Medical Conditions and Modifiable Risk Factors for Myelodysplastic Syndrome: A Systematic Review

Myelodysplastic syndromes (MDS) are a group of myeloid malignancies that result in dysplastic and ineffective hematopoiesis, and have a poor 5-year relative survival estimate (<50%; ref. 1). Surveillance data suggest that the age-adjusted MDS incidence rate is 4.6 per 100,000 in the United States (2); however, with a median age at diagnosis between 71 and 76 years (3, 4), the number of new cases is expected to increase as the average age of the U.S. population increases. The only curative treatment option for MDS is stem cell transplantation following high-dose chemotherapy and total body irradiation, which, given the toxicity of the treatment regimen, is typically reserved for younger, healthier patients. MDS in older patients or those without a suitable bone marrow donor is typically limited to supportive care with transfusions, hematopoietic growth factors, or chemotherapy (5). Given the poor survival rate, identification of modifiable risk factors for MDS could have a significant impact on reducing incidence and ultimately deaths due to this disease, yet there is a notable lack of studies on modifiable risk factors for MDS.

The classification of MDS has undergone several changes since the initial classification system developed in the 1970s by the French–American–British (FAB) Cooperative Group (6). The behavior code for MDS under the FAB classification system was 1 (uncertain whether benign or malignant), which resulted in MDS being considered a premalignant condition, and the lack of a requirement to report MDS to cancer registries. In 2001, the World Health Organization (WHO), building on evolving biological and clinical evidence related to myelodysplastic syndromes, changed the International Classification of Diseases for Oncology (ICD-O) behavior code for MDS to 3 (malignant), and the National Cancer Institute (NCI) Surveillance Epidemiology and End Results (SEER) program began collecting data on MDS incidence. The 2001 WHO update also resulted in the refractory anemia with excess blasts in transformation (RAEB-t) subtype being reclassified as MDS rather than a form of acute myeloid leukemia as it had been under the FAB system (7).

Further revisions were made to the WHO criteria in 2008 (8) and 2016 (9) to reflect continued advances in understanding of MDS biology and clinical biomarkers with prognostic relevance. The primary change in the most recent 2016 revision is that the MDS subtypes are no longer classified primarily as being either “refractory anemia” or “refractory cytopenia.” Instead, all of the subtypes are defined as MDS, with specifications made based on single versus multiple hematopoietic cell lineage, and the presence of blasts, ring sideroblasts, or the del(5q) chromosomal abnormality (9).

Most of the studies evaluating risk factors for MDS that were conducted prior to 2001 tended to evaluate risk factors for myeloid malignancies overall rather than for MDS specifically and were based on case definitions using the older FAB MDS diagnostic criteria. There have now been more than 15 years of MDS case accrual using the more uniform WHO definitions in the existing large prospective, observational cohort studies, as well as several MDS case–control studies that have used the WHO diagnostic criteria to evaluate risk factors for MDS.

Clinical studies have shown that chemotherapy (particularly alkylating agents) or radiotherapy from a previous cancer can lead to subsequent MDS (10–13). However, therapy-related cases are hypothesized to account for only 5% of MDS diagnoses (14, 15), and the majority of MDS cases are thought to occur de novo. Occupational exposure to benzene (16, 17), pesticides (18–20), and radiation (21) have also been reported to be associated with higher risk of MDS.

Lifestyle factors may also alter risk of MDS in the general population. Previous meta-analyses have reported higher risk of MDS among heavy smokers (22), and no association between alcohol consumption and MDS (22, 23). However, none of the previous meta-analyses restricted the inclusion criteria to a specific set of MDS diagnostic classification criteria (FAB vs. WHO), which may have introduced outcome misclassification. Therefore, the aim of this systematic review was to evaluate the association between medical conditions and modifiable factors and risk of MDS specifically among studies that used either the 2001 or 2008 WHO MDS diagnostic criteria to classify cases.

A literature search was conducted using PubMed, MEDLINE, and Scopus databases for articles published between January 1, 2001, and August 1, 2017. For PubMed, the search terms “myelodysplastic syndromes” (limited to title/abstract) and “epidemiology” were used. MEDLINE search terms included “myelodysplastic syndromes (/ep)” or “risk factors.” Scopus search terms included “epidemiology” and “myelodysplastic syndrome*” and “risk factor.”

