Cigarette Smoking and the Risk of Adult Myeloid Disease: A Meta-Analysis

Background The adult myeloid diseases, myelodysplastic syndrome and acute myeloid leukemia, have been reported to be associated with cigarette smoking, but the results have been conflicting. Previous studies may have ignored the relationship between myelodysplastic syndrome and acute myeloid leukemia, where approximately one-third of myelodysplastic syndrome cases will progress to acute myeloid leukemia, which could induce a serious bias in independent analyses. For the purposes of researching pathogenesis, we suggest that myelodysplastic syndrome and acute myeloid leukemia should be regarded as a single class of adult myeloid disease, and herein assessed the relationship between cigarette smoking and the risk of adult myeloid disease. Methods The PubMed, Cochrane Library, EBSCO, and EMBASE databases were systematically searched for reports published from 1990 to 2015. Two authors independently assessed the methodological quality and the extracted data. The odds ratios and adjusted odds ratios (OR), a sensitivity analysis, and the publication bias were analyzed using the CMA v2 (Comprehensive Meta Analysis Version 2) software program. Results Twenty-five studies were included in this meta-analysis. The publication dates ranged from 1990 to 2014. The pooled OR in current smokers and ever-smokers showed an increased risk of adult myeloid disease, with ORs of 1.45 (95% CI, 1.30–1.62; p<0.001) and 1.23 (95% CI 1.15–1.32; p<0.001) versus non-smokers, respectively. In the subset analyses, the OR of adult myeloid disease was increased regardless of the form of disease, geographical region, NOS (Newcastle Ottawa Scale) score, and source of controls. The smoking status was divided into <20 and ≥20 cigarettes per day, and these groups had ORs of developing adult myeloid disease of 1.24 (95% CI, 1.09–1.40; p = 0.001) and 1.32 (95% CI, 1.14–1.53; p<0.001), respectively. In the groups divided based on the number of years the subjects had smoked (<20 and ≥20 years), the ORs were 1.05 (95% CI, 0.90–1.23; p = 0.25) and 1.30 (95% CI, 1.16–1.45; p<0.001), respectively. Similarly, <20 and ≥20 pack-years were associated with ORs of 1.15 (95% CI, 1.03–1.29; p = 0.017) and 1.34 (95% CI, 1.18–1.52; p<0.001), respectively. Conclusions This meta-analysis, for the first time, combined myelodysplastic syndrome with acute myeloid leukemia to assess the overall risk of adult myeloid disease, and it demonstrated that cigarette smoking is associated with a significantly increased risk of adult myeloid disease.


Introduction
Adult myeloid diseases are a group of clonal diseases that affect the stem cells in bone marrow. Among the adult myeloid diseases, myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML) have high incidence rates of 4 and 3.7 per 100,000 people, respectively. Moreover, the number of MDS cases reaches 40-50 per 100,000 in patients older than 70 years old [1,2]. The median age at diagnosis is approximately 67 and 76 years for AML and MDS, and the conditions tend to occur earlier in Asian populations [2][3][4][5]. In addition, men have a significantly higher incidence rate than women [4,5]. A striking feature of MDS is its genetic instability, and approximately one-third of MDS cases result in AML [2,[6][7][8][9][10]. The pathogenesis of MDS or AML is uncertain, particularly why MDS can transform into AML.
Tobacco is an established factor that can cause carcinogenicity, teratogenicity and mutagenicity. Up to half of current smokers will eventually die of a tobacco-related disease [11]. The monograph by the International Agency for Research on Cancer (IARC), the National Comprehensive Cancer Network Clinical Practice Guidelines (NCCN Guidelines) and the European Society for Medical Oncology Practice Guidelines (ESMO Guidelines) reported that adult myeloid disease is linked to smoking [3,10,12]. Attributable risk calculations suggest that about 15% of all leukemia deaths and 24% of myeloid leukemia deaths are accounted for as a result of cigarette smoking [13].
Since the strongest evidence comes from cohort studies, some data were also available from case-control studies. Early in 1978, Paffenbarger [14] found that cigarette smoking was related to AML. Many subsequent case-control studies focused on investigating the relationship between cigarette smoking and the risk of AML or MDS, but the results had significant bias [15][16][17][18]. Fircanis [19] evaluated the relationship between smoking and AML, and Tong [20] investigated the relationship between smoking and MDS. However, neither of these researchers examined the interrelationship between MDS and AML, wherein approximately one-third of MDS cases transform into AML. Independent analyses of AML or MDS alone cannot accurately assess the association with smoking. In fact, both AML and MDS show identical cellular morphology, clonal hyperplasia of myeloblast. The number of myeloblasts is considered as one of important differential features between them. Myeloblasts are less than 20% in MDS, nevertheless more than 20% in AML [21]. We thought that MDS and AML should be grouped together as a single class of "adult myeloid disease" for the purposes of researching pathogenesis, and we herein assessed the relationship between cigarette smoking and the risk of developing adult myeloid disease.

