|Home | About | Journals | Submit | Contact Us | Français|
The association of smoking with outcomes following breast cancer prognosis is not well understood.
In a cohort study called Life After Cancer Epidemiology (LACE), 2265 women diagnosed with breast cancer were followed for a median of twelve years. We used multivariable proportional-hazards models to determine whether smoking, assessed approximately two years post-diagnosis, was associated with risk of death among these women. We also undertook a systematic review of all cohort studies to date that have examined the association between smoking and breast cancer mortality.
Compared with never smokers, women who were current smokers had a two-fold higher rate of dying from breast cancer [hazard ratio (HR)=2.01, 95% confidence interval (CI) 1.27–3.18] and an approximately four-fold higher rate of dying from competing (non-breast cancer) causes (HR=3.84, 95%CI 2.50–5.89). Among seven studies that met the inclusion criteria in the systematic review, three studies and our own reported significantly increased risk of breast cancer death with current smoking. We found little evidence of an association between former smoking and breast cancer mortality (HR=1.24, 95% CI 0.94–1.64).
Consistent with findings from our prospective observational study, the systematic review of seven additional studies indicates positive association of current smoking with breast cancer mortality, but weak association with former smoking.
Women who smoke following breast cancer diagnosis and treatment are at higher risk of death both from breast cancer and other causes.
Breast cancer is a highly heterogeneous disease with large variations in survival. Identification of modifiable lifestyle factors that may improve survival is of interest to women diagnosed with breast cancer as well as their caregivers. Smoking has been associated with increased mortality following diagnosis of a variety of cancers, including prostate[2,3], colorectal and vulvar cancers, leukemia and malignant melanoma. Smoking after breast cancer diagnosis has been shown to adversely affect overall survival[8,9], whereas the association with breast cancer specific survival appears equivocal [10,11,9,12,13]. Smoking is associated with several factors that lead to poorer outcomes among women with breast cancer, including lower socioeconomic status[14,15], decreased physical activity[16,17] and comorbidity[18,19]. Comorbidity may be the most life-threatening issue among women with breast cancer who are current or former smokers because of an increased mortality risk from a spectrum of smoking-associated conditions and the fact that comorbidity can often lead to receiving less aggressive or less complete cancer treatment .
Wells hypothesized that inconsistent epidemiologic evidence for the association of smoking with breast cancer may reflect both adverse and protective effects of smoking. Evidence for the adverse effects comes from work implicating smoking in increased metastatic potential of cancer cells and promotion of tumor angiogenesis and growth; protective effects, however, may work through the anti-estrogenic effects of smoking. The hypothesis that smoking may induce earlier menopause thus reducing the risk of breast cancer has received inconsistent support [23–26]. Subsequent epidemiological studies have shown either no association or a small increase in risk of breast cancer associated with active smoking [27,25,28]. Some support for the association of smoking with breast cancer risk comes from a recent laboratory study indicating that nicotine receptor mediated carcinogenic properties are involved in biological functions associated with the development of breast cancer .
Based on the biological and epidemiological evidence that smoking may influence breast cancer progression, we hypothesized that smoking would be associated with both an increased risk of death from breast cancer as well as from other causes in our prospective cohort of more than 2,000 women with breast cancer. In this study of the relationship of smoking with mortality, we considered death from breast cancer, competing causes and all causes for women in the National Cancer Institute funded population-based cohort of predominantly early stage breast cancer survivors, Life After Cancer Epidemiology (LACE ). This large cohort of over 2,000 breast cancer survivors was well suited to examining the consequences of cigarette smoking following initial breast cancer treatment while taking into account known prognostic factors in the clinical, lifestyle-related and sociodemographic domains. We assessed the extent to which the impact of smoking on survival differed as a function of tumor subtype [estrogen receptor (ER), progesterone receptor (PR) and Human Epidermal Growth Factor Receptor 2 (HER2) status], body mass index and menopausal status. To place our study-specific results in the context of the current evidence base, we also undertook a systematic review of all cohort studies to date that have examined the association between smoking and breast cancer mortality.