Titles, abstracts, and then full manuscripts were reviewed for relevance. Inclusion criteria included articles that were published in English in a peer-reviewed journal. In addition, articles had to present findings specifically for risk of MDS in a population of adults (≥18 years of age), and the study had to define MDS using the 2001 or 2008 WHO classification criteria (8, 24). Articles were excluded from the systematic review if: cases were classified using the FAB criteria, occupational exposures were evaluated, or if treatment-related MDS incidence or MDS mortality was the outcome of interest. Systematic reviews and meta-analyses were also excluded, although reference lists were reviewed for additional eligible articles.

All six authors of this systematic review participated in the review and quality rating assignments for the included studies. Each of the included articles was assigned to two reviewers based on the reviewers' area of expertise, with each author reviewing 4 articles. None of the authors of this article reviewed articles in which they were listed as an author or coauthor.

The following information was abstracted into an electronic data collection form (REDCap 7.4.11, Vanderbilt University, Nashville, TN): number of study participants, number of MDS cases, study design, method of case ascertainment (hospital/clinic-based, cancer registry, community/population-based, other), MDS classification system, potential risk factors evaluated, and study findings. For case–control articles, information on the method of identifying controls and matching criteria (if any) was abstracted.

A modified version of the Downs and Black quality rating scale (25) and the Academy of Nutrition and Dietetics Quality Criteria Checklist (26) was used to assign quality ratings (Supplementary Table S1). The quality checklist items assess articles for various dimensions of research quality, including completeness of reporting, and the potential for external validity, bias and confounding. For each quality checklist item, reviewers were asked to determine whether the article satisfied the checklist item (yes = 1 point), or whether the article did not satisfy the checklist item, or the reviewer was unable to determine whether the item was satisfied on the basis of the research reporting (no/unable to determine = 0 points). A third reviewer (M.R. Sweeney, H. Arem, or K. Robien) independently reviewed quality checklist items for which the two initial reviewers did not agree. Each article could receive a maximum of 14 points if all quality checklist items were satisfied. Articles with 0 to 4 points were considered to be of negative quality, 5 to 9 points considered to be neutral quality, and 10 to 14 points considered to be of positive quality.

This systematic review was registered in the International Prospective Register of Systematic Reviews (PROSPERO) database (CRD42017076443). The full project protocol is available at https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=76443.

The PRISMA diagram (Fig. 1) outlines the findings of the literature search. The initial PubMed search identified 896 manuscripts. Nine additional papers were then identified through the MEDLINE search, and 17 additional papers were identified through the Scopus search. An additional eight papers were identified via the reference sections of meta-analysis articles. After applying the inclusion/exclusion criteria, a total of 13 articles were included in this review.

Table 1 summarizes the articles and their findings by exposures. Nine of the articles were case–control studies (17, 27–34). The remaining four articles were cohort studies (35–38). The studies were conducted in the United States (n = 6), China (n = 4), Japan (n = 1), Israel (n = 1), and Sweden (n = 1). Three of the articles used the same case–control study population (28–30). One article included an all-female population (36); the other 12 articles included both men and women. Three of the 13 articles evaluated more than one potential MDS risk factor (17, 35, 38).

Agreement in quality scores between the two initial reviewers ranged from 43 to 100% (average: 85.0%, SD: 17.0), and all discrepancies in initial scores were successfully adjudicated by one of the three reviewers. All 13 papers were assigned positive quality scores, with scores ranging from 10 to 14 points. Seven of the 13 articles (27–29, 32, 33, 35, 36) received scores of 14 (maximum score). Two articles received scores of 10 (17, 31). The other four articles received scores of 12 (37) or 13 (30, 34, 38). There were several reasons that papers did not achieve full points. Reasons for lower quality ratings included: (i) did not clearly state how MDS was identified and classified in their study population (17, 31), (ii) did not include participants representative of the general population from which they were selected (17, 31), (iii) did not recruit participants or cases/controls over the same period of time (31), (iv) presented analyses that were not planned a priori (34, 37), (v) time period between exposure and outcome was not the same for cases and controls (31), (vi) did not adequately adjust for confounding (17, 37), or (vii) did not account for the strengths or limitations of the study design when drawing conclusions (17, 38).