Literature search and study selection
A systematic literature search was conducted by two independent reviewers (Wang and Liu) of the PubMed, Cochrane Library, EMBASE and EBSCO databases for papers published from 1990 to 2015. The following search terms were used: (1) "AML" OR "acute myeloid leukemia" or "myelodysplastic syndrome" OR "MDS" OR "myelodysplastic" OR "myelodysplasia" OR "preleukemia"; (2) "smoking" OR "tobacco" OR "cigarette". These search themes were combined using the Boolean operator "and" in several combinations without restrictions. In addition, the references reported in the identified studies were used to complete the search.
Studies eligible for inclusion in this meta-analysis met the following criteria: (1) the study design was a case-control study; (2) investigated the association between smoking and the risk of AML or MDS; (3) the case and control of studies must have been based on adults; (4) the diagnoses of AML and MDS were confirmed either histologically, pathologically, cytologically or by detailed medical records; (5) the odds ratios (OR) and the number of events could be calculated from the data presented;(6) published as a full text document in English. Two reviewers (Liu and Jiang) determined the study eligibility independently. Disagreements were resolved by consensus.

Data extraction and quality assessment
Two investigators (Liu and Jiang) read the full text of the manuscripts independently and extracted the following data from each eligible study: first author's name, publication year, country of origin, study period, source of controls, sample size, method of ascertainment of smoking, method of ascertainment of adult myeloid disease and the Newcastle Ottawa Scale (NOS) score [22]. The NOS score usually is used for assessing the quality of nonrandomized studies (case-control and cohort studies) in meta-analyses. To assess the quality of these studies, two reviewers (Liu and Jiang) independently read and scored each study according to the NOS [22]. NOS scores of 1-3, 4-6, and 7-9 were considered to indicate low, medium and high quality, respectively.

Statistical analysis
We calculated the pooled odds ratios (OR) with the 95% CI by using the Comprehensive Meta-Analysis software program (Version 2.2.064 July 27, 2011). The statistical heterogeneity within studies was evaluated using a χ 2 -based Cochran's Q statistic [23] and was further quantified using I 2 statistics (I 2 = 0-25%, no heterogeneity; I 2 = 25-50%, moderate heterogeneity; I 2 = 50-75%, large heterogeneity; and I 2 = 75-100%, extreme heterogeneity. For values of I 2 50%, the fixed-effects model was used; I 2 50%, the random-effects model was used) [23]. To investigate the potential heterogeneity, we also conducted a subgroup analysis based on the following four aspects: forms of disease, geographical region, source of controls and the NOS score. Publication bias was examined by Egger's regression and Begger's funnel plot (p 0.05 was considered to indicate the presence of significant publication bias) [24] and was adjusted by Duval and Tweedie's trim-and-fill method [25]. A sensitivity analysis was performed to evaluate the potential effects of the removal of one study in each turn. If the outcome was significantly changed after one study was removed, then the study was excluded from the included studies due to selection bias, and a new analysis was conducted.

Literature search
There were twenty-five articles included in our meta-analysis [15][16][17][18]. Two reviewers assessed seventy-six articles independently and excluded fifty-one articles. The reasons for exclusion of the articles were as follows: insufficient data to allow for estimation of the OR for eighteen studies, fourteen studies did not focus on our topic of interest, seven were based on cohort studies, ten were meta-analyses or reviews, and two were included in another article. Our search flow is shown in Fig 1.

