Interventions for preventing weight gain after smoking cessation
Abstract
This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:
The primary objective is to review the evidence from two kinds of trials:
(i) The effects of interventions specifically designed to limit weight gain on two outcomes: weight gain and smoking cessation.
(ii) The effects of antidepressants, nicotine replacement therapy and exercise interventions on weight gain.
For weight gain in (i) and (ii), the effects will be examined only in those with sustained abstinence from smoking.
Secondary ojectives include:
(i) The effects of the interventions on weight gain and smoking cessation at the end of the programme and at least six months post‐cessation, to determine whether any effects on weight gain and cessation are temporary.
(ii) To examine evidence of interactions between body characteristics, gender, and psychological variables such as fear of weight gain on (a) smoking cessation and (b) weight gain.
Background
Smoking is the single biggest preventable cause of ill health, and recent estimates indicate that there are still over 10 million smokers in England (HDA 2004). Cigarette smoking causes lung cancer, respiratory disease, and heart disease, and contributes to a range of other diseases and conditions (DOH 1998). In 2000, 4.83 million premature deaths worldwide were estimated to be attributable to smoking, with about half of these occurring in developing countries (Ezzati 2003). In the UK, it is estimated that each year 86,500 deaths were attributed to cigarette smoking over the period 1998‐2002 (HDA 2004). Results from the General Household Survey in England indicate that 66% of smokers in England want to give up smoking (ONS 2003).
Although smoking cessation is associated with substantial health benefits, research has demonstrated that it is usually accompanied by weight gain (Klesges 1997; Spring 1995). In the USA it is estimated that 80 percent of people who quit smoking gain weight (USDHHS 1990). In the Lung Health Study over a five year follow‐up period, O'Hara 1998 found that among people who sustained quitting for five years, women gained 5.2 kg in year one and a mean of 3.4 kg in years one to five, while men gained a mean of 4.9 kg in year one and a mean of 2.6 kg in years one to five. In another study, Williamson 1991 relates changes in body weight to changes in smoking status in adults who were weighed in the First National Health and Nutrition Examination Survey (NHANES I 1971‐5) and then weighed a second time in the NHANES I Epidemiologic Follow‐up Study (1982‐4). In the cohort that included continuing smokers and those who had quit smoking for a year or more, they found that the adjusted mean weight gain attributable to smoking cessation was 2.8 kg in men and 3.8 kg in women. They also found that 9.8 percent of men and 13.4 percent of women had a major weight gain greater than 13 kg.
Among smokers there is a high prevalence of concerns about weight gain post‐cessation, which has been cited as a primary reason for not trying to quit smoking by weight‐concerned smokers, especially in women (Clark 2004; Klesges 1989; Klesges 1992). Moreover, weight gain experienced during or after smoking cessation was found to affect withdrawal from treatment, and was associated with relapse in ex‐smokers (Klesges 1989; Klesges 1992).
Different smoking cessation interventions have been developed to promote smoking cessation and simultaneously control weight gain in challenging populations, such as weight‐concerned smokers. They include behavioural interventions, such as exercise and calorie restriction or eating advice. Dietary interventions might serve to encourage reluctant quitters to try to stop smoking if they can be reassured that weight gain might be limited. However, such interventions might also risk undermining the success of the quit attempt. There is evidence that hunger and cigarette cravings are related, and that hunger can undermine quit efforts (Hall 1992). Moreover, early weight gain was found to promote rather than harm long‐term cessation (Gritz 1988; Hall 1986; Hughes 1991).
There are a range of other health technologies for smoking cessation that have been developed independently of concerns about weight gain, with the sole aim of assisting smoking cessation. However, some of these, such as antidepressants, nicotine replacement therapy, and exercise might also plausibly influence weight gain as well as smoking cessation. The effects of these interventions on smoking cessation are evaluated in the relevant Cochrane reviews, but the effects on weight gain are summarised only in the exercise intervention review (Ussher 2005). The effects of these medications on weight gain will therefore be included in our review.
Objectives
The primary objective is to review the evidence from two kinds of trials:
(i) The effects of interventions specifically designed to limit weight gain on two outcomes: weight gain and smoking cessation.
(ii) The effects of antidepressants, nicotine replacement therapy and exercise interventions on weight gain.
For weight gain in (i) and (ii), the effects will be examined only in those with sustained abstinence from smoking.
Secondary ojectives include:
(i) The effects of the interventions on weight gain and smoking cessation at the end of the programme and at least six months post‐cessation, to determine whether any effects on weight gain and cessation are temporary.
(ii) To examine evidence of interactions between body characteristics, gender, and psychological variables such as fear of weight gain on (a) smoking cessation and (b) weight gain.
Methods
Criteria for considering studies for this review
Types of studies
Randomized controlled trials.