The LACE cohort consists of women diagnosed with stage I (≥1 cm), II, or IIIa breast cancer from 1997 to 2000 in the Kaiser Permanente Northern California (KPNC) Cancer Registry or the Utah Cancer Registry. Eligible women were diagnosed, on average 21 months (range 9–39 months) prior to enrollment, had completed completed breast cancer treatment (except adjuvant hormonal therapy) and were free of any documented recurrence during that period. In addition to the KPNC and Utah cancer registries, this cohort included women screened and eligible for the Women’s Healthy Eating and Lifestyle (WHEL) study, a dietary intervention trial examining the prevention of breast cancer recurrence. A total of 2,586 (45.7%) completed initial enrollment; subsequent review to confirm eligibility left 2,270 women in the cohort. The large majority of cohort members (82%) came from KPNC, 12% from Utah, and 6% from WHEL. The upper age restriction for enrollment to the study was 79 years. Of 2,270 women included in the cohort, data on smoking status was available for 2, 258 participants (Table 1). This final sample formed the study population in the present analysis.
The Institutional Review Boards at the University of California, San Francisco and the KPNC approved this study.
Smoking status was determined from baseline questionnaire that was completed on average 23 months (ranging from 11 to 39 months) after breast cancer diagnosis. The questionnaire asked whether they currently smoked cigarettes and whether they had ever smoked in the past. “Never smokers” were women who answered “no” to both questions. Women were also asked if they consumed more than 100 cigarettes in their lifetime; we will refer to this measure as “ever-smoking”. Those that responded affirmatively were additionally asked the average number of cigarettes smoked per day.
A health status update questionnaire was mailed to LACE participants to monitor health outcomes semi-annually until April 2006 and annually thereafter until June 2012. The health status update asked women about any events that might have occurred in the preceding 6 months (or 12 months on the revised questionnaire), including recurrences or new primary breast cancer, hospitalizations, and other cancers. Reported events were confirmed by medical record review. Participant deaths were determined through KPNC electronic data sources, a family member responding to a mailed questionnaire, or by phone call to a proxy. All reported deaths were confirmed by death certificate review to verify primary and underlying cause of death. This information was then categorized as breast cancer death or non-breast cancer death. Outcomes were updated regularly by surveillance of KPNC electronic data sources and mortality files (including data from the States of California and Utah Departments of Vital Statistics and the U.S. Social Security Administration) for all participants, including those who dropped out (n = 90) or were lost to active follow-up (n = 15). In this analysis, the outcomes of interest were mortality from breast cancer, other causes and all causes.
The covariates in these analyses were socio-demographic, lifestyle-related, and clinical prognostic factors that, based on the existing literature and a priori reasoning, could potentially confound an association between smoking status and mortality. The sociodemographic covariates included age (calculated as the difference between date of breast cancer diagnosis and reported date of birth), race/ethnicity (white, non-white), and education (≤ 12 years, 13–15 years, >15 years). Lifestyle-related factors included body mass index (BMI) at diagnosis (calculated as weight in kilograms/height in meters squared, or kg/m2, from self-reported weight and height), physical activity at enrollment, and alcohol consumption (≤0.5 g/day, 0.6–5 g/day, ≥ 6 g/day). Standard BMI categories (normal weight, < 25; overweight, 25 – 30, and obese, ≥ 30) were used. Physical activity was assessed with a questionnaire based on the Arizona Activity Frequency Questionnaire . Standard metabolic equivalent task (MET) values were assigned to each activity and then frequency was multiplied by duration and MET value and summed over all activities (other than the sedentary recreational and transportation activities), providing a summary measure of total activity in MET hours per week . Two categories of physical activity, above and below the median in this group, were used (≤46 MET-hours/week, >46 MET-hours/week).
Prognostic factors were obtained from medical chart review and electronic databases for the LACE participants who were KPNC members and from medical chart review only for those who were not, and included menopausal status (pre-menopausal, post-menopausal, missing), lymph node involvement (0, 1–3, ≥4), estrogen receptor (ER) status (positive, negative, missing), progesterone receptor (PR) status (positive, negative, missing), and human epidermal growth factor 2 (HER2) receptor status (positive, negative, missing) from primarily immunohistochemistry, chemotherapy treatment (yes/no), radiation treatment (yes/no), tamoxifen use (never user, former user, current user, missing), and type of surgery (conserving or mastectomy). Stage at diagnosis (I, IIA, IIB, IIIA) was classified according to the Tumor, Node, Metastasis (TNM) system based on the criteria of the American Joint Committee on Cancer. Multivariate models for all-cause and non-breast cancer mortality were further adjusted for comorbidity. Patient-reported comorbid medical conditions were used as an indicator of the comorbidity burden; these conditions included thyroid disease, hypoglycemia, diabetes, hypertension, myocardial infarction, angina, peripheral arterial disease, gallbladder disease, diverticulitis, Crohn disease, pancreatitis, colorectal polyps, irritable syndrome, kidney disease, arthritis, osteoporosis, cirrhosis, stroke, and lupus. The comorbidity burden was estimated using the Charlson comorbidity index (CCI[35,36]). The CCI was derived from the number and type of underlying diseases present at study entry from patient questionnaire data and was categorized as a binary variable (0, ≥1).