Meta-analysis
Current smokers. Thirteen [15, 16, 26, 28, 30, 35, 37-39, 41, 43, 45, 46] out of fourteen studies independently researched the relationship between MDS or AML and cigarette smoking, and one [17] study meanwhile researched MDS and AML. The pooled OR was 1.45 (95% CI, 1.30-1.62; p<0.001), and it was determined that current smokers had a 45% higher risk of developing adult myeloid disease compared with non-smokers, with low heterogeneity (Fig  2A). A funnel plot identified the presence of publication bias, and Duval and Tweedie's trimand-fill analysis was used to adjust the ORs (Fig 2B), with p = 0.46 in the Begg rank correction test and p = 0.89 in Egger's linear regression test.
Ever-smokers. Twenty-one [15, 16, 18, 26-30, 32, 33, 35-39, 41-46] out of twenty-two studies independently researched the relationship between MDS or AML and cigarette smoking, and one study [17] meanwhile researched MDS and AML. The pooled OR was 1.23 (95% CI, 1.15-1.32; p<0.001), which confirmed that ever-smokers had a 23% increased risk of developing adult myeloid disease compared with non-smokers, with moderate heterogeneity ( Fig  2C). A funnel plot identified the presence of publication bias, so Duval and Tweedie's trimand-fill analysis was used to adjust the ORs (Fig 2D), with p = 0.75 in the Begg rank correction test and p = 0.83 in Egger's linear regression test.
A sensitivity analysis was conducted, and the pooled OR was not significantly changed in the current smokers (S1 Fig)

Subgroup analysis
We performed subgroup analyses stratified by the forms of disease, geographical region, NOS score (high and medium quality studies) and the source of controls in this study (Table 2).
Ever-smokers. The subjects were also grouped based on the form of disease, geographical region, NOS score and source of control, all results indicated an significant increased OR, but a little less than current smokers. The results were shown in Table 2

Smoking status analysis
Cigarette smoking does not exert its carcinogenic effects via a single event, but is instead a cumulative process. To further investigate the relationship between cigarette smoking and adult myeloid disease, three aspects were analyzed: the number of cigarettes smoked per day, the number of years of smoking and the pack-years smoked. The outcomes are shown in Table 3. Patients who smoked < 20 ( Fig 5A) and 20 (Fig 5B) cigarettes per day had 1.24 (95% CI, 1.09-1.40; p = 0.001) and 1.32 (95% CI, 1.14-1.53; p<0.001) increased risks of developing adult myeloid disease, respectively. For patients with <20 ( Fig 5C) and 20 (Fig 5D) years of smoking, the risk of adult myeloid disease was 1.05 (95% CI, 0.90-1.23; p = 0.25) and

