Association between increased body mass index and a diagnosis of ...

78 downloads 88 Views 223KB Size Report
Association between increased body mass index and a diagnosis of acute promyelocytic leukemia in patients with acute myeloid leukemia. E Estey1, P Thall2, ...
Leukemia (1997) 11, 1661–1664  1997 Stockton Press All rights reserved 0887-6924/97 $12.00

Association between increased body mass index and a diagnosis of acute promyelocytic leukemia in patients with acute myeloid leukemia E Estey1, P Thall2, H Kantarjian1, S Pierce1, S Kornblau1 and M Keating1 Departments of 1Hematology and 2Biomathematics, University of Texas MD Anderson Cancer Center, Houston, TX, USA

We had observed that several patients with acute promyelocytic leukemia (APL) were markedly obese. Therefore we determined the relationship between obesity and a diagnosis of APL among patients with acute myelocytic leukemia (AML). Between 1980 and 1995 we saw 1245 patients with newlydiagnosed AML of whom 120 had APL. Increasing body mass index (BMI) was strongly associated with a diagnosis of APL (P = 0.0003). Like Douer et al (Blood 1996; 8: 303–313) we found APL to be more frequent in Latinos and younger patients (P , 10−4 for both). Logistic regression indicated that increasing BMI, decreasing age, and Latino origin were each independently associated with a diagnosis of APL (multivariate P values ,10−4, ,10−4, 0.0035, respectively). Since the mean BMI in the non-APL patients (25.1) resembled that of the general US population, it appears that APL patients are ‘heavy’ and not that non-APL patients are ‘thin’. Five of the APL patients (4.2%) had a BMI .50 (vs none of the other 1125 AML patients). Given the distribution of BMI in the general US population ages 17–74, the probability that five of 120 normal adults would have a BMI .50 is virtually nil. Excluding the five very heavy APL patients does not alter the conclusion that increasing BMI predicted for APL in patients with AML. Although the mechanism is unclear there appears to be an association between increasing BMI and a diagnosis of APL among patients with AML. There may also be an association between APL and obesity. Keywords: APL; BMI; obesity

Introduction Acute promyelocytic leukemia (APL) is distinguished from other types of acute myeloid leukemia (AML) by the presence of a PML-RAR fusion protein formed consequent to a t(15;17), a hemorrhagic diathesis, and a unique responsiveness to alltrans retinoic acid and anthracyclines.1 Recently, Douer et al2 reported a higher than expected incidence of APL among Latinos with AML, both at Los Angeles County-University of Southern California Medical Center and in separate cases identified by the Los Angeles County Cancer Surveillance Program. Over the last 15 years at MD Anderson we have noted that several patients with APL were markedly obese, often in excess of 300 pounds. Douer et al’s report prompted us to examine formally the relationship between APL, body mass index (BMI), ethnicity, and age in patients with newly diagnosed AML.

diagnosis had the t(15;17). Molecular testing for the t(15;17) was only begun in 1991. Since then six of the seven patients with APL without the t(15;17) on conventional cytogenetic analysis have had evidence of the t(15;17) on molecular analysis. This suggests that the great majority of patients who presented before 1991 with morphologic APL without the t(15;17) on conventional cytogenetic analysis would have had molecular evidence for this translocation had testing been performed. Regardless, however, we did the analyses described below including all patients with a morphologic diagnosis of APL, and separately, including only patients with evidence for the t(15;17) on conventional cytogenetic analysis. The values for age, height and weight were those recorded at presentation to the hospital for treatment. BMI, calculated as weight in kg/(height in meters),2 was used as the measure of obesity.5 The ethnic and racial origins of the patients were taken as those recorded by the patients on the MD Anderson hospital admission form. White, Black, Latino and Asian groups were recognized. The great majority of Latino patients were Texas residents of Mexican ancestry. The Asian group included patients of Chinese, Korean, Japanese and Filipino origin. The Wilcoxon Mann–Whitney test was used to compare the distributions of numerical-valued variables between patients with APL and patients with other types of AML. Association between qualitative variables was assessed by the generalized Fisher exact test. Effects of covariates on the probability an AML patient had APL were evaluated by logistic regression. All computations were carried out on a DEC Alpha Server 2100 5/250 running OSF/1. Results Table 1 and Figure 1 illustrate the relationship between BMI and diagnosis of APL among patients with AML. The association between BMI and APL was the same when attention was restricted to those APL patients documented to have the t(15;17) on conventional cytogenetic analysis (P = 0.0003).