Differences in means and proportions of selected covariates in the exposure groups were assessed using Student t test for continuous variables and Pearson χ2 for categorical variables. For categorical covariates with fewer than 5 participants in one or more categories, Fisher’s exact test was used instead of Pearson χ2. Kaplan-Meier plots were used to examine associations between smoking status and survival graphically and statistically respectively. Follow-up began at the time of diagnosis and ended at first confirmed date of death, depending on the specific analysis. Individuals who were alive were censored at the date of last contact (either most recent health status update questionnaire or electronic surveillance). Guided by a priori considerations , separate delayed-entry Cox proportional hazards models [38,39] with time since diagnosis as the time scale were used to estimate the risk of each outcome associated with smoking status and ever-smoking, accounting for varying times of enrollment into the cohort, and adjusting for covariates. Risk was expressed as a hazard ratio and 95% confidence interval (CI). The type I error was set at .05 and all reported P-values are two-sided.
After generating unadjusted Cox proportional hazards models for smoking status, known prognostic variables and those that showed significant relations with either the independent or dependent variable were added to the model (p<0.10). All Cox models were tested for proportionality of hazards graphically and statistically . In multivariate models, interaction terms were considered. All analyses were conducted in SAS version 9.2 and R version 2.15.0.
We identified papers published prior to July 1st, 2012 through a search in Medline (www.ncbi.nlm.nih.gov) and Google (www.google.com) using the following search terms: ‘smoking or tobacco’, ‘mortality or survival’ and ‘breast (neoplasm or cancer)’. We also performed a cited-reference search of retrieved articles and identified publications by review of the references in the retrieved articles. For each of the studies included in the review we abstracted characteristics of the study population; information on follow-up; outcome, exposure and covariate assessment; results and conclusions. We followed the PRISMA guidelines (http://www.prisma-statement.org; Transparent Report of Systematic Reviews and Meta-Analysis) as a methodological template for this review (see Figure 1 and Appendix 1).
We included studies with the following characteristics: (i) included women diagnosed with invasive breast cancer and of at least 18 years of age at diagnosis; (ii) included a measure of smoking status categorized as current-, past- and never-smokers; (iii) considered breast cancer-specific mortality as one of the outcomes; and (iv) were of cohort design.
We excluded studies that were published in languages other than English, and for which full text was not available. Intervention studies of smoking cessation in breast cancer survivors were also excluded.
A flow diagram of our literature search algorithm is provided in Figure 1. For each of the studies included in the review we abstracted characteristics of the study population; information on follow-up; outcome, exposure and covariate assessment; results and conclusions were.
Distributions of baseline characteristics for selected variables by smoking status are presented in Table 1. At study initiation 52.9% (1194/2258) of the study participants reported that they had never smoked, 39.5% (891/2258) were former smokers, and 7.7% (173/2258) were current smokers. The majority of the women were early-stage breast cancer survivors with 81% in stages I or IIa at the time of diagnosis. The median age was 58 years (SD=11.0 years; range 25–79 years) at the time of breast cancer diagnosis; 20% of the participants were non-white and nearly 27.4% had a high school level education or below. Never and former smokers tended to be older, and more educated than current smokers.
We noted a significant variation in the distribution of tumor stage (p-value=0.02): 83.2% of current, 81.9% of former and 78.7% of never smokers had stage I or IIa disease. Current and former smokers were more likely to have HER2 negative tumors compared to never smokers (78%, 79.7% and 71.9% respectively, p-value=0.002). Significant differences were also observed in menopausal status with 60.1% of never, 69.7% of former, and 56.6% of current smokers reporting post-menopausal status (p-value<0.0001). Former and current smokers were more likely to consume larger quantities of alcohol than never smokers: 28.3% of current smokers and 30.0% of former smokers consumed more than 6 grams of alcohol per day compared to 13.5% of never smokers (p-value<0.0001). Never smokers were also more likely to receive chemotherapy than former or current smokers (60.6% vs. 52.6% and 56.6% respectively, p-value=0.002).
