INDUSTRY FUNDING AND PLACEBO QUIT RATE IN CLINICAL TRIALS OF NICOTINE REPLACEMENT THERAPY: A COMMENTARY ON ETTER ET AL. (2007)

Etter and colleagues showed that industry-sponsored trials of nicotine replacement therapy (NRT) were more likely to report statistically significant results compared with non-industry trials [1]. They suggest that one possible reason for this difference may be greater resources in industry-sponsored trials, leading to higher treatment compliance and therefore greater efficacy. However, differential efficacy may be due to increased response in the active arm, reduced response in the placebo arm, or both. Moreover, there is anecdotal evidence that researchers have attempted to increase the likelihood of obtaining a statistically significant result in trials by reducing the rate of placebo responding [2]. We sought to explore this possibility by investigating quit rates in the placebo arm of industry-sponsored and non-industry trials of NRT.

We identified clinical trials of NRT from the relevant review in the Cochrane Database of Systematic Reviews [3]. Study level data were extracted independently by two authors (N.G., E.T.) and discrepancies resolved by a third author (S.G.). Data on 12-month quit rates in the active and placebo arms of each study were extracted. Where data were available only on 6-month quit rates these values were used to impute 12-month quit rate values, given the high correlation between these variables (r = +0.90). In addition, data on individual study characteristics, including year of publication, location of study (United States, other), level of behavioural support (low, high), type of nicotine replacement therapy (NRT) (oral/nasal spray, patch), definition of abstinence (point prevalence, sustained), outcomes measured (6-month, 12-month, both), biochemical validation [none/not stated, exhaled carbon monoxide (CO), cotinine/other biochemical] and industry support (present, absent) were extracted. Two studies conducted across multiple sites in the United States and other countries were coded according to whether or not the majority of sites were in the United States. Linear regression was used to investigate the association between industry support and placebo quit rate. Analyses were performed unadjusted and adjusted for individual study characteristics, including active quit rate to adjust for overall quit success in each sample. All analyses were weighted by the inverse of the individual study standard error (SE).

A total of 113 studies of NRT were identified from the Cochrane review and screened for potential eligibility. Studies were excluded if they did not specify the level of adjunctive behavioural support (k = 3), or offered free choice or combined NRT (k = 3), resulting in the exclusion of six studies, so that a total of 107 studies were included in the final analyses (70 industry-sponsored trials and 37 non-industry trials) (Table 1). Linear regression indicated an association between presence of industry support and reduced placebo quit rate in the unadjusted model [B = −4%, 95% confidence interval (CI) −8%, −1%, P = 0.008], which was attenuated somewhat in the fully adjusted model (B = −2%, 95% CI −4%, −0%, P = 0.048). Given the large number of predictors in our model, we also fitted a reduced model with predictor removal contingent on P > 0.2. Both models yielded similar results (Table 2), and accounted for >70% of the variance in placebo quit rate. When active quit rate was used as the outcome variable and placebo quit rate included as a covariate, there was no association with source of funding.

Table 1.  Characteristics of studies of nicotine replacement therapy.

		Industry support (k = 70)	No industry support (k = 37)	P-value^a
^a Statistical significance calculated using the Mann–Whitney test for continuous variables, and Pearson's χ^2 test for categorical variables. Values represent median (interquartile range) for continuous variables, and n (%) for categorical variables. CO: carbon monoxide.
Year		1995 (1991–1998)	1991 (1987–1997)	0.037
Sample		231 (140–450)	182 (113–305)	0.086
Quit rate	Placebo	10% (6–14%)	14% (8–21%)	0.012
	Active	17% (12–26%)	21% (10–30%)	0.60
Location	USA	33 (47%)	17 (46%)	0.91
	Other	37 (53%)	20 (54%)
Support	Low	16 (23%)	12 (32%)	0.28
	High	54 (77%)	25 (68%)
Type	Oral/spray	35 (50%)	31 (84%)	0.001
	Patch	35 (50%)	6 (16%)
Definition	Point prevalence	20 (29%)	13 (35%)	0.48
	Sustained	50 (71%)	24 (65%)
Outcome	6 months only	20 (29%)	9 (24%)	0.059
	12 months only	24 (34%)	21 (57%)
	Both	26 (37%)	7 (19%)
Validation	None/not stated	8 (11%)	5 (13%)	0.55
	Exhaled CO	47 (67%)	21 (57%)
	Cotinine/other	15 (22%)	11 (30%)

Table 2.  Regression model of presence or absence of industry support and placebo quit rate.

Predictor	B	95% CI	P-value
	Full model
Year	+0.000	−0.002, +0.001	0.26
Location	−0.008	−0.026, +0.010	0.36
Support	+0.008	−0.012, +0.029	0.43
Type of NRT	−0.006	−0.026, +0.013	0.53
Definition of abstinence	−0.016	−0.035, +0.002	0.086
Reported outcomes	+0.005	−0.007, +0.017	0.39
Biochemical validation	+0.014	+0.000, +0.029	0.046
Quit rate in active arm	+0.544	+0.464, +0.624	<0.001
Industry support	−0.018	−0.037, −0.000	0.048

	Reduced model
CI: confidence interval; NRT: nicotine replacement therapy.
Definition of abstinence	−0.013	−0.030, +0.005	0.15
Biochemical validation	+0.018	+0.004, +0.031	0.009
Quit rate in active arm	+0.568	+0.497, +0.639	<0.001
Industry support	−0.020	−0.037, −0.003	0.023

Our analysis suggests that the differential efficacy of NRT between industry-sponsored and non-industry trials may result from differences in the quit rate in the placebo arm of these trials, rather than differences in the active arm. A number of factors may account for this difference between industry-funded and non-industry trials, such as the characteristics of participants recruited (in particular heaviness of smoking or level of tobacco dependence), the exclusion of participants with confounding comorbidity (such as alcohol abuse or mood disorder) or the number of study sites [2]. Interestingly, we did not observe an association between placebo response rate and either year of publication or whether the study was conducted in the United States, in contrast with recent evidence from trials of antidepressant medication [4,5]. These results suggest that there may be important differences in the design and conduct of industry-sponsored trials compared with non-industry trials, which impact specifically upon placebo response rates to increase the likelihood of observing a statistically significant treatment effect.

Declarations of interest

None.

Acknowledgements

M.R.M. is a member of the UK Centre for Tobacco Control Studies, a UKCRC Public Health Research Centre of Excellence. Funding from the Economic and Research Council, British Heart Foundation, Cancer Research UK, the Department of Health and the Medical Research Council, under the auspices of the UK Clinical Research Collaboration, is gratefully acknowledged. The authors are grateful to Jon Heron for statistical advice.