Combination Therapy of Varenicline With Nicotine Replacement Therapy Is Better Than Varenicline Alone

A Systematic Review and Meta-analysis of Randomized Controlled Trials

Ping-Hsun Chang; Chien-Hsieh Chiang; Wei-Che Ho; Pei-Zu Wu; Jaw-Shiun Tsai; Fei-Ran Guo


BMC Public Health. 2015;15(689) 

In This Article


Search Strategy

We conducted a comprehensive search in November 2014. The databases included MEDLINE (from 1966 to November 2014), EMBASE (from 1966 to November 2014), (from 2000 to November 2014) and the Cochrane Library. Search terms were varenicline, nicotine replacement therapy (including nicotine patch, gum, inhaler, nasal spray, lozenge). We combined "varenicline" and "nicotine replacement therapy" by the Boolean operator "and" for screening in titles, abstracts and key words. Then the results were limited to "randomized controlled trial". After selecting articles, we searched the reference lists for relevant citations. We limited the language to English. We did not limit countries of publications.

Selection Criteria

Only published RCTs with an adult population were included. Trials had to investigate combination treatment of varenicline and nicotine replacement therapy. The required outcomes were abstinence rates with biochemical verification, safety profile, or tolerability of the therapy. Exclusion criteria included non-RCT studies, trials without outcome measurements, trials using smoking cessation medications but not aiming to stop cigarette smoking (eg. stop alcohol use or long term NRT use), and articles that the full-text was not available.

Study Selection

One author (WC) searched the electronic databases. The results were independently assessed by two authors (WC and PZ). The authors identified articles eligible for further review by screening the titles and abstracts. The second step of selection was based on the full-text of articles. The disagreements were resolved by consensus between authors. A standardized data extraction form was used to collect population characteristics, study inclusion and exclusion criteria, intervention details, and outcome data from each study.

Data Extraction

Information was extracted from each included trial on: (1) characteristics of trial participants (including age, sex, location, Fagerström Test for Nicotine Dependence (FTND) score), and the trial's inclusion and exclusion criteria; (2) type of intervention (including type, dose, duration and frequency of varenicline and NRT; behavioral counseling); (3) type of outcome measure (exhaled carbon monoxide, self-reporting), length of follow-up, adverse effects. One author (WC) extracted the data from included studies and the second author (PZ) checked the extracted data. Disagreements were resolved by discussion between the two review authors; if no agreement could be reached, it was planned a third author (PH) would decide.

Quality Assessment and Publication Bias

The quality of the studies was assessed by the Jadad score.[12] The score ranges from 0 to 5 according to randomization, blinding, and patient dropout. We assessed the risk of publication bias by funnel plots. Asymmetry in a funnel plot was considered as a risk of publication bias. We used Begg's rank correlation test and Egger's regression test for statistical verification of bias.[13] The test results were generated by Comprehensive Meta-Analysis Version 2 software (Biostat, Englewood, NJ, USA).

Statistical Analysis

Pooled ORs were used to compare the effects of treatments. We defined the early outcome as the quit rate assessed before or at the end of treatment completion. The late outcome was the quit rate assessed for a period of time after the end of treatment completion, majorly at 24 or more weeks. The case numbers of adverse events were aggregated and the event rates were expressed as percentages. We calculated the ORs and 95 % confidence intervals (95 % CI) of adverse events by fixed effect model. Between-study heterogeneity was estimated using the χ 2 -based Q statistic[14] and heterogeneity was considered statistically significant when P-value was less than 0.1. If heterogeneity was significant, the pooled estimate was calculated based on the random effects model.[15] In the absence of significant heterogeneity, the pooled estimate was calculated using the fixed effect model.[16] A statistical test with a P value less than 0.05 was considered significant in pooled estimates. The forest plots and pooled estimates were generated by Review Manager (RevMan Version 5.3 Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014).