We included RCTs of any duration beyond 4 weeks. RCTs had to compare NRT with an inert control (eg. placebo or standard of care). We chose 4 weeks to include the timeframe of maximum nicotine withdrawal symptoms so that adverse events may be differentiated from withdrawal symptoms 8 . We additionally sought out observational studies to examine the proportion of events occurring 9 . We evaluated adverse effects reported at any point in the duration of the studies. We included any form of NRT delivery (i.e. lozenge, skin patch, gum, nasal spray, inhaler, and tablet). We did not examine efficacy in this analysis. We excluded post-hoc analyses, maintenance therapy, or relapse prevention studies.

In consultation with a medical librarian, we established a search strategy. We searched independently, in duplicate, the following 10 databases (from inception to November 20, 2009): MEDLINE, EMBASE, Cochrane CENTRAL, AMED, CINAHL, TOXNET, Development and Reproductive Toxicology, Hazardous Substances Databank, Psych-info and Web of Science. Given that observational studies are poorly indexed in many databases, we also searched databases that include the full text of journals (ScienceDirect, and Ingenta, including articles in full text from approximately 1700 journals since 1993) 10 . In addition, we searched the bibliographies of published systematic reviews and health technology assessments 4 5 6 11 12 13 14 15 16 . Searches were not limited by language, sex or age.

Two investigators (EM, PW) working independently, in duplicate, scanned all abstracts and obtained the full text reports of records indicating that the study was either an RCT or observational study evaluating NRT on the outcomes of interest. After obtaining full reports of the candidate studies (either in full peer-reviewed publication or press article) the same reviewers independently assessed eligibility via full text review.

Two reviewers (EM, PW) conducted data extraction independently using a standardized pre-piloted form. Reviewers collected information about the NRT intervention tested, the population studied (age, sex, underlying conditions), treatment dosages and dosing schedules, the specific measurement of abstinence (prolonged or point-prevalence), and the methods of biochemical confirmation. The reviewers extracted data on adverse events characterized by the study authors as physical or mental adverse events. Recognizing that adverse events may include both physical and mental effects concomitantly, we defined physical adverse events as effects confined to physical parts of the body and mental adverse events as symptoms accompanied by psychological conditions. We characterized serious adverse events as unexpected life-threatening events occurring during the trial period. A priori, we examined the follow life-threatening adverse events: all-cause mortality, myocardial infarction, all-cause strokes, incidence of all-cancers, all-hospitalizations, suicidal ideation, depression, and incidence of diabetes. Study quality evaluation included general methodological reporting features including allocation concealment, sequence generation, blinding status, intention-to-treat, and appropriate descriptions of loss to follow-up. In rating quality, failure to report a quality component of study design (e.g. blinding) was treated the same as not employing it. We entered the data into an electronic database such that duplicate entries existed for each study. When the two entries did not match, we resolved differences through discussion and consensus. In the absence of an inert control group, we considered randomized NRT dosing studies to be observational studies and collected data on proportion of combined events. For cohort studies, we additionally calculated events as proportions of events.

