Original Article
De Novo Acute Myeloid Leukemia Risk Factors A Texas Case-Control Study Sara S. Strom, PhD1; Robert Oum, PhD1; Kplola Y. Elhor Gbito, MD, MPH1; Guillermo Garcia-Manero, MD2; and Yuko Yamamura, BA1
BACKGROUND: Acute myeloid leukemia (AML) is comprised of several bone marrow-based cancers and is the most common type of leukemia in the United States. The etiology of AML is not well understood. A case-control study was conducted at The University of Texas M. D. Anderson Cancer Center to investigate associations between lifestyle characteristics and the risk of AML in Texas. METHODS: This study included 638 adult patients with de novo AML (cases) and a group of 636 matched controls. Intervieweradministered questionnaires were used to collect demographic and occupational data. The distribution of cases by World Health Organization (WHO) subtype was 71 patients (11%) with recurrent cytogenetic abnormalities (AML-RCA), 134 patients (21%) with multilineage dysplasia (AML-MD), and 389 patients (61%) with AML not otherwise categorized (AML-NOC). Multivariate logistic regression analyses were performed among all AML cases and among both sexes and each WHO subgroup. RESULTS: Among men, heavy smoking (30 pack-years; odds ratio [OR], 1.86) and occupational solvent exposure at low levels (OR, 2.87) or moderate/high levels (OR, 4.13) statistically significantly increased the risk of AML. Among women, obesity (OR, 1.62) and solvent exposure to low levels (OR, 2.73) or moderate/high levels (OR, 3.90) increased the risk of AML. Across WHO subtypes, obesity was associated with a statistically significantly increased risk of AML-RCA (OR, 3.15), whereas solvent exposure increased the risk in all subtypes at low levels (AML-RCA: OR, 4.11; AML-MD: OR, 2.54) and moderate/high levels (AML-RCA: OR, 5.13; AML-MD: OR, 3.02). A joint effect between smoking and solvent exposure was observed, and the highest risk was observed among smokers who had solvent exposure (OR, 4.51). CONCLUSIONS: The current results suggested that several factors play a role in AML predisposition with possible C 2012 American Cancer Society. joint effects. Risk profiles for AML differed by sex and WHO subtype. Cancer 2012;118:4589-96. V KEYWORDS: acute myeloid leukemia, epidemiology, risk factors, smoking, solvents, obesity.
INTRODUCTION Acute myeloid leukemia (AML) is comprised of several bone marrow-based neoplasms that have clinical similarities but distinct morphologic, immunophenotypic, and cytogenetic subtypes. AML is the most common type of leukemia in the United States, accounting for 75% to 80% of all adult acute leukemias.1 It is estimated that there were 12,330 new cases of AML and 8950 deaths in the United States in 2010.2 The age-adjusted incidence rate for 2001 through 2005 was 3.6 per 100,000 population. AML cases are classified into de novo and therapy-related AML (t-AML), defined as AML preceded by chemotherapy and/or radiation treatment, which constitutes 5% to 15% of all AMLs.3 On the basis of morphologic and immunophenotypic characteristics, the World Health Organization (WHO) divides AML into 4 homogenous subgroups: AML with recurrent cytogenetic abnormalities (AML-RCA); AML with multilineage dysplasia (AML-MD), including AML after a myelodysplastic/myeloproliferative disorder; t-AML; and AML not otherwise categorized (AMLNOC). Although several studies have examined the epidemiologic risk factors associated with AML, its etiology is still not well understood. Studies have been hampered by inconsistent diagnosis, changing classification standards, the inclusion of t-AML with de novo AML, and small sample sizes. Some of the previous studies have suggested several potential risk factors for AML, including exposure to benzene and other organic solvents,4-6 chemotherapeutic agents,7 agrichemicals,8,9 tobacco smoke,10,11 and obesity.12 However, exposure to these risk factors is relatively rare and does not explain the majority of AML cases.13 The current population-based case-control study was conducted at the University of Texas M. D. Anderson Cancer Center (MDACC), 1 of the largest AML treatment centers in the United States. The objectives were to identify
Corresponding author: Sara S. Strom, PhD, Department of Epidemiology, Unit 1340, The University of Texas M. D. Anderson Cancer Center, 1155 Holcombe Boulevard, Houston, TX 77030; Fax: (713) 792-9568;
[email protected] 1 Department of Epidemiology, The University of Texas M. D. Anderson Cancer Center, Houston, Texas; 2Department of Leukemia. The University of Texas M. D. Anderson Cancer Center, Houston, Texas.