The median follow-up in our analytic cohort of 2,258 women was 12.3 years (standard deviation=2.9 years; range 1.5–15.5 years). The median follow-up among current smokers was significantly shorter than the median follow-up among former and never smokers (11.9 vs. 12.2 vs. 12.4 years respectively; p=0.0005). During this period, a total of 485 deaths was observed: 215 among 1194 never smokers, 213 among 891 former smokers, and 57 among 173 current smokers. Of these 485 deaths, 241 of deaths were due to other causes (105 among never smokers, 105 among former smokers, and 32 among current smokers) and 244 were due to breast cancer (111 among never smokers, 108 among former smokers and 25 among current smokers).
Kaplan-Meier plots of survival by smoking status for all-cause survival (Figure 2a), competing-cause survival (Figure 2b), and breast cancer-specific survival (Figure 2c) revealed differences by smoking status in all-cause, competing-cause, and to lesser extent in breast cancer survival. In Table 2, we present results indicating that current smokers had significantly higher risk of all-cause, competing-cause and breast cancer-specific mortality than never smokers. Current smokers had an increased risk of death from any cause in unadjusted (HR=2.03, 95% CI 1.51–2.72), and covariate-adjusted models (HR=2.63, 95% CI 1.93–3.58). Former smoking was also associated with increased risk of all-cause mortality in both unadjusted (1.38, 95%CI 1.14–1.67) and covariate-adjusted models (1.28, 95%CI 1.05–1.56). Both current and former smoking were associated with increased risk of competing-cause mortality, with covariate-adjusted hazard ratios of 3.84 (95%CI 2.50–5.89), and 1.33 (1.00, 1.78), respectively. Compared to never smokers, current smokers had an approximately two-fold increase in the risk of breast cancer death in unadjusted (HR=1.71, 95%CI 1.11–2.64), and covariate-adjusted models (HR=2.01, 95%CI 1.27–3.18). Former smoking was associated with increased risk of breast cancer mortality in unadjusted models (HR=1.35, 95%CI 1.04–1.76). After adjusting for all potential confounders, the association was attenuated and did not reach statistical significance (HR=1.24, 95%CI 0.94–1.64). We further examined the association between ever-smoking and all mortality outcomes and found similar magnitude and direction of associations (data not shown).
We next examined whether tumor characteristics such as ER, PR and HER2, menopausal status, and BMI at cohort entry modified the effect of smoking on breast cancer mortality. When interaction terms were included in multivariate models, we found significant interactions with normal BMI (p-value= 0.003), and HER2 positivity (p-value=0.02), but not with ER, PR and menopausal status. Stratified analyses revealed increased risk of breast cancer death in current smokers with both ER positive and negative tumors (Table 3); the association did not reach statistical significance among those with ER negative tumors, likely due to the small number of breast cancer deaths in this group. We found no evidence of effect modification by PR or HER2 status.
Possible variations in the effect of smoking status on breast cancer survival according to BMI were also evaluated (Table 3). Fully adjusted models suggest that both current and former smoking increases the risk of breast cancer death across all BMI strata. While the association was statistically significant among current smokers (HR=4.46, 95%CI 2.03–9.76) and former smokers (HR=1.70, 95%CI 1.00–2.90) of normal weight (BMI<25), the association did not reach statistical significance among overweight and obese women (BMI ≥25) who were former or current smokers.
We next evaluated whether the association of smoking status with breast cancer mortality varied according to menopausal status. We found an increased risk of breast cancer death in pre and postmenopausal women who were current smokers (Table 3). Among former smokers, we found an increased risk of breast cancer mortality among pre-menopausal women (HR=2.73, 95%CI 1.40–5.32), but not post-menopausal women.
For the systematic review, we identified six published cohort studies of the association, and these are summarized in Table 4 together with our own study. In general, we observed increased a stronger risk of breast cancer mortality among women with breast cancer who were current smokers than among those who were former smokers. Notably, three of the seven studies in the systematic review reported risk ratios for estimates of breast cancer mortality and four reported hazard ratios for this outcome. Given these differences in the effect measures reported, we chose not to combine the studies quantitatively using meta-analysis.