In order to assess inter-rater reliability on inclusion of articles, we calculated the Phi statistic (ϕ), which provides a measure of inter-observer agreement independent of chance 17 . We calculated the Odds Ratios [OR] and appropriate 95% Confidence Intervals [CIs] of outcomes according to the number of events reported in the original studies or sub-studies. Given that zero events in one treatment arm prevents a useful ratio from being developed, in circumstances of zero outcome events in one arm of a trial, we added 0.5 to each arm, as suggested by Sheehe 18 . We first pooled studies of all NRT interventions versus all controls using the DerSimonian-Laird random effects method 19 , which recognizes and anchors studies as a sample of all potential studies, and incorporates an additional between-study component to the estimate of variability 20 . We calculated the I² statistic for each analysis as a measure of the proportion of the overall variation that is attributable to between-study heterogeneity 21 . Given that we are examining adverse events, interpreting heterogeneity estimates can be challenging as even pooled analysis with large heterogeneity may provide important insights into the likelihood of events 9 . We considered an I² above 50% as moderate to large heterogeneity and examined explanations of heterogeneity by applying a random effects meta-regression with the following co-variates: oropharyngeal formulation vs. skin patch; duration of study (in months); reporting of allocation concealment, and reporting of blinding status. We then calculated the residual heterogeneity and present it as the residual I². Forest plots are displayed for each primary analysis, showing pooled study effect measures with 95% CIs, and the overall DerSimonian-Laird pooled estimate. For studies considered as observational studies, we calculated pooled weighted proportions by first stabilizing the variances of the raw proportions (r/n) using a Freeman-Tukey type arcsine square root transformation and applying a random effects model. While several methods of pooling proportions exist, the Freeman-Tukey method works well with both fixed and random effects meta-analysis and truncates at zero 22 . This is a variance-stabilizing transformation that removes the dependence of the variance on the mean of the transformed proportion (ie. it corrects for overdispersion). Assessing heterogeneity in pooled proportions may be misleading 23 24 , therefore we report the I² value, as this measure is less affected by the number of studies as the more commonly used I². The square root of this number (i.e. tau [ τ ]) is the estimated standard deviation of underlying effects across studies 25 . As with the RCT analysis, we applied a random-effects meta-analysis. Analyses were conducted using StatsDirect (version 2.5.2) and Comprehensive Meta-analysis (version 2).

One hundred and twenty studies met our inclusion criteria (ϕ = 0.91) 26 27 28 29 30 31 32 33 34 35 36 37 38 39 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 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 110 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 . Figure 1 displays the search break-down. Additional file 1 and Table 1 provide study characteristics. Ninety-two studies 26 27 28 29 30 31 32 33 34 35 36 37 38 39 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 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 were RCTs involving 32,185 participants and 28 56 75 93 110 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 were considered as observational studies involving 145, 205 participants.

Of RCTs, eighty-three 26 27 28 29 30 31 32 33 34 37 38 39 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 66 105 106 107 108 109 65 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 87 88 89 90 92 93 94 95 96 97 99 100 101 102 112 114 115 116 117 142 (83/92) used a placebo control. Forty-two 39 40 41 43 45 47 48 49 51 52 53 54 55 60 61 64 66 68 69 70 71 72 74 75 76 77 81 83 87 89 90 92 93 98 99 102 103 105 106 107 112 115 evaluated the nicotine patch; 26 26 27 28 29 30 31 32 33 34 35 36 37 38 42 56 58 79 88 95 100 101 108 109 113 116 142 the nicotine gum; 6 44 50 57 65 73 111 the nicotine nasal spray; 6 46 62 63 67 78 96 the nicotine inhaler; 4 80 84 97 114 the nicotine tablet; 1 85 the nicotine lozenge; and, the remainder (n = 35) evaluated NRT combination therapies. Duration of treatment ranged from 1 to 24 months, with varying levels of dosages for each form of NRT.

Seventy-four RCTs 26 27 28 29 30 31 33 34 35 36 37 38 39 40 41 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 67 68 69 70 72 73 74 75 76 77 78 79 80 82 83 84 85 86 87 88 92 93 95 96 99 100 101 103 106 107 108 109 114 115 116 117 were conducted in healthy adult populations. An additional 6 36 66 89 97 105 113 were conducted among populations with medical and psychiatric co-morbidities (.eg. smoking-related diseases, chronic diseases, alcoholism, depression), 4 81 98 104 142 among pregnant women, 3 42 71 91 among hospitalized patients, 3 90 94 111 among adolescents, 1 102 among postmenopausal women, and 1 112 among surgical patients. Fifty-nine 27 28 29 30 33 36 37 38 41 42 45 48 49 50 51 52 53 54 55 56 58 59 61 62 63 66 67 71 73 74 76 77 79 81 83 86 89 90 91 92 94 97 98 99 100 101 102 104 105 106 107 108 111 113 114 115 116 117 142 RCTs included co-interventions, of which 20 27 36 49 50 51 52 54 61 71 79 81 91 98 100 101 102 111 114 116 117 provided general counseling (eg. group counseling or individual counseling), 19 28 41 53 56 61 63 67 76 85 89 90 92 94 99 104 105 106 107 113 provided behavioural or psychological treatment, 12 29 30 33 37 42 45 55 62 66 74 86 97 provided varying forms of advice or support, 3 38 108 142 provided educational sessions, 4 58 59 73 77 provided an additional NRT and/or placebo, and 1 115 provided rimonabant, an endocannabinoid antagonist used as an appetite suppressant.

