with simvastatin in 20,536 high-risk individuals: a randomised placebo-controlled trial. Lancet. 2002;360:7–22
  71. Rosenheck RA, Leslie DL, Sindelar J, Miller EA, Lin H, Stroup TS, et al. Cost-effectiveness of second-generation antipsychotics and perphenazine in a randomized trial of treatment for chronic schizophrenia. Am J Psychiatry. 2006;163:2080–2089
  72. DeAngelis C, Drazen JM, Frizelle FA, Haug C, Hoey J, Horton R, et al. Clinical trial registration: a statement from the International Committee of Medical Journal Editors. Med J Aust. 2004;181:293–294

 Conflict of interest: None of the authors has any conflict of interest with respect to the topic of this manuscript. All authors had full access to the data of this study.

PII: S0895-4356(09)00231-5

doi:10.1016/j.jclinepi.2008.09.019

Journal of Clinical Epidemiology
Volume 63, Issue 2 , Pages 117-125, February 2010

Article Tools

* Image Usage & Resolution