Two articles examined the risk of MDS in relation to obesity, as measured by body mass index (BMI). Having a BMI ≥ 30 kg/m² was associated with a more than doubled risk of MDS among both men and women in the United States in the NIH-AARP Diet and Health cohort study (ref. 35; 64 MDS cases, HR: 2.18; 95% CI: 1.51–3.17), compared with individuals with a normal BMI (18.5–25 kg/m²; 51 MDS cases). In a case–control study of 256 MDS cases and 1,388 controls based in Minnesota, Poynter and colleagues (27) found higher odds of MDS among women who were overweight and obese at age 35 in Minnesota, respectively (overweight, n = 23 MDS cases, OR: 2.02; 95% CI: 1.10–3.70; obese, n = 10 MDS cases, OR: 3.18; 95% CI: 1.34–7.51), compared with normal BMI women, but no statistically significant differences among men (overweight, n = 80 MDS cases, OR: 1.11; 95% CI: 0.74–1.61; obese, n = 28 MDS cases, OR: 1.53; 95% CI: 0.90–2.61). The investigators also found that women who were underweight at age 18 had 74% lower odds of developing MDS (OR: 0.26; 95% CI: 0.10–0.70) compared with women with a BMI in the normal range at age 18.

Physical activity was associated with a lower risk of MDS in the NIH-AARP prospective cohort study (35). Those who exercised vigorously ≥3 times/week had 32% lower risk of MDS (n = 82 cases, HR: 0.68; 95% CI: 0.49–0.95) compared with those who exercised vigorously ≤3 times/month (68 cases).

Smoking was evaluated as a risk factor for MDS in two large prospective cohort studies (35, 38) and two case–control studies (17, 31). All studies reported a higher risk of MDS with smoking. Among the cohort studies, Ma and colleagues (35) observed a higher risk of MDS among ever (n = 145 cases, HR: 1.91; 95% CI: 1.35–2.71), former (n = 109 cases, HR: 1.68; 95% CI: 1.17–2.41), and current smokers (n = 36 cases, HR: 3.17; 95% CI: 2.02–4.98) compared with nonsmokers in the NIH-AARP study. For current smokers, the risk was greater for those who smoked more than one pack per day (OR: 4.70; 95% CI: 2.68–8.24) than for those who smoked one pack per day or less (OR: 2.42; 95% CI: 1.39–4.21). The risks were similar for former smokers regardless of the number of packs smoked per day (≤1ppd, OR: 1.55; 95% CI: 1.04–2.32; >1 ppd, OR: 1.87; 95% CI: 1.23–2.83). Ugai and colleagues (38) reported a marginally statistically significant 2-fold higher risk of MDS among current smokers (HR: 2.11; 95% CI: 0.91–4.89) in a prospective cohort study in Japan with 70 MDS cases. Similar findings were observed for men with >30 pack-years of smoking (HR: 2.22; 95% CI: 0.95–5.19). No female smokers were diagnosed with MDS during the study follow-up period.

Smoking was also found to be associated higher risk of MDS in the case–control studies. Björk and colleagues (31) report slightly higher odds of MDS per pack-year of smoking among current smokers (OR: 1.013; 95% CI: 1.001–1.03), and a marginally statistically significantly higher odds among recent smokers (OR: 1.8; 95% CI: 1.0–3.3) compared with never smokers in a small case–control study with 75 MDS cases and 278 population- and hospital-based controls. Lv and colleagues (17) observed 44% higher odds of MDS overall (n = 141 cases, OR: 1.44; 95% CI: 1.04–2.01) and a more than doubled risk of the refractory anemia with excess blasts subtype (RAEB; n = 37 cases, OR: 2.32; 95% CI: 1.12–4.77) when comparing ever smokers to never smokers, but no association was observed with the refractory cytopenia with multilineage dysplasia subtype (RCMD, OR: 1.10; 95% CI: 0.73–1.65). Duration (1–19 years, OR: 1.32; 95% CI:0.67–2.61; ≥20 years, OR: 1.22; 95% CI: 0.91–1.63), number of cigarettes/day (1–19 cigarettes/day, OR: 1.21; 95% CI: 0.82–1.78; ≥20 cigarettes/day, OR: 1.06; 95% CI: 0.75–1.50), and pack-years (1–19 pack-years, OR: 1.11; 95% CI: 0.76–1.61; ≥20 pack-years, OR: 1.06; 95% CI: 0.73–1.55) were not associated with MDS risk. However, smoking ≥20 pack-years was associated with more than doubled odds of the RAEB subtype (n = 27 cases, OR: 2.66; 95% CI: 1.06–6.69).