Seventeen observational studies 75 93 110 119 120 121 122 123 124 126 127 128 129 131 132 133 141 used the nicotine patch; 2 125 135 used a nasal spray, 1 129 used the nicotine gum; and 8 used a combination of NRTs. Duration of treatment ranged from 4 to 26 weeks, with varying levels of dosages for each form of NRT. Sixteen 56 75 93 110 118 122 123 125 126 127 129 130 132 133 134 137 of the observational studies were conducted among healthy adult populations, 7 120 121 131 135 138 139 140 among mixed healthy and unhealthy adult populations, 2 136 141 among adult populations with medical co-morbidities, 2 124 128 among adolescent populations, and 1 119 among an older adult population. The majority of observational studies 110 119 120 121 122 123 124 125 126 128 133 134 56 118 127 130 132 137 138 (19/28) included co-interventions, with 12 110 119 120 121 122 123 124 125 126 128 133 134 providing general counseling; 3 56 130 137 providing counseling specific to behaviour or behaviour modification; 3 118 132 138 providing educational sessions; and 1 127 providing a self-help booklet.

Studies reported methodological issues variably. Thirty-eight RCTs 28 29 31 37 41 44 46 59 61 62 63 65 70 71 72 73 74 75 77 78 81 82 84 85 90 93 97 98 99 101 102 104 105 112 115 116 117 142 reported sequence generation of randomization; 17 28 29 31 37 59 63 67 69 71 73 74 75 81 93 98 99 115 reported allocation concealment; and, 81 26 27 28 29 30 31 32 33 34 37 38 39 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 66 105 106 107 109 65 67 68 69 70 71 72 73 74 75 76 77 78 79 80 82 83 84 85 87 88 89 90 92 93 94 95 96 97 99 100 101 102 112 114 115 116 117 142 reported on patient blinding in the study. To confirm smoking abstinence, eighty four 26 27 28 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 66 105 107 108 109 65 67 69 70 71 72 73 74 75 76 77 78 79 80 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 100 101 102 103 104 111 112 114 115 116 117 142 studies used exhaled carbon monoxide (CO); 16 34 35 37 41 45 55 57 62 63 79 81 90 94 107 111 142 used salivary cotinine; 5 29 36 47 60 101 used serum cotinine; and 2 99 114 used urinary cotinine. No studies required participants to pay for the NRT. In 80 studies, 26 27 28 29 31 32 33 34 35 36 37 38 39 40 41 43 44 45 46 47 48 49 50 51 52 53 54 56 57 58 59 60 61 62 63 64 66 105 106 107 108 109 113 65 67 68 70 71 72 74 75 76 77 79 80 83 84 85 86 87 89 90 92 93 94 97 98 99 100 101 102 103 104 111 112 114 115 116 117 142 participants were planning on quitting. All observational studies are considered as non-comparative single-arm studies reporting prevalence of the adverse events in the community.

We evaluated all forms of NRT for adverse events. Additional file 1 displays the study characteristics and table 2 displays the RCT findings.

We used meta-regression to explain heterogeneity. We found large heterogeneity (I² = 62%) in our analysis of pooled events of nausea and vomiting. We were able to explain a large amount of heterogeneity examining the covariate of skin patch vs. oropharyngeal administration (OR 0.73, 95% CI, 0.56-0.97, P = 0.02, I² = 42%) as well as reporting of allocation concealment (OR 1.50, 95% CI, 1.07-1.59, P = 0.002, I² = 37%). We also found large heterogeneity (I² = 66%) in the event of mouth and throat soreness, but were unable to explain this using our pre-specified covariates. Considerable heterogeneity (I² = 72%) was also identified for the adverse event of coughing. We found that duration of study (B coefficient 0.07, 95% CI, 0.01-0.12, P = 0.01, I² = 42%) and allocation concealment (OR 0.53, 95% CI, 0.31-0.91, P = 0.02, I² = 43%) contributed to heterogeneity observed in this analysis. Our analysis of sweating found large heterogeneity (76%) that was predominantly contributed by a study that provided concomitant bupropion, indicating that the dual use of both drugs resulted in a significantly higher incidence of sweating (OR 29.24, 95% CI, 3.96-215.48). We also found heterogeneity in our analysis of insomnia that was explained by the duration of the trials. Longer duration trials had reduced rates of insomnia (B coefficient -0.07, 95% CI, -0.13 to -0.008, p = 0.02).

