Stopping Smoking Shortly Before Surgery and Postoperative Complications: A Systematic Review and Meta-analysis

Cigarette smoking has been implicated as a risk factor for postoperative complications across a spectrum of surgical specialties.^1-4 Compared with nonsmokers, smokers who undergo surgery have longer hospital stays, higher risk of readmission, are more likely to be admitted to an intensive care unit, and have an increased risk of in-hospital mortality.^5-7 Although stopping smoking before surgery can mitigate these risks^8,9 and longer periods of preoperative abstinence confer stronger benefits, the existing data do not allow clear advice on an optimal period of preoperative smoking cessation.¹⁰

The picture is complicated by a concern that has been circulating over the past 2 decades that stopping smoking within a few weeks before surgery may not just be ineffective in reducing postoperative complications, but that it can contribute to them. This concern originated from a 1989 article that found postoperative pulmonary complications in 6 of 18 continuing smokers, compared with 12 of 21 ex-smokers who quit for less than 8 weeks prior to surgery.¹¹ The report did not include statistical analysis, but the authors suggested that losing the cough-promoting effect of cigarettes before any improvement in sputum clearance might predispose to retention of secretions and postoperative pulmonary complications. Although the difference between the 2 groups was not statistically significant (χ² = 2.2; P = .20), the warning has in some instances become accepted as a proven fact. For example, an influential guidance document from the London Health Observatory states that “Cessation should occur at least 8 weeks prior to surgery to minimize the increase in pulmonary complications in recent quitters.”^12(p13) The 8-week cutoff point has also been recommended by other sources.^13,14 Patients are often scheduled for operations at relatively short notice, and an opportunity to discuss smoking may arise fairly late. Clinicians faced with smoking patients, or even with patients who proactively ask for help with stopping smoking, are often unsure whether they should provide smoking cessation treatment shortly before surgery. The National Institute of Clinical Excellence, which provides guidance to the English National Health Service, has recently convened a meeting of experts to consider revoking its general recommendation to provide smoking cessation advice to preoperative patients. This was motivated by concerns about possible harm to patients who would be encouraged to quit less than 8 weeks before their operation. Similar concerns are anecdotally expressed by anesthetists and surgeons globally.

Our preparatory examination of the existing literature on this topic identified 2 important methodologic issues. First, most existing studies focus on comparisons of early quitters (usually, those smoke free for more than 2 months before their surgery) and recent quitters (those smoke free for only a few weeks or up to 2 months before their surgery) with never smokers.¹⁰ Of these 3 groups, recent quitters often have the poorest outcomes. This seems to form one of the sources of warnings about recent quitting. However, showing that recent quitters have more complications than early quitters and/or never smokers may simply mean that recent quitting is less beneficial than early quitting. Only a comparison with continuing smokers can show whether recent quitting poses a risk.

The second issue concerns biochemical validation of self-reported abstinence. Hospital patients are often acutely aware that smoking may have contributed to their illness; consequently, they worry about the disapproval of clinical staff and tend to misreport their smoking status.^15,16 If the sample of patients classified as recent ex-smokers contains a proportion who are in fact still smoking, this is likely to dilute any potential risks or benefits of recent quitting. Compared with studies based on self-reported smoking status, studies that objectively validate self-reported abstinence from smoking provide more reliable evidence. The 2 validation methods commonly used, salivary cotinine levels and end-expired carbon monoxide readings, are both reliable.¹⁷

We present herein an analysis of the existing literature that takes into account these methodologic issues. The study sought to determine the following: Is there any evidence that stopping smoking within 8 weeks before surgery is associated with postoperative complications?

We identified and analyzed all existing studies that allowed comparisons of postoperative complications in patients who stopped smoking 8 weeks or less prior to surgery (recent quitters) with those who continued to smoke. All types of postoperative complications were considered from all specialties and populations. All study designs were included if they provided data on postoperative complications (as defined by the authors in that specialty) in both recent quitters and continuing smokers.