Discussion
Recently, Fircanis [19] and Tong [20] analyzed association of smoking with AML and MDS by meta-analysis, respectively. Since approximately one-third of MDS cases could progress to AML, we supposed that the inclusion of some transformational cases (where the phenotype was AML but where the disease had developed from MDS) could have induced serious bias in independent analyses. Therefore, we conducted this meta-analysis that combined MDS with AML together as a single class of "adult myeloid disease" to eliminate bias. To our knowledge, this study is the first meta-analysis that combined MDS with AML to assess the overall risk of adult myeloid disease due to cigarette smoking. We extracted twentyfive studies that included 8,074 cases and 31,805 controls, and demonstrated a statistically significant association between current smoking and an increased incidence of adult myeloid disease. Current smokers were shown to have a 45% higher risk of developing adult myeloid disease than non-smokers. Ever-smokers had a 23% higher risk of developing adult myeloid disease than non-smokers, indicating that current smokers had a higher risk of adult myeloid disease than ever-smokers. One possible reason was that current smokers might have a higher total cumulative dose such as nicotine, benzene and polycyclic aromatic hydrocarbon, and longer exposure time of smoking than ever smokers. In addition, we found that the risk of myeloid disease had no significant drops after cessationin ever-smokers. As we known, tobacco is one of the most potent carcinogenic mixtures and the carcinogenic effect of cigarette smoking can persist for decades after smoking cessation. Few studies have reported on the dynamics and mechanism following cessation. Most recent two [47,48] studies showed that DNA methylation levels in ever-smokers approached those of never smokers several years after smoking cessation. Therefore, the DNA methylation maybe one factor that the risk of myeloid disease had no significant drops after cessationin ever-smokers. The truly irreversible nature of smoking-related and associated biomarkers are needed to be investigated.
According to the sub-analysis of the smoking status, there was a dose-relationship between cigarette smoking and adult myeloid disease. First, the risk of adult myeloid disease increased from 24% to 32% when smokers smoked <20 vs. 20 cigarettes per day. Second, the risk of adult myeloid disease increased from 5% to 30% in smokers who smoked <20 years compared with those who smoked 20 years. Finally, the risk of adult myeloid disease increased from 15% to 34% when the comparison was made of smokers who had smoked <20 vs. 20 packyears. Based on these results, we confirmed that there were close, positive, and cumulative effects of smoking on the development of adult myeloid disease. Moreover, we also found that cigarette smoking in North America was associated with a higher ORs than in Europe and  Asia, but limited studies were inadequate for a clear conclusion that people who live in North America were susceptible population. Now that cigarette smoking had a high risk to induce adult myeloid disease, Does it play a key role in MDS transformation? There have as yet been few detailed studies on the MDS transformation into AML by the individuals' smoking habits, however, indirect studies may provide us related clues. Over the years, a large number of studies have been presented that indicate a causal link between MDS development and genetic abnormality, related genes liked bcl-2 [49], C-CBL [50], CD95 [51], HOXA9 [52], TNFα [53]. In addition, CpG methylation [54, 55], telomere shortening [56,57] and microsatellite instability [58][59][60] also were found association with MDS progression to AML. Most importantly all above mentioned factors were affected by cigarette smoking [61][62][63][64][65][66][67][68][69][70]. Based on these studies, it indirectly indicated that cigarette smoking may be an induced factor on MDS development.
Benzene of smoking is a known clastogen, causing chromosomal aberrations in vitro, and prolonged exposure to mean concentrations of >64 mg/m 3 may be associated with chromosomal aberrations. In addition, increased risks of developing all types of leukemias were found at all concentrations of exposure, while exposure to levels above 40 ppm increased the risk of acute nonlymphocytic leukemia [71]. Zhang [72] conducted a chromosome-wide aneuploidy study (CWAS) in workers exposed to benzene, and their findings suggested that the development of aneuploidy may be a potential mechanism underlying benzene-induced leukemia. Benzene is mainly found in the atmosphere, and almost all human exposure is through inhalation (95% of daily intake) [71]. Personal exposure assessment research has indicated that an average cigarette smoker inhales 6-10 times the benzene inhaled by average non-smokers, and approximately 90% of a smoker's benzene exposure is from smoking [73]. Consequently, the tobacco use may lead to chromosomal and karyotype aberrations due to increased benzene exposure, thus resulting in MDS or AML.
Our study is associated with some limitations that may have affected the results. First, the results were calculated from published studies, and our included studies cannot cover published studies that could have led to a publication bias. In addition, published studies are often positive, and the omission of unpublished studies (with negative findings) may lead to an overestimation of the pooled OR. Second, the methods used to evaluate the smoking habits were mainly questionnaires or reports, rather than based on determinations of the blood levels of cotinine. Third, some of the subset analyses, although specified a priori, were performed on small datasets. Finally, methodological differences and confounding factors in the included studies were unavoidable.

Conclusion
In summary, we herein demonstrated that cigarette smoking leads to a significantly increased risk of adult myeloid disease in adults. Cigarette smoking does not act as a carcinogenic factor via a single exposure, and instead leads to a cumulative process. We found that there was a positive risk between smoking and adult myeloid disease when a threshold smoking dose was crossed.
More and larger prospective studies should be performed, and future studies should focus on investigating the pathway(s) by which cigarette smoking can induce adult myeloid disease.