Anthropometric factors and risk of melanoma in women: A pooled analysis

Abstract

Anthropometric factors such as height, weight and body mass index are related to the occurrence of certain malignancies in women including cancers of the breast, ovary and endometrium. Several studies have investigated the relation between height and weight or body mass and the risk of cutaneous melanoma in women, but results have been inconsistent. We conducted a collaborative analysis of these factors using the original data from 8 case–control studies of melanoma in women (2,083 cases and 2,782 controls), with assessment of the potential confounding effects of socioeconomic, pigmentary and sun exposure-related factors. Women in the highest quartile of height had an increased risk of melanoma [pooled odds ratio (pOR) 1.3, 95% confidence interval (CI) 1.1–1.6]. We also found an elevated risk associated with weight gain in adult life of 2 kg or more (pOR 1.5, 95% CI 1.1–2.0). Stratifying by age at melanoma diagnosis (<50, ≥50 yr), we found this risk greater among women <50 yr of age. Associations were unaffected by adjustment for other known risk factors for melanoma. There was no evidence that the effects varied for different histologic subtypes of cutaneous melanoma. There was no association with body weight per se, body mass index, or body surface area, either recent or in young adulthood. In aggregate, data from these studies suggest that greater height and weight gain may be risk factors for cutaneous melanoma in women. © 2007 Wiley-Liss, Inc.

Anthropometric factors such as height, weight and body mass index are related to the occurrence of certain malignancies in women including cancers of the breast, ovary and endometrium.1, 2 Increasing height has been associated with increased risk of cancers of the breast and colon.2, 3 Obesity consistently has been associated with an increased risk of hormone-related cancers such as colon, breast (in postmenopausal women) and endometrium, and also kidney.4 Evidence for an effect on risk of melanoma, however, remains inconclusive.

Several studies have investigated the association between height and weight or body mass and the risk of melanoma in women.5–24 Interpretation of the results has been hampered by differences in methodology and inconsistent approaches to defining overweight and obesity. Moreover, only a few studies have examined associations according to histologic subtype14, 15, 25 or menopausal status,7, 8 which may be important given the possible link between reproductive and hormonal factors and risk of melanoma in women.6, 14, 26–31 The prevalence of overweight and obesity is growing dramatically in most parts of the world, and generally is higher in women than in men.32 This is predicted to have major consequences on the incidence of obesity-related diseases. Quantitative assessment of the association between obesity and melanoma is thus important, and may provide further insight into disease etiology. We have conducted a pooled analysis of 8 case–control studies to examine the relationship between height, weight, body mass index and weight gain on the risk of melanoma in women. The study is restricted to women as the pooled analysis collaboration originally was designed to examine the reproductive and sex hormone effects on the risk of melanoma in women.

BMI, body mass index; BSA, body surface area; CI, confidence interval; LMM, lentigo maligna melanoma; NM, nodular melanoma; OR, odds ratio; pOR, pooled odds ratio; SE, standard error; SSM, superficial spreading melanoma; WHO, World Health Organization.

Methods

A more detailed description of the general methods used in our collaborative analysis is provided elsewhere.33, 34 Strict criteria were used to minimize inter-study heterogeneity and ensure comparable study quality. Briefly, we analyzed studies completed as of July 1994 that included melanomas diagnosed both on an inpatient and outpatient basis, collected data through a personal interview on important risk factors for melanoma (i.e., pigmentary traits and sun exposure history), and included at least 100 women with melanoma and 100 women controls. Data were available for all but 1 study that met these criteria.35 Descriptive statistics for each of the analysis variables were compared with published results and provided to the original study investigators to ensure their accuracy. Table I summarizes the characteristics of the 8 studies that collected anthropometric data and that met our inclusion criteria.8, 11, 14, 26, 36–39 Seven of the 8 studies included in the pooled analyses were population-based; only the study by Bataille et al.36 used hospital/clinic controls.

Table I. Characteristics of the 8 Studies in the Pooled Analyses of Anthropometric Measures and Melanoma in Women