Table 1

Relationship between body mass index and APL

APL Other AML (120 patients) (1125 patients)

Patients and methods 1245 patients with newly diagnosed AML as previously defined3,4 received initial treatment between 1980 and 1995. Of the patients 120 (9.6%) had APL. APL was diagnosed morphologically.4 Seventy-six percent of the patients with this

Correspondence: EH Estey, Department of Hematology, Box 61, UT MD Anderson Cancer Center, 1515 Holcombe Blvd, Houston, Texas 77030, USA Received 27 March 1997; accepted 28 May 1997

25th percentile of BMI distribution 50th percentile of BMI distribution 75th percentile of BMI distribution Minimum, Maximum BMIs

22.8

21.3

26.0

24.2

28.9

27.6

16, 60.7

9.9, 49.1

P = 0.0003, Wilcoxon Mann–Whitney test for hypothesis of no difference between distribution of BMI in APL and other AML patients.

Increased BMI and APL E Estey et al

1662

Table 3A Logistic regression model predicting APL among AML patients including all 120 APL patients

Variable

Coefficient estimate

s.e.

P value

Intercept BMI Age Latino

−2.246 0.0720 −0.0436 0.7279

0.4663 0.0151 0.0063 0.2400

— ,10−4 ,10−4 0.0035

Model deviance 699.5 on 1241 dF. Predictive equation

pˆ = −2.246 + 0.0720 • BMI − 0.0436 1 − pˆ • Age + 0.7279 • Indicator (Latino)

In

Figure 1 Histograms indicating number of cases of APL (top) and other types of AML (bottom) with indicated values for BMI.

Like Douer et al,2 we found a strong tendency for Latino patients with AML to have APL (Table 2). Again results were similar if only patients with the t(15;17) on standard analysis were considered. There was also a suggestion that those patients of Oriental background who had AML were relatively likely to have APL, although the small number of Oriental patients (22) prevents a firm conclusion (Table 2). Finally, an analysis analogous to that illustrated in Table 1 indicated that APL patients were younger than patients with other types of AML (P , 10−4, with median ages of 36 and 56 years, respectively). BMI and race were interrelated (P , 10−4, Kruskal–Wallis test) with mean BMI values of 22.33, 25.09, 26.38 and 26.84 for Orientals, Whites, Latinos and Blacks, respectively. Similarly, age and ethnic origin were interrelated (P , 10−4 Kruskal–Wallis test) with Latinos and Orientals on average 10 years younger than Whites and Blacks (mean ages 43.9, 42.9, 52.8 and 51.1 years for these four groups, respectively). BMI and age were not related. Logistic regression was used to sort out these interrelationships and determine which factors were predictive of APL among patients with AML. The final logistic model (Table 3A) indicates that BMI, age and Latino (as opposed to White, Black or Oriental) origin were all independent predictors of APL. This can also be appreciated by examining the contour plots of Figure 2a and b which show the estimated probability of APL according to BMI and age in both Latinos and nonlatinos. As can be seen in Figure 1, five of our 120 patients with APL had BMI .50 (vs none of the 1125 patients with other types of AML). These five (three males and two females) Table 2

Relationship between ‘ethnicity’ and APL

Ethnic group

White Black Latino Oriental

Patients

APL (%)

95% CI for % APL

980 87 156 22

75 (7.7) 9 (10.3) 32 (18.2) 4 (20.5)

6.2–9.5 5.7–18.7 15.0–27.7 7.8–40.3

P , 10−4, Fisher exact test for hypothesis of no difference between percent of patients with APL in the different ethnic groups.

pˆ = Estimated probability of APL BMI and age are continuous; (Latino): 1 = if Latino, 0 = if White, Black or Oriental.