Our prospective observational study shows that smoking following breast cancer diagnosis and treatment is detrimental not only to overall and non-breast cancer mortality but also to breast cancer specific mortality in women diagnosed with breast cancer. While former smoking affects non-breast cancer mortality, we found little evidence that former smoking increases the risk of breast cancer-specific mortality. We evaluated whether the association of smoking status with breast cancer mortality varied according to menopausal status and found that both premenopausal current and former smokers were at increased risk of breast cancer death, which is consistent with a recently published study. The association of current smoking with breast cancer-specific mortality was stronger among women of normal weight (BMI<25) than among overweight and obese women (BMI ≥25).
Of the seven prospective observational studies in our systematic review that examined current versus never smoking and breast cancer mortality, three large studies[9,42,13] and our own study indicated a positive association. Although none of the seven studies in our systematic review demonstrated a statistically significant association between former smoking and the risk of breast cancer death, the effect estimates were elevated in four reports and our own study (Table 4). We extend the literature on smoking and risk of death after breast cancer diagnosis to clarify that the association is stronger with current than with former smoking. The small number of primary studies in this systematic review precluded pooling of the results using meta-analysis; it also precluded an examination of subgroups and potential modifying effects of BMI, tumor subtype, and menopausal status in the association of smoking with mortality following breast cancer diagnosis.
The link between smoking and breast cancer risk or mortality has attracted considerable research attention. Ambrosone et al.  reported that only those postmenopausal women with the slow acetylation phenotype of the polymorphic Nacetyltransferase 2 (NAT2) gene showed an association between active smoking and incident breast cancer risk; fast acetylators showed no association. Among postmenopausal slow acetylators, Morabia et al.  reported that active smokers had a significantly higher breast cancer risk than never smokers whereas among premenopausal women, there were no significant differences in the relative risk of breast cancer between slow and rapid acetylators who were active versus never smokers.
Since breast cancer is a highly heterogeneous disease, it is plausible that smoking differentially affects estrogen-responsive tumors versus more aggressive forms of cancer. In addition, cigarette smoke has been associated with elevated metastatic potential of tumor cells and stimulation of angiogenesis . The association of cigarette smoke with increased pulmonary metastatic propensity in mouse models  gives rise to the hypothesis that smoking could increase the risk of breast cancer recurrence and death. Consistent with this, cross-sectional epidemiologic evidence has shown that smokers with breast cancer had more and larger lymph node metastases than nonsmokers, after controlling for primary tumor size and other variables [46–48]. In addition, smoking has been associated with a younger age at diagnosis , hormone receptor negative breast cancer  and an increased risk of lung metastases among breast cancer patients [45,51]. Although mechanisms that lead to the altered tumor behavior associated with smoking are not well established, it has been shown that some complications of radiation therapy may be more frequent and severe in smokers . Notably, never smokers were more likely to receive chemotherapy than current or former smokers. Analysis of smoking status among women undergoing breast cancer treatment may help better target therapeutic options.
A limitation of our study is that smoking status was self-reported and lacked details of the number of packs consumed and whether smoking continues throughout the follow-up period. Hence, we were unable to evaluate a dose-response relationship between smoking and outcomes, or evaluate the effect of change in smoking status later after breast cancer diagnosis. We were also unable to investigate age of starting smoking and how that factor may influence outcomes after breast cancer diagnosis. The proportion of women who were current smokers at the time of enrollment was low. In addition, smoking status and other measures were self-reported, and self-reported smoking exposures were not validated by smoking biomarkers. Since former smokers were likely motivated by a breast cancer diagnosis to stop smoking, relatively few current smokers were identified at the time of enrollment. It is important to note that the finding of this study regarding current smoking and breast cancer-specific survival was based on only 25 events in current smokers, thus in spite of statistical significance is liable to uncertainty. Our analyses consider cancer deaths, which may also reflect the role of smoking on cancer diagnosis or treatment rather than the direct effect of smoking on cancer mortality. Another potential limitation is the possible misclassification of cause of death. If some of the breast cancer deaths are in fact deaths from other causes, we would find at least some positive association with smoking and breast cancer mortality because deaths would include those from smoking-related conditions such as cardiovascular disease or lung cancer. Finally, we were also unable to evaluate the role of smoking in the development of any metastases.