Results of two prospective cohort studies and two case–control studies are inconsistent. However, when restricted to the two studies that evaluated the association between alcohol use and MDS risk with never drinkers as the comparison group (17, 35), the findings suggest no association with risk of MDS.

In the NIH-AARP cohort study, Ma and colleagues (35) reported no association between alcohol intake and MDS risk for the highest third of consumption (n = 57 MDS cases, HR: 0.81; 95% CI: 0.55–1.19). However, the lowest third of alcohol consumption (number of drinks per day not specified) was associated with a 40% lower risk of MDS (n = 35 cases, HR: 0.58; 95% CI: 0.38–0.88) compared to never drinkers (58 cases). In a prospective cohort study based in Japan, Ugai and colleagues (38) found that regular alcohol consumption (1–299 g ethanol/week, n = 14 cases, HR: 0.41; 95% CI: 0.21–0.78) was associated with a 59% lower risk of MDS compared with current nondrinkers (defined as <1 day of alcohol consumption per month or former drinkers). Higher alcohol consumption was associated with lower risk of MDS among men (≥300 g/week, HR: 0.45; 95% CI 0.20–0.99), although there were only 9 cases in this group. Alcohol consumption was not associated with MDS risk in women in this study; however, only one female MDS case reported any alcohol use. All types of alcohol (sake, distilled spirits, beer) suggested protective effects, but only the findings for beer intake reached statistical significance (1–299 g/week, n = 5 cases, vs. nondrinkers n = 24 cases, HR: 0.36; 95% CI: 0.13–0.98).

In a case–control study in China (208 MDS cases, 208 controls), Liu and colleagues (30) also found that participants who reported drinking alcohol had a 60% lower odds of developing MDS compared with “never or hardly ever” drinkers (n = 76 cases, OR: 0.41, 95 % CI: 0.21–0.80). The lowest odds of MDS was observed for light drinkers (≤12.5g/day, n = 19 cases, OR: 0.27; 95% CI: 0.12–0.61) and wine drinkers (n = 4 cases, OR: 0.12; 95% CI: 0.02–0.79). The authors also conducted a joint association analysis for smoking and alcohol intake, and found that the lower odds of MDS associated with overall and light alcohol intake remained significant only for nonsmokers (overall, n = 21 cases, OR: 0.35; 95% CI: 0.15–0.81; light, n = 8 cases, OR: 0.19; 95% CI: 0.06–0.60). Conversely, another case–control study in China with 403 MDS cases (17) did not find an association between alcohol consumption and risk of MDS (ever consumer, OR: 1.26; 95% CI: 0.89–1.78; current consumer, OR: 1.40; 95% CI: 0.82–2.38; former consumer, OR: 1.26; 95% CI: 0.80–1.99).