Table 3B Logistic regression model predicting APL among AML patients excluding the five of 120 APL patients with BMI .50

Variable

Coefficient estimate

s.e.

P value

Intercept BMI Age Latino

−1.9065 0.0560 −0.0420 0.7383

0.4985 0.0173 0.0063 0.2419

— 0.0016 ,10−4 0.0034

Predictive equation

pˆ = −1.9065 + 0.0560 • BMI − 0.0420 1 − pˆ • Age + 0.7383 • Indicator (Latino) In

pˆ = BMI, age and (Latino) are as noted in Table 3A.

weighed 316, 369, 400, 418 and 480 pounds. The first of these five was a 37-year-old white woman, the second a 23year-old Latino woman, the third a 40-year-old white woman, the fourth a 40-year-old white man and the fifth a 28-yearold Latino man. Omitting these five patients from the analysis did not alter the conclusion that increasing BMI is an independent predictor of development of APL (Table 3B) Discussion While confirmatory of Douer et al’s2 report of an increased incidence of APL in Latinos with AML, our report is, we believe, the first to note an association between increasing BMI and APL. The mean BMI in our 1125 non-APL patients was 25.08 with a standard deviation of 5.28 vs 27.65 and 7.79, respectively, in our APL patients. While these means do not appear to differ greatly, the apparently small difference in BMI across the two samples of APL and non-APL patients has a profound effect on the probability of APL (P , 10−4 Table 3). As noted above, this effect is not due to the extreme BMI values in five of the APL patients, since it persists when these patients are excluded. Since the mean BMI in the non-APL patients resembled that of the general US population (see below), it appears that APL patients are ‘heavy’ and not that patients with other types of AML are ‘thin’ or ‘cachectic’. Such cachexia could in theory develop during a pre-leukemic

Increased BMI and APL E Estey et al

Figure 2 (a) Estimated probability of developing APL among Latinos with AML according to age and BMI. For example, a 20-yearold Latino with AML and a BMI of 60 would have an estimated probability of APL of 0.87 and an estimated probability of other types of AML of 0.13. A 20-year-old Latino with AML and a BMI of 25 would have an estimated probability of APL of 0.35 and of other types of AML of 0.65. (b) Same as (a) except that non-Latinos are now considered.

phase of the illness, such phases being more common in patients with other types of AML than in patients with APL. It remains to be seen if the association between increasing BMI and a diagnosis of APL among patients with AML will be limited to patients presenting to us. An international metaanalytic study would be useful in confirming our results. It was interesting that 20.5% of Orientals with AML had APL; although the 95% confidence interval for this rate was wide, it barely overlapped the 95% confidence interval for the proportion of white AML patients who had APL (Table 2). However, the number of Oriental AML patients was insufficient to see if ‘Oriental’ was a predictor of APL in patients with AML. In two large studies of the Japan Acute Leukemia Study Group (JALSG) the incidence of APL was 20 and 18% among 326 and 252 patients, respectively, with AML.6,7 This compares with an incidence of 10% in our 1245 AML patients. The median age of the patients in the two JALSG studies was 48 years, while the median age in our 1245 patients was 54 years. The younger age of the Japanese patients may account for the higher proportion of APL cases, since APL patients in Japan, as at MD Anderson, appear younger than patients with other types of AML.8 It should be pointed out that most of our Oriental patients were not of Japanese background. A metaanalysis would again be useful to investigate the incidence of APL in ‘Oriental’ patients with AML. Our finding that younger AML patients are more likely than older patients to have APL has been reported by others.9–11 As did Douer et al,2 we restricted attention to patients with AML. Thus we cannot comment on a possible link between obesity and development of APL in the general population.