Study	Diagnosis years	Age range	Geographic location	Number of:		Case sources (response rate)	Histology	Control sources (response rate)	Exposure measures
				Cases	Controls
SSM, superficial spreading melanoma; NM, nodular melanoma; LMM, lentigo maligna melanoma; Other, unclassified or other melanoma.
Bataille et al. (1996)36	1989–1993	16–75	NE Thames region, UK	255	253	Hospital pathology reporting (61%)	153 SSM, 41 NM, 15 LMM, 46 Other	Hospital outpatients, general practice surgeries (95%)	Self-reported current height, weight
Gallagher et al. (1985)14	1979–1981	20–79	Western Canada	361	361	Cancer registries (88%)	269 SSM, 66 NM, 26 Other	Medical plan subscribers (59%)	Self-reported current height, weight; usual weight at age 15–21
Green et al. (1985)37	1979–1980	15–81	Queensland, Australia	114	114	Pathology laboratories (97%)	79 SSM, 9 NM, 23 LMM, 3 Other	State electoral roll (92%)	Self-reported current weight; weight at age 21
Holly et al. (1995)26	1981–1986	25–59	San Francisco, CA	452	935	Cancer registry (79%)	355 SSM, 61 NM, 13 LMM, 23 Other	Random digit dialing (77%)	Self-reported height, usual weight over past 5 yr; weight at age 25
Kirkpatrick et al. (1994)11	1984–1987	25–65	Seattle, WA	127	145	Cancer registry (80%)	88 SSM, 39 Other	Random digit dialing (80%)	Self-reported height, weight 2 yr prior to diagnosis for cases (comparable reference date for controls); weight at age 18
Østerlind et al. (1988)38	1982–1985	20–79	East Denmark	280	536	Cancer registry (92%)	187 SSM, 56 NM, 34 Other	Population register (81%)	Self-reported height, weight; weight in 20s
Smith et al. (1998)8	1987–1989	≥18	Connecticut, USA	308	233	CT tumor registry (76%)	191 SSM, 34 NM, 46 LMM, 37 Other	Random digit dialing (70%)	Self-reported height, weight
Zanetti et al. (1990)39	1984–1987	19–92	Turin, Italy	186	205	Cancer registry (89%)	97 SSM, 21 NM, 21 LMM, 47 Other	National health service register (51%)	Self-reported current height, weight

Analysis variables

Adult weight was reported by all 8 studies (at interview,14, 36–39 1 yr prior to interview,8, 11 or usual weight over the past 5 yr26), and height by all except Green and Bain.37 Weight in young adulthood was available for 5 studies,11, 14, 26, 37, 38 although it was reported differently for each study, as follows: at ages 15–21,14 at age 18,11 at age 21,37 at age 25,30 and “in your 20s”.38 Our definition therefore includes individuals considered by the WHO as “adolescents” (10–19 yr),40 and those defined as “youths” by the United Nations (15–24 yr). Weight was expressed as quartiles of the control-distribution within each study because the average weight in controls was not homogeneous between studies. Height was expressed as quartiles of the distribution for the combined control group of all studies. BMI (in adult/young adulthood) was classified using the World Health Organization (WHO) standard categories41: underweight (<18.5 kg m⁻²), normal weight (18.5–24.9 kg m⁻²), overweight (25–29.9 kg m⁻²) and obese (≥30 kg m⁻²), using normal weight as the reference category. Adult body surface area (BSA) (in square meter) was calculated using the method of Mosteller,42 and was expressed as quartiles of the distribution for the combined control group of all studies because the average height in controls across studies was very similar. Absolute adult lifetime weight change was computed as the difference between adult weight and weight in young adulthood. For the analysis of weight change, 3 categories were created: loss (2 or more kilograms), no change (loss less than 2 kg−gain less than 2 kg), and gain (2 or more kilograms), using no change as the reference category. We also further categorized weight gain (2–4.9 kg, 5–9.9 kg, 10 or more kilograms).

Statistical analysis

We used a 2-stage method of analysis to obtain study-specific odds ratios (ORs) and pooled odds ratios (pORs) and 95% confidence intervals (CIs).33 In the first stage, each study was analyzed according to its original design unless otherwise specified. For pair-matched studies, we used conditional logistic regression and for frequency matched studies, we used unconditional logistic regression and stratified by age (<35, 35–44, ≥45 yr). To evaluate inter-study variability, we examined the study-specific ORs and tested for statistical heterogeneity using a χ² test. The pooled exposure effect was estimated in a second-stage linear, mixed model as the average of the study-specific ORs and standard errors (SEs), weighted by the inverse marginal variances. The marginal variance is the sum of the individual study variances and the random study effect. In the absence of heterogeneity, the average of the individual study estimators was weighted by the inverse of the study-specific variances alone.33 In the presence of heterogeneity, the pooled estimator is the average of individual study estimators weighted by the sum of the individual study variances and the random effects as described by Stukel et al.33 We examined the data for potential sources of heterogeneity by stratifying on type of control group, population versus hospital-based controls and style of questionnaire (i.e., telephone versus in-person interview).

We evaluated the potentially confounding effects of several factors including hair, eye and skin color, freckling, family history of melanoma in a first degree relative, ethnicity, number of nevi on the arms, skin sensitivity to sun exposure, sun exposure history and educational level.34 Whenever possible we used standardized groupings.34, 43, 44 Described in detail in an earlier report,34 questions relating to sun exposure history varied considerably across studies. Therefore, we included the sunlight-related factors most strongly related to melanoma risk within each study. These variables included history of sunburns, sun exposure and migration to Australia.34 All estimates were age-adjusted. To assess the impact of other potentially confounding factors, we examined the percent change in the age-adjusted pOR with the addition of each factor. Variables resulting in a 10% or greater change in the estimate were included in the final models. To assess stratum-specific effects (e.g., age category or anatomical site), we broke the pairs and stratified on age (e.g., <35, 35–49, ≥45 yr).