Tea drinking was evaluated as a potential risk factor for MDS in one case–control study of 208 cases of MDS (28). Any tea drinking was associated with 60% lower odds of MDS compared with non-tea drinking (n = 89 cases, OR: 0.40; 95% CI: 0.24–0.67). Similarly, any duration of tea drinking was inversely associated with odds of MDS (≤20 years, n = 40 cases, OR: 0.40; 95% CI 0.23–0.73; >20 years, n = 49 cases, OR: 0.39; 95% CI: 0.20–0.74) when compared with non-tea drinkers (119 cases), as was the number of cups (≤1 cup/day, n = 25 cases, OR: 0.31; 95% CI: 0.16–0.63; ≥2 cups/day, n = 64 cases, OR: 0.45; 95% CI: 0.25–0.79) and the grams of dried tea leaves consumed per year (<750 g/year, n = 37 cases, OR: 0.39; 95% CI: 0.21–0.72; ≥750 g/year, n = 52 cases, OR: 0.40; 95% CI: 0.21–0.77). When stratified by sex, only intake of ≥750 g dried tea/year was associated with a statistically significant 60% lower odds of MDS among men (n = 45 cases, OR: 0.42; 95% CI: 0.19–0.91), whereas among women, only intake of <750 g dried tea/year was associated with a statistically significant 65% lower odds of MDS (n = 15 cases, OR: 0.34; 95% CI: 0.15–0.80). In this study, the authors also stratified the analyses by MDS subtype (WHO 2008 classification), and found that tea drinking was associated with nearly 70% lower odds of the RAEB subtype (n = 37 cases, OR: 0.32; 95% CI: 0.13–0.79) compared with no tea consumption. Similarly, duration of tea consumption of ≤20 years (n = 10 cases, OR: 0.21; 95% CI: 0.06–0.68), consuming ≥2 cups of tea/day (n = 27 cases, OR: 0.32; 95% CI: 0.12–0.85), and use of ≥750 g dried tea leaves/year (n = 23 cases, OR: 0.27; 95% CI: 0.09–0.80) were associated with lower odds of the RAEB subtype compared with non-tea drinkers. No significant associations with duration or amount of tea consumed were observed for the RCMD subtype.

Two studies assessed dietary risk factors for MDS including fruit, vegetable, and meat intake. In the NIH-AARP Diet and Health study (193 MDS cases), no statistically significant associations between daily fruit and vegetable intake and risk of MDS were observed (servings per day, second tertile, HR: 1.06; 95% CI: 0.75–1.50; third tertile, HR: 1.02; 95% CI: 0.71–1.47; ref. 35). However, when Liu and colleagues (29) evaluated associations between dietary isoflavones (found in plant foods, especially soybeans) in a case–control study conducted in China (208 MDS cases, 208 controls), they found that adults who consumed the highest amounts of isoflavones, including daidzein, had 60% lower odds of developing MDS (isoflavones overall, n = 50 cases, OR: 0.40; 95% CI: 0.20–0.81; daidzein specifically, n = 51 cases, OR: 0.43; 95% CI: 0.21–0.85). Individuals reporting higher daily intake of genistein (6.0–13.4 mg/day, n = 57 cases, OR: 0.49; 95% CI: 0.26–0.94; 13.5–57.3 mg/day, n = 54 cases, OR: 0.36; 95% CI: 0.18–0.74) or glycitein (1.1–2.1 mg/day, n = 53 cases, OR: 0.38; 95% CI: 0.19–0.74; 2.2–10.3 mg/day, n = 59 cases, OR: 0.49; 95% CI: 0.25–0.97) had significantly lower odds of an MDS diagnosis compared with those with lower genistein or glycitein intake. No association was observed between meat intake and risk of MDS in the NIH-AARP Diet and Health study (grams/day, second tertile, HR: 1.08; 95% CI: 0.76–1.53; third tertile, HR: 1.20; 95% CI: 0.85–1.70; ref. 35).

In the first study to evaluate the association between allergies and risk of MDS specifically, Linabery and colleagues (36) found no significant associations for allergy overall (HR: 1.20; 95% CI: 0.70–2.03), or asthma (HR: 2.00; 95% CI: 0.93–4.32), hay fever (HR: 1.06; 95% CI: 0.42–2.70), skin allergy (HR: 0.73; 95% CI: 0.29–1.85), or other individual allergy (HR: 1.22; 95% CI: 0.64–2.34), specifically, in an analysis of data from the Iowa Women's Health Study prospective cohort (59 MDS cases).