This review found that NRT is associated with an increased risk of gastrointestinal complaints and insomnia. There was also an observed increased risk of skin irritation with the nicotine patch and oropharyngeal complaints with orally administered NRT. Although NRT was associated with an increased risk of heart palpitations, the review did not observe an increased incident of heart attack or death. With the exception of insomnia, NRT does not appear to be associated with serious adverse psychiatric effects.

The reviewers actively sought serious adverse events. The most serious adverse event consistently reported in both RCTs and observational studies were heart palpitations and chest pains (OR 2.06, 95% CI, 1.51-2.82, P < 0.001). NRT has been implicated in reports of atrial fibrillation and myocardial infarction among patients with risk factors 143 144 145 146 147 148 . Several possible explanations for this exist. First, among patients using NRT who continue to smoke, high serum concentrations may stimulate the sympathetic nervous system, so increasing blood pressure, stroke volume and cardiac output 149 . Second, previous and current smokers may have established cardiovascular disease. Patients with unstable coronary syndrome, a common manifestation of coronary artery disease, may have unrecognized recent plaque ruptures including coronary vasoconstriction and increased strain placed on the heart due to tachycardia and hypertension 143 . Unofficial guidelines 149 caution the continued use of NRT in patients with known cardiovascular disease in the absence of a physician. They recommend that patients be warned of these risks and counseled to desist smoking and arrange intensive behavioral support 149 . In our analysis, we did not observe an increased risk for myocardial infarction or death from NRT.

Almost all studies demonstrated localized irritation related to NRT use, skin irritation with the use of NRT patch and mouth soreness and ulcerations with orally administered NRT. It is possible this is due to the success of cessation rather than NRT as mouth ulcers occur in about 40% of all individuals achieving tobacco abstinence regardless of cessation intervention 150 . It has previously been understood that mouth lesions are associated with smoking cessation and not NRT 150 . However, this review found a significantly increased risk of mouth ulcers with orally administered NRT users compared to inert controls who had ceased smoking. In order to prevent relapse due to treatment discontinuation, strategies should be developed to assist patients unable to continue oral administered NRT due to mouth ulcers such as increasing the nicotine patch dose, using the nicotine nasal spray, or switching to an alternative form of smoking cessation pharmacotherapy such as buproprion or varenicline.

Psychological adverse events, particularly suicidal ideation, are a major concern in patients initiating smoking cessation 151 . We found only one large retrospective observational study that discussed this topic and reported no significant difference in fatal and non-fatal self-harm associated with NRT compared to other frequently used pharmacotherapies, bupropion (HR 1.17, 95% CI, 0.59-2.32) or varenicline (HR 1.12, 95% CI, 0.67-1.88) 139 .

A criticism of smoking cessation trials has been that they infrequently enroll participants with psychological difficulties, thus making generalizable statements about their safety difficult 152 . The present review found that 56 RCTs specifically excluded participants with mental disorders. Only two RCTs targeting participants with concomitant psychological difficulties, specifically alcoholism and depression, were identified 89 92 . In these studies, the risk of insomnia was higher among those taking NRT, compared to controls, (OR 1.42, 95% CI, 1.21-1.66, P < 0.001). Sleep disturbance can significantly worsen psychological distress and mental illness and impair quality of life 153 . Therefore, clinicians should remain vigilant about NRT-related sleep disturbance among patients with a history of psychiatric illness.

An important issue to examine when describing adverse events from smoking cessation therapies is whether the adverse events are related to a pharmacotherapy or whether they are related to tobacco withdrawal 154 . For example, insomnia and sleep disturbances could be related to tobacco abstinence. One way to assess this affect would be to compare side effects in those that have quit smoking in both groups. However, since individuals that quit smoking may differ from those that continue, this analysis would remove the benefits of randomization and introduce bias.