DOI: 10.1002/cncr.27442, Received: October 12, 2011; Revised: November 30, 2011; Accepted: December 28, 2011, Published online February 1, 2012 in Wiley Online Library (wileyonlinelibrary.com)
Cancer
September 15, 2012
4589
Original Article
demographic and epidemiologic factors associated with the risk of developing adult, de novo AML among Texas residents and explore differences in risk between sexes and WHO subtypes. MATERIALS AND METHODS This population-based case-control study included 638 adult patients with de novo AML (cases) and a control group of 636 individuals from Texas. Cases were adult patients (ages 18-80 years) who registered at MDACC between 2003 and 2007 with a confirmed diagnosis of AML, and there were no restrictions on sex or ethnicity for their inclusion. Cases were identified at their first visit and were enrolled prospectively into the study. The overall participation rate among cases was approximately 87%. Reasons for nonenrollment included refusals (9%) or too ill to participate/died before interview and no proxy was available (4%). Proxy interviews (with close family members) were required for 11% of the completed interviews, because the cases were either too ill or died before completing the full questionnaire after giving their informed consent. On the basis of WHO classification criteria, in the case group, there were 71 patients with AML-RCA, 134 patients with AML-MD, and 389 patients with AML-NOC in our study. WHO classification was not available for 44 patients. Clinical information, including karyotypes, was obtained from the MDACC clinical database. In accordance with previous conventions, cases were categorized according to their pretreatment cytogenetics as favorable (translocation 8,12 [t(8,12)], inversion 16 [inv(16)]), poor (5, 7, 3, or complex [3 abnormalities]), or intermediate (diploid, other abnormalities not otherwise categorized).14 Distributions of WHO and cytogenetic categories are presented in Table 1. By using our previously described methodology,15 controls were recruited using random digit dialing. Briefly, a standardized telephone script was used to introduce the study, establish eligibility, and determine willingness to participate. Controls had no previous history of invasive cancer and were frequency-matched to cases on age (5 years), sex, race, and county of residence. Seventy-seven percent of the eligible controls participated in the study. Informed consent was obtained before data collection in accordance with institutional review board requirements. Interviewers administered the structured questionnaire within 2 months of patient registration (cases) or enrollment (controls) to assess demographic information, medical history, occupational history, lifetime smoking, exposure to second-hand smoke, family history 4590
Table 1. Demographic Characteristics of Cases and Controls
No. of Patients (%) Characteristic
Cases, n 5 638
Controls, n 5 636
P
293 (45.92) 345 (54.08)
293 (46.07) 343 (53.93)
NS
504 (79) 87 (13.64) 34 (5.33) 13 (2.04) 54.2415.21
492 (77.36) 88 (13.84) 44 (6.92) 12 (1.89) 53.013.90
NS
430 (67.40) 208 (32.60)
298 (47.60) 328 (52.40)
< .0001
588 (92.16) 50 (7.84)
618 (97.17) 18 (2.83)
< .0001
71 (12) 134 (22.6) 389 (65.5)
— — —
68 (11.1) 363 (59.2) 182 (29.7)
— — —
Sex Women Men
Race White Hispanic Black Asian Age at diagnosis/interview: MeanSD, y
NSa
Education status 100 cigarettes in their lifetime were defined as ‘‘ever-smokers’’ and were further categorized as ‘‘current smokers’’ or ‘‘former smokers,’’ defined as those who had quit >1 year before diagnosis (cases) or interview (controls). Pack-years were calculated from the average number of packs smoked per day number of years smoked. ‘‘Heavy smoking’’ was defined as smoking 30 pack-years over the lifetime of the participant. The body mass index (BMI) (in kg/m2) was calculated using self-reported weight and height at diagnosis/ interview. Participants with a BMI 30 kg/m2 were categorized as obese. Participants who reported having any first-degree relatives with lymphoma, leukemia, multiple myeloma, or myelodysplastic syndromes were categorized as having a positive family history of hematopoietic cancer. Descriptive analyses were conducted using chisquare tests and Student t tests with SPSS statistical software (version 17.0; SPSS, Inc., Chicago, Ill). For categorical variables, Cochran-Armitage trend tests were used to estimate P trend values. Odds ratios (ORs) and corresponding 95% confidence intervals (CIs) were calculated using unconditional logistic regression. Variables with univariate ORs with P values .1 were evaluated for inclusion in a multivariable model that was constructed in a forward, stepwise manner and adjusted for relevant variables, including education and family history of hematopoietic cancer. Independent models were run for each sex and for each WHO subtype. Results were considered statistically significant at the a ¼ .05 level. RESULTS Demographic characteristics of the study participants are presented in Table 1. Because of successful matching, there were no differences between cases and controls with respect to age, sex, or race/ethnicity. Controls were more likely to have completed a bachelor’s degree (P < .0001) and were less likely to report having a positive family history of hematopoietic cancer (P < .0001). Because the incidence of AML differs by sex,2 we determined crude ORs by sex (Table 2). In both men and Cancer