Anderson and colleagues (32) observed significant positive associations for a history of any autoimmune condition (n = 564 cases, OR: 1.50; 95% CI: 1.35–1.66) and odds of MDS, as well as for several individual autoimmune conditions among older adults in a large population-based case–control study using SEER-Medicare data (2471 MDS cases). However, only rheumatoid arthritis (OR: 1.52; 95% CI: 1.27–1.81) and pernicious anemia (OR: 2.38; 95% CI: 1.98–2.86) remained statistically significantly associated with higher odds of MDS after Bonferroni corrections were applied. Similarly, only rheumatoid arthritis (OR: 1.52; 95% CI 1.19–1.92), polymyalgia rheumatica (OR: 1.53; 95% CI: 1.05–2.22), and pernicious anemia (OR: 1.68; 95% CI: 1.24–2.26) remained statistically significantly associated with higher MDS odds in a 5-year lag analysis.

Titmarsh and colleagues (33) evaluated the association between community acquired infections and odds of MDS in a population-based case–control study of Medicare participants (3,072 MDS cases). The investigators found that a diagnosis of bronchitis (n = 1,051 cases, OR: 1.25; 95% CI: 1.16–1.36), influenza (n = 427 cases, OR: 1.29; 95% CI: 1.16–1.44), pharyngitis (n = 501 cases, OR: 1.22; 95% CI: 1.11–1.35), pneumonia (n = 815 cases, OR: 1.52; 95% CI: 1.40–1.66), sinusitis (n = 828 cases, OR: 1.25; 95% CI: 1.15–1.36), cellulitis (n = 943 cases, OR: 1.51; 95% CI: 1.39–1.64), herpes zoster (n = 233 cases, OR: 1.31; 95% CI: 1.14–1.50), cystitis (females only; n = 1270 cases, OR: 1.26; 95% CI: 1.17–1.36), or gastroenteritis (n = 148 cases, OR: 1.38; 95% CI: 1.17–1.64) in the previous year were all associated with higher odds of MDS (Bonferroni corrected). They also conducted a latency analysis, and found that for bronchitis (P = 0.002), influenza (P = 0.004), pharyngitis (P = 0.006), and cellulitis (P = 0.016), the odds of MDS decreased with greater time since the infection occurred.

Two studies, one cohort and one case–control study, evaluated the association between anemia and the risk of MDS. In a case–control study using the SEER-Medicare database (6,067 MDS cases, 100,000 controls), Murphy and colleagues (34) found that elderly individuals with pernicious anemia (vitamin B₁₂ deficiency anemia) in the previous year were at higher odds of developing MDS (OR: 2.87; 95% CI: 2.53–3.26). Similarly, a cohort study in Israel found that among 116,517 adults with anemia, those with the lowest levels of hemoglobin (<9 g/dL) had the highest risk of developing MDS (n = 418 MDS cases, HR: 5.09; 95% CI: 2.99–8.67) compared with individuals with hemoglobin levels between 11 and 12 g/dL (37). A previous history of diabetes mellitus (type of diabetes not specified) was associated with a higher risk of MDS (n = 55 cases, HR: 1.57; 95% CI: 1.16–2.11), whereas a previous history of cardiovascular disease was associated with a lower risk of MDS (n = 17 cases, HR: 0.39; 95% CI: 0.26–0.57).

Lv and colleagues (17) found no association between use of hair dyes (at least 2 times/year, OR: 1.31; 95% CI: 0.88–1.93) and odds of MDS after adjusting for covariates. Similarly, use of traditional Chinese medicines was not associated with odds of MDS overall (OR: 1.62; 95% CI: 0.82–3.22), or the RAEB (OR: 0.61; 95% CI: 0.21–1.78) or RCMD (OR: 1.77; 95% CI: 0.76–4.10) subtypes (WHO 2001 classification). However, use of antituberculosis drugs was associated with higher odds of MDS (n = 19 cases, OR: 3.15; 95% CI: 1.22–8.12; ref. 17).

While only a small number of studies of potential risk factors for MDS using the 2001 or 2008 WHO diagnosis criteria have been conducted to date, these studies have reported that higher BMI, smoking, a history of autoimmune disorders, community acquired infections, or anemia, and use of antituberculosis drugs were associated with higher risk of MDS. Vigorous physical activity, and tea and dietary isoflavone intake were associated with lower MDS risk. Overall, findings suggest that there is no association between alcohol use and risk of MDS; however, findings related to type of alcohol, and amount or duration of alcohol use are inconsistent and warrant further investigation. Similarly, use of traditional Chinese medicines, hair dye use, meat intake, and fruit and vegetable intake were not reported to be associated with risk of MDS.