Electronic databases were examined using the KA24 (Knowledge Access 24 hours a day) Internet-based resource. The databases included the British Nursing Index (BNI) (from 1985), the Cumulative Index to Nursing and Allied Health Literature (CINAHL) (from 1981), The Cochrane Library database (from 1800), Embase (from 1980), Medline (from 1950) and PsycINFO (from 1806), all searched through May 2010. The search terms were developed within the study team.^18-21 The following search terms were used: (smoking AND cessation) or (tobacco AND cessation) or (stop$ AND smok$) or quit$ or (smoking AND status) and (post-op$ or postop$ or pre-op$ or preop$ or peri-op$ or periop$) and (complication$ or morbidity or mortality or pulmonary). The search was limited to electronic databases and to articles in English.

Figure 1 shows the flow of studies through the reviewing process. One reviewer filtered titles and abstracts of the articles returned in the search for information on postoperative complications (complications were defined by the authors in that specialty and so differed depending on surgical procedure) in recent quitters (stopped smoking within 8 weeks before surgery) and continuing smokers. Reports identified as potentially relevant were screened by a second reviewer. Hard copies were obtained of all 57 articles classified as potentially relevant. A hand search of the references from these articles identified a further 3 publications. The 60 articles were examined independently by 2 of us (K.M. and H.M.). Their conclusions were checked by a third reviewer (P.H.), who also reconciled any differences. The 9 articles that included interpretable data were rated for methodologic quality independently by the first 2 reviewers, and their conclusions were again checked by the third reviewer, who reconciled any differences.

For each study we extracted the number of postoperative complications for both continuing smokers and recent quitters. We also extracted data on the study period, duration of abstinence in recent quitters, whether their smoking status was biochemically validated, study design, the type of surgery, and the postoperative complications that were assessed (Table).

Where a study collected relevant data but did not report them in a form usable for our analysis, we contacted the authors. David Lindström, MD, PhD, provided data on pulmonary complications that were not reported in his 2008 article.²⁵

The quality of the included articles^{1,2,5,11,22-26} was assessed according to national guidelines for undertaking systematic reviews²⁷ and using indicators of susceptibility to bias specific for the purpose of the review. The key issue for our purpose concerns the reliability of classifying patients as recent quitters. Misclassification of smoking status presents by far the most serious risk of bias. Real differences could be diluted or even masked if the classification of abstinence was inaccurate. Studies that validated self-reported abstinence biochemically thus present the best evidence available. This consideration is included in the Table. It reflects the reliability of data of interest to the present review rather than the overall quality of the studies, which had mostly different purposes and priorities.

The included studies differed in a number of other ways, but we are not aware of any other feature that would exaggerate or diminish the difference between the 2 study groups in a systematic manner. The included studies used a range of study designs, but the key comparison of postsurgical complications in patients who stopped smoking shortly before surgery and those who continued to smoke is largely independent of study design and not affected by it. For example, while 2 studies randomized smokers to either a stop-smoking intervention or a control procedure, some smokers in the control group stopped smoking and many in the intervention group did not. For our purpose, the comparison of the randomized groups is less informative than the comparison of quitters and continuing smokers across both conditions (both studies provide the necessary information). For these reasons, no quality points were assigned to study designs. The different designs would not be expected to differ in accuracy of detecting postsurgical complications or produce any systematic bias in reporting complications in continuing smokers and recent quitters. Similarly, as the subanalysis of interest to us was not the main focus of the included studies, publication bias was unlikely: there was no obvious reason that would lead researchers or publishers to prefer one result over another.

Data extracted from the included studies were entered into the RevMan (Review Manager) program (version 5.0) developed for Cochrane meta-analyses. Analyses of adverse effects typically use the Peto odds ratio, which is the preferred statistic for rare occurrences, but in this case the rate of complications was high (over 50% in some studies), and we calculated a pooled risk ratio using a fixed-effects model. We assessed statistical heterogeneity using the I² statistic. Where there was significant heterogeneity, a random-effects model was used.²⁸ We conducted 3 meta-analyses. The first meta-analysis included all available studies to check for any effects of recent quitting, beneficial or detrimental. We repeated this analysis using only studies with validated self-reported abstinence and least risk of bias. Finally, we analyzed separately studies that focused specifically on pulmonary complications to assess possible detrimental effects in this particular area.