Nonetheless, assuming that BMI is normally distributed with a mean of 25.3 and standard deviation of 4.0 for US males ages 18–74,12 the probability of a male with a BMI >50 is ,0.0001. Hence, the probability that three out of our 60 males with APL would have a BMI >50 is virtually nil under the null hypothesis that our male APL population and the general male population do not differ in BMI. Similar considerations apply to the female population (mean BMI 25.0, standard deviation 5.6 for US females ages 18–74).12 Hence our data suggest a relationship between APL and obesity. Although our data indicate an association between increasing BMI and a diagnosis of APL, at least among AML patients, the explanation for the association is obscure. We considered two explanations for the phenomenon we discovered although both were purely speculative and none appeared satisfactory. First, was the possibility that adipose tissue sequesters endogenous retinoids. Such sequestration could prevent functional activation of PML-RAR fusion protein and, consequentially, differentiation of promyelocytes. 13 This hypothesis presupposes that not all people with this fusion protein develop APL; only those with inadequate concentrations of retinoids in promyelocytes do so. In addition to the conceptual difficulty with this formulation, it is known that although adipocytes are important sites of synthesis of retinol-binding protein and although adipose tissue accounts for a substantial proportion of total body endogenous retinoid, retinoids in adipose tissue are in dynamic equilibrium with retinoids in other sites.14 Hence, a sequestration hypothesis is unlikely. Second, we considered the possibility that a gene involved in obesity could be localized to a chromosomal area involved in the genesis of the t(15;17). The most likely current candidate would appear to be the gene coding for the leptin receptor. Because plasma concentrations of leptin are highly correlated with percentage of body fat (although concentrations do vary at any single percent),15 it has been proposed that obesity is associated with leptin resistance resulting from a disordered leptin receptor. The human gene coding for this receptor has now been localized to chromosome 1p.16 It is of interest that the leptin receptor has recently been shown to have a nearly identical ligand binding domain as a sequence (B219) that is expressed in enriched hematopoietic stem cells.17 It is tempting to speculate that abnormalities in B219 play a role in development of both obesity and APL although it is unclear how the t(15;17) would fit into this formulation. Acknowledgments We thank Soon Woo for her expert assistance in preparing the manuscript. We have also benefitted from discussions of the manuscript with Drs Scott Lipmann and Waun Ki Hong at MD Anderson, Peter Davis at the University of Texas Medical School in Houston, William Blaner at Columbia University, New York and Jules Hirsch at Rockefeller University, New York.

References 1 Grignani F, Fagioli M, Alcalay M, Longo L, Pandolfi PP, Donti E, Biondi A, Lo Coco F, Grignani F, Pelicci PG. Acute promyelocytic leukemia: from genetics to treatment. Blood 1994; 83: 10–25. 2 Douer D, Preston-Martin S, Chang E, Nichols PW, Watkins KJ, Levine AM. High frequency of acute promyelocytic leukemia among Latinos with acute myeloid leukemia. Blood 1996; 87: 308–313.