The strengths of the study include its prospective population-based cohort design, its large size and our ability to take into account multiple covariates in the tumor-related, lifestyle and socio-demographic domains. In addition, information on molecular subtypes allowed us to compute subtype-specific estimates of association between smoking and breast cancer mortality. Efforts to investigate this question further should include additional investigations of smoking and subtype-specific mortality in studies with more detailed data on the intensity and duration of smoking exposure and a larger sample size for evaluating outcomes stratified by tumor subtype.
In conclusion, the results from this large cohort study and systematic review add to the scant evidence examining the association of smoking with mortality after breast cancer diagnosis. Taking the results of our cohort study and systematic review together, we have found evidence for a positive association of current smoking with breast cancer mortality. The evidence for an effect of former smoking on breast cancer mortality is weak. In addition, a stronger association was observed between current smoking and overall mortality than breast cancer-specific mortality. While this work is unable to provide a definitive answer on the nature of the relationship between smoking and breast cancer mortality, it provides more support for the importance of current than former smoking. In addition, the stronger association was observed between current or former smoking and overall mortality. As is the case in survivors of other cancer and in the general population, these results underscore the importance of promoting smoking cessation efforts among breast cancer survivors. Our data suggest that for breast cancer survivors who are current smokers, reduction in breast cancer mortality may be yet another motivation for smoking cessation.
This research was supported by Grant# KG110940 from Susan G. Komen for the Cure and by Grant# 121891-MRSG-12-007-01-CPHPS from the American Cancer Society. Additional support was obtained from the National Institutes of Health, National Cancer Institute, Bay Area Breast Cancer SPORE (P50 CA 58207). The data were derived from the Life after Cancer Epidemiology (LACE) study, which is supported by the National Cancer Institute (RO1 CA129059-05 to B. J. Caan).
|Section/topic||#||Checklist item||Reported on page #|
|Title||1||Identify the report as a systematic review, meta-analysis, or both.||1|
|Structured summary||2||Provide a structured summary including, as applicable: background; objectives; data sources; study eligibility criteria, participants, and interventions; study appraisal and synthesis methods; results; limitations; conclusions and implications of key findings; systematic review registration number.||1|
|Rationale||3||Describe the rationale for the review in the context of what is already known.||3|
|Objectives||4||Provide an explicit statement of questions being addressed with reference to participants, interventions, comparisons, outcomes, and study design (PICOS).||3|
|Protocol and registration||5||Indicate if a review protocol exists, if and where it can be accessed (e.g., Web address), and, if available, provide registration information including registration number.||NA|
|Eligibility criteria||6||Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, language, publication status) used as criteria for eligibility, giving rationale.||7|
|Information sources||7||Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched.||7|
|Search||8||Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated.||7|
|Study selection||9||State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis).||7|
|Data collection process||10||Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.||NA|
|Data items||11||List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made.||NA|
|Risk of bias in individual studies||12||Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis.||NA|
|Summary measures||13||State the principal summary measures (e.g., risk ratio, difference in means).||NA|
|Synthesis of results||14||Describe the methods of handling data and combining results of studies, if done, including measures of consistency (e.g., I2) for each meta-analysis.||NA|
|Risk of bias across studies||15||Specify any assessment of risk of bias that may affect the cumulative evidence (e.g., publication bias, selective reporting within studies).||NA|
|Additional analyses||16||Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, meta-regression), if done, indicating which were pre-specified.||NA|
|Study selection||17||Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram.||Figure 2|
|Study characteristics||18||For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations.||Table 4|
|Risk of bias within studies||19||Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12).||NA|
|Results of individual studies||20||For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.||Table 4|
|Synthesis of results||21||Present results of each meta-analysis done, including confidence intervals and measures of consistency.||NA|
|Risk of bias across studies||22||Present results of any assessment of risk of bias across studies (see Item 15).||NA|
|Additional analysis||23||Give results of additional analyses, if done (e.g., sensitivity or subgroup analyses, meta-regression [see Item 16]).||NA|
|Summary of evidence||24||Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers).||10|
|Limitations||25||Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias).||10|
|Conclusions||26||Provide a general interpretation of the results in the context of other evidence, and implications for future research.||12|
|Funding||27||Describe sources of funding for the systematic review and other support (e.g., supply of data); role of funders for the systematic review.|
From: Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group (2009). Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Med 6(6): e1000097. doi:10.1371/journal.pmed1000097