September 15, 2012
women, being a current smoker (men: OR, 1.60; 95% CI, 1.08-2.38; women: OR, 1.93; 95% CI, 1.26-2.97) and a heavy smoking history (ie, 30 pack-years [men: OR, 2.45; 95% CI, 1.57-3.83; women: OR, 1.76; 95% CI, 0.98-3.16]) were associated with an increase in risk, which reached statistical significance only in men. Exposure to second-hand smoke was not associated significantly with AML risk in men or women (data not shown; N ¼ 1059 [men: OR, 0.82; 95% CI, 0.55-1.22; women: OR, 1.06; 95% CI, 0.68-1.67]). The magnitude of AML risk associated with intensity of exposure to solvents was similar between men and women. Low levels were associated with an almost 3-fold increase (men: OR, 2.99; 95% CI, 1.82-4.91; women: OR, 2.76; 95% CI, 1.46-5.21), and moderate/high levels were associated with an increase >4-fold (men: OR, 4.26; 95% CI, 3.00-6.04; women: OR, 4.18; 95% CI, 2.287.64). The OR associated with the cumulative exposure index indicated that both the duration and the intensity of exposure led to similar increases in risk between men and women. However, there were marked differences in the prevalence of exposures by sex. On the basis of the calculations of population-attributable risk, exposure to moderate/high levels of solvents accounted for 41% of the overall AML risk in men and 14% in women. Moderate/ high levels of agrichemical exposure also increased AML risk in men (OR, 4.71; 95% CI, 2.51-8.85). Because of the small number of women exposed to agrichemicals, ORs were not calculated. Obesity was associated with an increased risk in both men (OR, 1.32; 95% CI, 0.881.98) and women (OR, 1.9; 95% CI, 1.3-2.8), reaching statistical significance only in women. Including factors that were identified as significant in the univariate analyses, we constructed independent multivariable models for men and women (Table 3). A history of heavy smoking significantly increased the risk of developing AML only in men (OR, 1.86; 95% CI, 1.15-3.02). However, among women, being a current smoker was associated with AML risk (OR, 1.75; 95% CI, 1.12-2.72). Occupational exposure to solvents increased AML risk in both men (OR, 4.13; 95% CI, 2.79-6.12) and women (OR, 4.04; 95% CI, 2.14-7.64). The occupations most commonly associated with solvent exposure in men were auto mechanic (25%), oil field worker (13%), chemical plant operator/worker (13%), and gas station operator (10%); in women, the occupations were hairdresser (33%), chemical/rubber plant worker (23%), laborer (15%), and cosmetologist (8%). Obesity significantly increased the risk of AML among women (OR, 1.57; 95% CI, 1.02-2.41), and there was a 4591
Original Article Table 2. Univariate Analysis by Sex
Women
Men
No. (%) Variable
No. (%)
Cases, n 5 293
Controls, n 5 293
OR [95%CI]
170 (58) 51 (17.4) 72 (24.6)
196 (66.9) 54 (18.4) 43 (14.7)
1.00 1.09 [0.71-1.68] 1.93 [1.26-2.97]a
170 46 45 32
196 46 30 21
(66.9) (15.7) (10.2) (7.2)
210 (71.7) 33 (11.3) 50 (17.1)
263 (89.8) 15 (5.1) 15 (5.1)
5.97
Cases, n 5 345
Controls, n 5 343
OR [95%CI]
145 (42) 123 (35.7) 77 (22.3)
190 (55.4) 90 (26.2) 63 (18.4)
1.00 1.79 [1.27-2.53]a 1.60 [1.08-2.38]a
1.00 1.15 [0.73-1.82] 1.73 [1.04-2.87]a 1.76 [0.98-3.16] .013
145 63 64 73
190 55 59 39
(55.4) (16) (17.2) (11.4)
1.00 1.50 [0.98-2.29] 1.42 [0.94-2.15] 2.45 [1.57-3.83]a < .0001
127 (36.8) 50 (14.5) 168 (48.7)
235 (69.3) 31 (9.1) 73 (21.5)
2.05
1.00 2.76 [1.46-5.21]a 4.18 [2.28-7.64]a < .0001 1.03 [1.01-1.05]a
10.15
1.00 2.99 [1.82-4.91]a 4.26 [3.00-6.04]a < .0001 1.02 [1.01-1.03]a
267 (91.1) 11 (3.8) 15 (5.1)
284 (96.9) 6 (2) 3 (1)
1.00 1.95 [0.71-5.35] —
269 (78) 25 (7.2) 51 (14.8)
323 (95.3) 3 (0.9) 13 (3.8)
1.00 — 4.71 [2.51-8.85]a
115 (39.3) 77 (26.3) 101 (34.5)
138 (47.3) 89 (30.5) 65 (22.3)
1.00 1.04 [0.70-1.54] 1.87 [1.25-2.78]a .005
111 (32.2) 134 (38.8) 100 (29)
101 (29.5) 173 (50.4) 69 (20.1)
1.00 0.71 [0.50-1.00] 1.32 [0.88-1.98] .320
272 (92.8) 21 (7.2)
285 (97.3) 8 (2.7)
1.00 2.75 [1.20-6.32]a
316 (91.6) 29 (8.4)
333 (97.1) 10 (2.9)
1.00 3.06 [1.47-6.37]a
Smoking status Never Former Current
Pack years of smoking Nonsmoker