To meet primary objective (ii), we will examine weight changes reported in trials included in the following Cochrane reviews. Trials will be included if they provide data on weight changes or if such data can be obtained from the authors:
-
Antidepressants for smoking cessation (Hughes 2004)
-
Nicotine replacement therapy for smoking cessation (Silagy 2004)
-
Exercise interventions for smoking cessation (Ussher 2005)
Types of participants
Adult smokers attempting to quit smoking.
Types of interventions
Any interventions that are designed specifically for limiting weight gain during and after smoking cessation. We will also examine the effects of specific pharmacological and behavioural interventions that are not designed primarily to limit post‐cessation weight gain but which might plausibly influence it, e.g. nicotine replacement therapy, antidepressants, and exercise. As the effects of these interventions on quitting are already documented in Cochrane reviews, our aim will be primarily to examine their effects on weight.
Types of outcome measures
There will be two primary outcome measures:
(i) Changes in body weight (as percentages and as absolute differences where possible) from baseline, at the end of treatment and at least six months from the quit date.
(ii) Smoking status at least six months from the start of treatment in participants smoking at baseline for those trials that report specific combined interventions.
For studies designed to limit weight gain that are not included in other Cochrane reviews, both outcomes will be fully examined. For studies that examine the effects of interventions that might plausibly influence weight gain and where the effects of these interventions on quitting are already described in other Cochrane reviews, the smoking cessation outcomes will be briefly reported, and the weight change outcomes in full.
Search methods for identification of studies
We will search the Cochrane Tobacco Addiction Group's Specialized Register, using the following search terms in title, abstract or keywords: food, calorie restrict*, intake, diet*, body mass index, bmi, Quetelet, waist‐hip ratio (WHR), weight, body‐weight, weight‐changes. The specialized register includes trials indexed in MEDLINE, EMBASE, PsycINFO and Web of Science, together with hand searching of specialist journals, conference proceedings, online registers of controlled trials and reference lists of previous trials and overviews. In addition, we will perform ad hoc searches of MEDLINE, EMBASE, and PsycINFO databases to identify studies for possible inclusion in this review. We will also contact authors for additional data or information where necessary.
Data collection and analysis
Two authors (MS and AP) will search for relevant trials, will obtain full copies of trials based on abstracts where available, and will select which trials to include in the review We will also conduct data extraction on each paper independently. We will resolve any disagreements by discussion with all authors, or by recourse to the editorial base. For each included trial we will extract information about the methods used, the participants, the interventions and the outcome measures, and report them in the table 'Characteristics of Included Studies'. The following data will be recorded where possible:
Methods
-
Method of selection of participants, in particular whether or not participants were concerned about weight gain, had had problems with weight gain in previous quit attempts, or were overweight when selected.
-
Methods of randomization and allocation concealment.
-
Blinding of investigators, participants and outcomes assessors, noting that in some studies blinding may not be possible.
Participants
-
Country and setting.
-
Number of participants in each comparison group.
-
Definition of 'smoker' used.
-
Demographic characteristics of participants: e.g. ethnicity, average age, sex, and average cigarettes smoked per day, Fagerstrom Test for Nicotine Dependence (FTND) score.
Interventions
-
Detailed description of the intervention, including intensity, frequency and duration.
-
Dose and duration of pharmocotherapy, if used.
-
Who delivered the intervention?
-
What was given to control group? e.g. placebo, sham exercise, dietary instruction.
-
Attendance/adherence to intervention and control programmes
-
Attrition rate
-
Drop‐out (lost to follow up) rate
Outcome Measures
For all included studies:
-
Quit rate in the intervention and control groups at least six months from the start of the trial.
-
Average change in body weight at least six months from baseline for quitters and non‐quitters.
-
Number of patients with follow‐up data.
-
Number of drop‐outs.
-
Any adverse events.
For studies designed primarily to limit post‐cessation weight gain and not included in other Cochrane reviews, we will also report the following outcomes:
-
Definition of abstinence from smoking used
-
Methods of validating abstinence.
Analysis
We will rate the quality of the included trials on methods of randomization and allocation concealment, as described in the Cochrane Handbook for Systematic Reviews of Interventions (Handbook 2005). Trials with adequate allocation concealment will be coded A. Code B will be assigned for trials with unclear allocation concealment, and code C for those with inadequate allocation concealment or where such concealment was not used.
We will identify any inconsistency in results across studies (heterogeneity) using the I2 test (Higgins 2003). If there is no significant heterogeneity, we will perform meta‐analysis, using the Mantel‐Haenszel fixed‐effect method for pooling results on smoking status, and a fixed‐effect method that uses the generic inverse variance approach for pooling results on weight change. If heterogeneity is present, we will perform random‐effects meta‐analysis if we cannot explain the heterogeneity by some characteristics of the studies.
We will prepare a funnel plot to detect evidence of possible publication and/or reporting bias.
1.Smoking Status.