1663

Increased BMI and APL E Estey et al

1664

3 Bennett JM, Catovsky D, Daniel MT, Flandrin G, Galton DA, Gralnick HR, Sultan C. Proposed revised criteria for the classification of acute myeloid leukemia. Ann Intern Med 1985; 103: 620–625. 4 Dalton W Jr, Stass SA. A morphologic and cytochemical approach to the diagnosis of acute leukemia. In: Stass SA (ed). The Acute Leukemia: Biologic, Diagnostic, and Therapeutic Determinants. Marcel Dekker: New York, Basel, 1987, p 27. 5 Berdanier C. Obesity. In: Munson PL, Mueller RA, Breese GR (eds). Principles of Pharmacology: Basic Concepts and Clinical Applications. Chapman and Hall: New York, 1995, p 1045. 6 Kobayashi T, Miyawaki S, Tanimoto M, Kuriyama K, Murakami H, Yoshida M, Minami S, Minato K, Tsubaki K, Ohmoto E, Oh H, Jinnai I, Sakamaki H, Hiraoka A, Kanamaru A, Takahashi I, Saito K, Naoe T, Yamada O, Asou N, Kageyama S, Emi N, Matsuoka A, Tomonaga M, Saito H, Ueda R, Ohno R for the Japan Adult Leukemia Study Group. Randomized trials between behenoyl cytarabine and cytarabine in combination induction and consolidation therapy, and with or without ubenimex after maintenance/intensification therapy in adult acute myeloid leukemia. J Clin Oncol 1996; 14: 204–213. 7 Ohno R, Kobayashi T, Tanimoto M, Hiraoka A, Imai K, Asou N, Tomonaga M, Tsubaki K, Takahashi I, Kodera Y, Yoshida M, Murakami H, Naoe T, Shimoyama M, Tsukada T, Takeo T, Teshima H, Onozawa Y, Fujimoto K, Kuriyama K, Horiuchi A, Kimura I, Minami S, Miura Y, Kageyama S, Tahara T, Masaoka T, Shirakawa S, Saito H. Randomized study of individualized induction therapy with or without vincristine, and of maintenance-intensification therapy between 4 or 12 courses in adult acute myeloid leukemia. Cancer 1993; 71: 3888–3895. 8 Kanamaru A, Takemoto Y, Tanimoto M, Murakami H, Asou N, Kobayashi T, Kuriyama K, Ohmoto E, Sakamaki H, Tsubaki K, Hiraoka A, Yamada O, Oh H, Saito K, Matsuda S, Minato K, Ueda T, Ohno R, and the Japan Adult Leukemia Study Group. Blood 1995; 85: 1202–1206. 9 Schiffer CA, Lee EJ, Tomiyasu T, Wiernik PH, Testa JR. Prognostic

10

11

12

13

14

15

16 17

impact of cytogenetic abnormalities in patients with de novo acute nonlymphocytic leukemia. Blood 1989; 73: 263–270. Mertelsmann R, Thaler HT, To L, Gee TS, McKenzie S, Schauer P, Friedman A, Arlin Z, Cirrincione C, Clarkson B. Morphological classification, response to therapy, and survival in 263 adult patients with acute nonlymphoblastic leukemia. Blood 1980; 56: 773–781. Keating MJ, Cork A, Broach Y, Smith T, Walters RS, McCredie KB, Trujillo J, Freireich EJ. Toward a clinically relevant cytogenetic classification of acute myelogenous leukemia. Leukemia Res 1987; 11: 119–133. Vital and Health Statistics: Anthropometric Reference Data and Prevalence of Overweight United States 1976–80. DHHS Publication No. (PHS) 87-1688. US Department of Health and Human Services, Public Health Service, National Center for Health Statistics: Hyattsville, MD, October 1987. Mangelsdorf DJ, Umesono K, Evans RM. The retinoid receptors. In: Sporn MB, Roberts AB, Goodman DS (eds). The Retinoids: Biology, Chemistry, and Medicine, 2nd edn. Raven Press: New York, 1994, p 319. Soprano DR, Blaner WS. Plasma retinol-binding protein. In: Sporn MB, Roberts AB, Goodman DS (eds). The Retinoids: Biology, Chemistry, and Medicine, 2nd edn. Raven Press: New York, 1994, p 257. Maffei M, Halaas J, Ravussin E, Pratley RE, Lee GH, Zhang Y, Fei H, Kim S, Lallone R, Ranganathan S, Kern PA, Friedman JM. Leptin levels in human and rodent: measurements of plasma leptin and ob RNA in obese and weight reduced subjects. Nature Med 1995; 1: 1155–1161. Chung WK, Power-Kehoe L, Chun M, Leibel RL. Mapping of the OB Receptor to 1p in a region of nonconserved gene order from mouse and rat to human. Genome Res 1996; 6: 431–438. Cioffi JA, Shafer AW, Zupancic TJ, Smith-Gbur J, Mikhail A, Platika D, Snodgrass HR. Novel B219/OB receptor isoforms: possible role of leptin in hematopoiesis and reproduction. Nature Med 1996; 2: 585–589.