Obesity, smoking, and alcohol consumption have been linked to increased risk of a number of different cancers (39). Although specific mechanisms by which obesity may contribute to MDS risk are unclear, alterations in endogenous hormone concentrations, insulin sensitivity, the insulin-like growth factor system, lipid signaling, adipokine and inflammatory cytokine responses, and the mTOR and MAPK/ERK signal transduction pathways may be contributing factors (40–43). Smoking exposes users to a number of potentially carcinogenic compounds, including benzene (38, 44).

Ethanol and its acetaldehyde metabolite have been shown to be genotoxic, either directly or through the creation of reactive oxygen species, and to alter DNA methylation (45). The World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) Continuous Update Project Expert Report 2018 identifies convincing evidence that alcohol intake is associated with higher risk of cancers of the mouth, pharynx, larynx, esophagus (squamous cell carcinoma), liver, colorectum, and postmenopausal female breast (39). Hematologic malignancies have not been included in the WCRF/AICR reviews to date; however, a recent meta-analysis by Psaltopoulou and colleagues (46) that used the WCRF methodology to evaluate the association between alcohol consumption and risk of hematologic malignancies found that alcohol intake appears to confer a protective effect on risk of non-Hodgkin lymphoma, but was not associated with risk of Hodgkin lymphoma or leukemias. The meta-analysis did not evaluate the association between alcohol intake and risk of MDS. Two earlier meta-analyses reported no association between alcohol intake and risk of MDS (using older FAB diagnostic criteria; refs. 22, 23).

The studies included in this systematic review reported conflicting results; however, it should be noted that the studies that used “never drinkers” as the comparison group reported no overall association between alcohol use and MDS risk, whereas studies that included former drinkers or rarely drinkers in the comparison group (30, 38) reported lower risk of MDS among light drinkers. These findings are consistent with a recent methodologic analysis showing that including former drinkers and individuals who no longer drink due to health reasons in the “no alcohol” (abstainer) comparison group creates an abstainer bias that strengthens observed associations among light to moderate drinkers in studies of the association between alcohol and mortality (47).

Fruits, vegetables, and other plant-based foods are complex mixtures of different micronutrients and phytochemicals, many of which may be protective against development of MDS through a variety of biologic mechanisms, including acting as antioxidants, providing key micronutrients needed for normal hematopoiesis, or through modulation of intracellular signaling or inflammation pathways. While Ma and colleagues (35) found no association between overall fruit and vegetable intake and MDS risk, Liu and colleagues found higher intake of tea (28) and dietary soy isoflavones (29) to be associated with lower odds of MDS. Further studies are needed determine which aspects of plant-based foods might be most relevant to prevention of MDS.

Like diet, physical activity may lower risk of MDS through a variety of different mechanisms, including achieving/maintaining a healthy weight, regulation of hormone levels (including insulin), and beneficial effects on the immune system (48). To date, only Ma and colleagues (35) have evaluated the association between physical activity and risk of MDS using the more homogenous WHO MDS diagnostic criteria. Further studies are needed to confirm these findings, and to evaluate whether any observed effects are modified by body composition.

The findings of an association between history of autoimmune disorders and community acquired infections and risk of MDS are consistent with the hypothesis that disordered immune responses increase risk of MDS as a result of exposure to the medications used to treat these conditions, and/or though abnormal immune responses directly damaging the bone marrow (32, 49). In addition, chronic antigenic stimulation that can occur with autoimmune conditions and community-acquired infections has been suggested as a mechanism that might increase risk of myeloid malignancies and MDS (50, 51). Given the limited number of studies in this area and the potentially conflicting biologic mechanisms by which allergies may influence risk of MDS, additional studies are needed to confirm these findings.