A total of 889 patients participated in the relevant studies. The characteristics of the 9 studies that met the inclusion criteria are listed in the Table.

Two additional studies were close to meeting the inclusion criteria. One of these reported data separately for patients who stopped smoking 2 to 4 weeks prior to the operation and those who quit earlier.²⁹ Those quitting 2 to 4 weeks preoperatively did not differ significantly from continuing smokers in the incidence of postoperative pulmonary complications (54% vs 44%). However, the analysis included a comparison of smokers who stopped smoking within 2 weeks before the operation with continuing smokers. This makes the interpretation of the findings difficult, and the study is not included in the meta-analyses results reported herein. However, including this study in the relevant meta-analyses did not alter the results.

The second study concerned pulmonary complications following resection of lung cancer.³⁰ It reported data separately for patients quitting 2 to 4 weeks prior to the operation but also merged those who quit within 2 weeks before the operation with continuing smokers. The incidence of complications was not significantly different in recent quitters and continuing smokers (6.2% vs 6.9%). The study is not included in the meta-analysis, but its inclusion does not affect the results.

Two other excluded studies require a special mention because they have been cited as demonstrating risks of recent quitting and an explanation of such risks.¹² Bluman et al¹³ found that 36 smokers who self-reported reducing their smoking rate in the weeks to months prior to surgery had a higher risk of postoperative pulmonary complications than 105 who reported smoking at their usual rate (relative risk, 6.7; 95% confidence interval [CI] 2.6-17.1). Self-reported smoking reduction may not reflect any real change in the inhaled volume of smoke. Even when there is a genuine reduction in the number of cigarettes, the reduction is usually undermined by compensatory smoking^31,32; this finding is therefore difficult to interpret.

Yamashita et al³³ compared intraoperative sputum production among recent quitters, continuing smokers, and nonsmokers.³³ The study is not included in our meta-analysis because volume of intraoperative sputum production is not a surgical complication. Patients were categorized into having or not having a high sputum volume. Those abstaining from smoking for less than 2 months preoperatively were more likely to have a higher sputum volume than nonsmokers (23% vs 9%; P < .01), but there was no significant difference between recent ex-smokers and current smokers (23% vs 18%).

Only 1 of the 9 included studies reported a significant result, and this was in favor of recent abstainers.¹ When all 9 studies are combined, there is no beneficial or detrimental effect of quitting within 8 weeks before surgery compared with continued smoking (Figure 2). Because there is substantial heterogeneity between these studies, the result needs to be interpreted with caution.

Figure 3 presents the results of the 3 reports that validated self-reported abstinence and had the highest quality scores.^1,11,25 These studies provide the best evidence available. The results again show no significant benefit or detrimental effect of recent quitting compared with continued smoking, but there is once more substantial heterogeneity in the data.

The effect of smoking cessation on postoperative pulmonary complications is shown in Figure 4. The results are homogeneous, and they show no significant increase in risk in those who stopped smoking less than 8 weeks prior to surgery compared with those who continued smoking.

The present analysis does not support the suggestion that quitting smoking less than 8 weeks before surgery has a negative impact on surgical outcomes. A hypothetical explanation of the presumed risks of quitting before surgery, as suggested by Warner et al¹¹ and repeated by others,¹³ is based on an assumption that stopping smoking leads to a decrease in coughing and an increase in sputum production. Both of these claims remain unconfirmed. Smoking seems to suppress rather than enhance cough reflex sensitivity.^34-36 Stopping smoking was reported to lead to a mild increase in coughing in 2 studies^37,38 and a decrease in 2 other reports.^39,40 Regarding mucociliary clearance shortly after smoking cessation, we are aware of only 1 study in which intraoperative sputum production was measured. The investigation found no difference between recent quitters and continuing smokers.³³

There are several limitations to this systematic review. The results are based on observational data. Quitting behavior may be determined by the same factors that determine postoperative outcomes. For example, patients who stop smoking may be more likely to access postoperative care, perhaps masking a detrimental effect of recent quitting. Conversely, those with more serious illness may be more likely to stop smoking, possibly obscuring beneficial effects of recent quitting. Another potential limitation is that our search covered only studies in English.