We will compare smoking status between the intervention and control groups at least six months from the start of treatment. We will prefer biochemically verified results over self‐report, and continuous (complete) or prolonged abstinence (occasional lapse in the first two weeks) over point prevalence. We will use the odds ratio as our summary statistic in the meta‐analysis. We will perform a full intention‐to‐treat (ITT) analysis, which will include all participants randomized to their original groups. We will regard participants who drop out or are lost to follow up as continuing smokers.
2. Weight Change.
The mean difference in body weight will be used as a summary statistic for the treatment effect on weight.
To assess the impact of smoking cessation interventions on changes in body weight we will consider the following outcomes across each group:
-
Average change in body weight from baseline in those exhibiting prolonged abstinence during or at the end of treatment, and for at least six months from the quit date, and at the longest follow‐up, to determine whether effects on weight gain are temporary.
-
Average change in body weight from baseline in those who fail to quit.
We hypothesize that most information will come from studies of those with prolonged abstinence, as those who revert to smoking or with only point prevalence abstinence at follow up may well gain little or no weight. If we considered the mean weight change for an active treatment such as nicotine replacement therapy (NRT) versus placebo, it may well be the case that those who quit on NRT gain less weight than those who quit on placebo. However, as there are likely to be more people continuously abstinent in the NRT group than in the placebo group, the mean weight change overall may well be greater for the whole group on NRT than on placebo, even though the mean weight change for the quitters on NRT might be less than for the quitters on placebo.
In light of this, we will report as a primary outcome weight change in those participants who were continuously abstinent, and the proportion of this group who have missing data. We will also report the mean change in weight for all those who did not achieve continuous abstinence, and for the whole group as allocated. The proportion of people with missing weight information is likely to be greatest for those who did not achieve abstinence, and these data could be biased by this. We will use data from studies with good follow‐up information to derive estimates for those with missing data and impute these, by way of sensitivity analysis. We anticipate that many studies will not provide mean weight change and standard deviations for subgroups defined by whether or not they achieved continuous abstinence; in such cases we will attempt to obtain the data from the authors. We will follow the same procedure to get estimates of any effect modification by gender and baseline body characteristics.
Where appropriate, we will try to categorize specific weight gain prevention interventions according to their nature, intensity and length and assess the effect of each category on quitting and weight.
The absolute mean difference in body weight will be used as a summary statistic for the size of the effects of interventions on weight. On each paper, if the absolute mean difference in body weight is not reported explicitly, we will calculate it by subtracting the baseline mean weight from the post‐intervention mean weight for the intervention and control groups. If the standard deviations for the changes in body weight are not present, we will calculate them using any present statistical analyses comparing the changes themselves (e.g. confidence intervals, t‐values or P‐values) following methods described in the Cochrane Handbook. To calculate standard deviations of the changes in weight from their associated confidence intervals for studies with large sample size, we will use the following formula:
(SD = √N x (upper limit ‐ lower limit) /standard error wide).
For studies with 95% confidence intervals for difference in means we will divide by 3.92 standard errors wide, 3.29 for studies with 90% confidence intervals and 5.15 for studies with 99% confidence intervals. If sample size is less than 60, the above standard error wide numbers will be replaced with numbers specific to both the t‐distribution and the group sample size minus 1.
To calculate the standard deviations (SDs) of the changes using their correspondence t‐values, we will use the following formula:
Standard deviation = (standard error in difference in means)/√((1/NE) + (1/NC))
[N E and N C are the sample sizes in the two groups]
The standard error of difference in means will be calculated by dividing the difference in means by the t‐value.
P‐values obtained from t‐tests also can be used to calculate the standard deviations of changes in means. We will obtain the t‐value that corresponds to the given P‐value from the t‐distribution table and then use the above formula to calculate the standard deviations. If the P‐values are reported as less than a certain value, we will take that value as a conservative estimate of the true P‐value.
In some cases, studies may present both starting and final mean weights for participants, but no information from which it would be possible to calculate the SDs as described above. In that event, we will calculate the summary statistic as final mean weight minus baseline mean weight. For studies that do not report standard deviations for changes in means where statistical analyses comparing the changes themselves are not presented, we will impute the standard deviations for the changes following techniques described by Follmann (Follmann 1992). To calculate those standard deviations, we will estimate the correlation coefficient, which describes how similar the baseline and final measurements were across participants, from any other included studies that report standard deviations for mean weight at baseline, final measurement, and changes in means. To estimate the correlation coefficient for the intervention and control groups, we will use the following formula:
r = (SD (B)2 + SD (F)2 ‐ SD (C)2) / (2 X SD(B) X SD (F)).
We will use the imputed correlation coefficients from each group to calculate the missing standard deviations for changes in means for the intervention and control groups in other studies by using the following formula:
SD (C) = √((SD (B)2 + SD (F)2) ‐ (2 X r X SD (B) X SD (F))
[where r= correlation coefficient, SD= standard deviation for the changes in means, B= baseline, F= final measurement, and C= change in mean weight measurement.]
We will report any adverse events described in trial reports.