Ineffective hematopoiesis and low blood counts are a component of the diagnostic criteria for MDS, thus reverse causality is a concern when evaluating the association between anemia and risk of MDS. The two studies included in this systematic review that evaluated the association between history of anemia and MDS risk both used administrative data, and may lack access to key information necessary for full interpretation of the findings. In the study by Meytes and colleagues (37), all participants had low hemoglobin levels (≤ 12 g/dL), and there was no control group without anemia. The underlying cause of anemia was not evaluated. The authors also did not conduct sensitivity analyses excluding participants with more recent MDS diagnoses (median time between hemoglobin assessment and MDS diagnosis was 3.5 years), thus reverse causality cannot be ruled out as an explanation for their findings. Murphy and colleagues (34) evaluated the association between a diagnosis of pernicious anemia (vitamin B₁₂ deficiency anemia) and risk of MDS. While deficiencies in micronutrients such as folate, iron and the B vitamins necessary for normal DNA synthesis and hematopoiesis could be risk factors for cancer initiation, further studies using different study designs with more detailed data on dietary intake, clinical factors, and timing of nutrient deficiencies relative to the MDS diagnosis are needed to confirm this finding.

Lv and colleagues (17) are the first to report an association between the use of antituberculosis drugs and odds of MDS. Previous studies have shown that exposure to the antituberculosis medications (isoniazid, streptomycin, rifampin, and pyrazinamide) can cause DNA damage, including chromosome aberrations, sister chromatid exchanges, and aneuploidy in mice (52) and lymphocytes from patients with tuberculosis (53). Clinically, these medications are known to alter blood counts, and it is recommended that patients be supplemented with vitamin B₆ (pyridoxine) primarily to prevent peripheral neuropathy, but also to prevent microcytic anemia. Data on dietary or supplemental vitamin B₆ intake was not collected by Lv and colleagues (17), but this should be evaluated in future studies.

A majority of the studies included in this review (10 of 13; 77%) were case–control studies, which are limited in their ability to establish temporality of the exposure relative to the outcome and increase the likelihood of nondifferential recall bias. Differential recall bias is also a concern for case–control studies, where a cancer diagnosis may heighten awareness or recollection of factors that may have contributed to development of disease. However, the case–control studies generally included larger numbers of MDS cases than the included prospective cohort studies (case–control: 75 to 6,067 cases; cohort: 59 to 418 cases). Two of the case–control studies included more than 2,000 MDS cases, and only one case–control study included fewer than 100 MDS cases.

All thirteen of the articles included in this systematic review received positive quality ratings, likely due to the fact that all have been published since 2009, and most journals now require authors to include more methodological information for greater transparency. When the included articles did not receive the full allotment of points for the quality rating, it was most commonly because the authors did not provide sufficient information for the reviewers to determine whether the MDS cases were properly identified/classified, whether the study participants were representative of the general population, or whether the participants (cohort studies) or cases/controls (case–control studies) were recruited over the same period of time.

With the exception of the studies that evaluated autoimmune disorders (32) and community-acquired infections (33) as potential MDS risk factors, the studies included in this systematic review all used self-reported data on participant exposures, and thus may be subject to recall bias.

Strengths of this review include the use of a quality rating system and the focus on studies that used the WHO classification system for MDS (2001 or 2008 version). Multiple reviewers independently assessed and rated each study. The exclusion of studies with cases ascertained prior to 2001 ensures that the definition of MDS is consistent across all included studies. This review was limited in that we were only able to include English language publications. In addition, the small body of research on any given exposure of interest precluded the ability to conduct meta-analyses.

The findings of this systematic review indicate that higher BMI, smoking, a history of autoimmune disorders, community acquired infections, history of anemia, and use of antituberculosis drugs are associated with higher risk of MDS. Vigorous physical activity, and tea and dietary isoflavone intake were associated with lower MDS risk. Findings suggest no association between the other factors and risk of MDS. However, research on risk factors for MDS is limited, and further research in larger studies is needed.

No potential conflicts of interest were disclosed.

Conception and design: J.N. Poynter, K. Robien

Development of methodology: M.R. Sweeney, H. Arem, K. Robien

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.R. Sweeney, K. Robien

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.R. Sweeney, K.M. Applebaum, H. Arem, B.H. Braffett, J.N. Poynter, K. Robien

Writing, review, and/or revision of the manuscript: M.R. Sweeney, K.M. Applebaum, H. Arem, B.H. Braffett, J.N. Poynter, K. Robien

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.R. Sweeney, K. Robien

Study supervision: K. Robien

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.