Our meta-analyses combined different types of surgical procedures and various definitions of postoperative complications. Although no detrimental effect was found in any single investigation or the various study combinations, it remains possible that there are particular types of surgery and/or complications that may reveal a different result. Amalgamating different types of studies makes it necessary to interpret the results with some caution. This, however, does not disqualify the approach we took. Given the relatively limited number of studies available, as long as the appropriate caveats are kept in mind, such a meta-analysis, combined with a systematic review of all individual studies, is the only way to arrive at a conclusion based on all currently available evidence.

Two of the meta-analyses included substantial heterogeneity, and where this is the case, the review of individual studies is more informative. It is reassuring that no single study within either group showed a significant risk, and that in the key analysis concerning pulmonary complications, the test of heterogeneity was nonsignificant.

Apart from showing no evidence of risk of recent quitting on pulmonary or any other complications, the results also show no evidence of benefit in wound healing or other areas of postsurgical recovery. Stopping smoking shortly before surgery may be unrelated to its immediate aftermath, but smokers prompted to quit in this context are of course likely to benefit in the long term.

Studies that did not validate smoking status may misclassify smokers as abstainers, making it less likely that any beneficial or detrimental effects of quitting will be detected. Future studies should verify self-reported abstinence biochemically and report details of exactly how and when smoking status was established.

No data are available on the effects of only a few days' abstinence from smoking. Early abstinence generates more intense withdrawal discomfort,^41,42 but there is no clear rationale to expect this to translate into postoperative complications. Nevertheless, future studies should focus on smokers quitting within 1 week of surgery and report their results separately.

In view of these limitations and the relatively small number of available studies, our findings are necessarily only tentative and may be modified when more data become available. The combined number of participants in the relevant trials was 889, and a small detrimental or beneficial effect may still emerge with further extension of this work. This is of course true for most meta-analytical verdicts.

In conclusion, there is currently no suggestion, either from any single study or from combinations of studies, that quitting smoking shortly before surgery increases postoperative complications. Future studies should focus on patients with a very short duration of abstinence and should use biochemical validation of self-reported abstinence. In the meantime, until some new evidence of harm emerges, firm advice to stop smoking and an offer of smoking cessation treatment to those who need it can be provided to presurgical patients at any time.

Correspondence: Katie Myers, MSc, CPsychol, 55 Philpot St, London E1 2JH, England (k.myers@qmul.ac.uk).

Accepted for Publication: January 7, 2011.

Published Online: March 14, 2011. doi:10.1001/archinternmed.2011.97

Author Contributions: All authors had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Myers, Hajek, and McRobbie. Acquisition of data: Myers, Hajek, and McRobbie. Analysis and interpretation of data: Myers, Hajek, Hinds, and McRobbie. Drafting of the manuscript: Myers, Hajek, Hinds, and McRobbie. Critical revision of the manuscript for important intellectual content: Myers, Hajek, Hinds, and McRobbie. Statistical analysis: Myers, Hajek, and McRobbie. Obtained funding: Myers and McRobbie. Administrative, technical, and material support: Myers, Hajek, and McRobbie. Study supervision: Myers.

Financial Disclosure: Drs Hajek and McRobbie have received research funds from and provide consultancies to the following manufacturers of smoking cessation medications: Glaxo SmithKline, Novartis, Pfizer Global, and Johnson & Johnson.

Funding/Support: Dr McRobbie is supported by the UK Centre for Tobacco Control Studies, a UKCRC Public Health Research Centre of Excellence.

Additional Information: Analyses in RevMan software are available from Ms Myers.