Our main analyses were based on all melanomas combined. Additionally, we separately computed odds ratios for each of the primary exposure variables by histologic subtype [superficial spreading melanoma (SSM), nodular melanoma (NM) and lentigo maligna melanoma (LMM)]. In the analysis of rarer histologic subtypes (e.g., NM and LMM) we broke the pair-matched sets to increase statistical power and analyzed the studies adjusting for the original age categories using conditional logistic regression. For LMM, we were unable to analyze weight in young adulthood, weight change since young adulthood and body mass index due to small strata. Analyses were conducted using SAS (SAS Institute, Cary, NC).

Results

A total of 2,083 cases and 2,782 controls were included in the analysis of adult weight; 1,969 cases and 2,668 controls for height and adult BMI; 1,334 cases and 2,091 controls for weight in young adulthood; and 1,220 cases and 1,977 controls for BMI in young adulthood and weight change (Table I). The combined study group was predominantly White as most studies comprised over 98% White subjects. The means and ranges for height, BSA, weight and BMI, and distribution of BMI using the WHO cut-points for cases and controls by study are presented in Table II. The prevalence of obesity among controls ranged from 79, 38 to 11%,8 and the prevalence of overweight ranged from 1311 to 30%36 (Table II).

Table II. Means and Ranges for Height, Weight and BMI, and Distribution of BMI Using Who Cut-Points for Cases and Controls by Study

Study	Height (cm), mean (range)	Adult weight (kg), mean (range)	Adult BMI (kg/m^2), mean (range)	Adult BSA (m^2), mean (range)	Weight in young adulthood (kg), mean (range)	BMI in young adulthood (kg/m^2), mean (range)	Weight change1 (kg) mean (range)	WHO classification of adult BMI (kg/m^2)
								<18.5 n (%)	18.5–24.9 n (%)	25–29.9 n (%)	≥30 n (%)
BSA, body surface area. 1 Weight change calculated as (adult weight−weight in young adulthood).
Bataille et al. (1996)36
Cases	160.7 (140–181)	63.1 (41–115)	24.5 (17.6–44.9)	1.71 (1.29–2.22)				2 (1)	137 (50)	103 (38)	32 (11)
Controls	161.8 (144–180)	66.2 (41–111)	25.4 (16.4–43.6)	1.67 (1.31–2.26)				5 (2)	151 (60)	75 (30)	20 (8)
Gallagher et al. (1985)14
Cases	163.9 (137–185)	63.3 (39–112)	23.6 (14.2–45.3)	1.69 (1.32–2.22)	54.9 (39–90)	20.5 (14.6–38.6)	8.4 (−36 to 56)	21 (6)	241 (67)	66 (18)	31 (9)
Controls	163.5 (137–185)	64.3 (43–118)	24.1 (17.2–53.9)	1.70 (1.35–2.38)	56.1 (36–121)	21.0 (11.2–47.4)	8.2 (−55 to 50)	14 (4)	226 (63)	85 (24)	34 (9)
Green et al. (1985)37
Cases		63.8 (43–102)			55.0 (38–83)		8.5 (−12 to 45)
Controls		62.8 (44–121)			55.9 (41–89)		7.1 (−32 to 57)
Holly et al. (1995)26
Cases	165.3 (142–183)	63.0 (43–136)	23.1 (17.0–46.1)	1.70 (1.34–2.56)	58.3 (41–118)	21.3 (15.8–46.2)	4.8 (−46 to 55)	19 (4)	335 (74)	61 (14)	36(8)
Controls	164.2 (132–183)	61.9 (38–129)	22.9 (16.4–49.5)	1.67 (1.27–2.49)	57.3 (38–116)	21.2 (15.4–38.7)	4.6 (−46 to 43)	33 (3)	714 (77)	120 (13)	63 (7)
Kirkpatrick et al. (1994)11
Cases	165.1 (150–183)	65.9 (46–159)	24.2 (18.3–54.9)	1.73 (1.39–2.74)	57.4 (40–114)	21.0 (14.8–39.2)	8.5 (−15 to 46)	1 (1)	85 (67)	32 (25)	8 (6)
Controls	165.0 (150–180)	65.5 (42–123)	24.1 (17.0–45.2)	1.73 (1.36–2.43)	56.9 (41–82)	20.9 (16.5–32.0)	8.6 (−22 to 57)	3 (2)	98 (68)	31 (21)	13 (9)
Østerlind et al. (1988)38
Cases	164.6 (150–182)	61.6 (42–104)	22.8 (15.8–37.2)	1.69 (1.25–2.34)	55.8 (44–73)	20.9 (17.1–27.8)	6.9 (−22 to 33)	20 (7)	194 (70)	54 (19)	6 (4)
Controls	164.4 (145–183)	62.9 (36–115)	23.3 (14.8–39.4)	1.67 (1.36–2.23)	57.0 (38–80)	21.3 (14.0–30.1)	6.3 (−20 to 67)	30 (6)	371 (69)	95 (18)	37 (7)
Smith et al. (1998)8
Cases	163.2 (130–183)	63.0 (41–131)	23.8 (16.3–51.8)	1.68 (1.31–2.45)				25 (8)	195 (64)	54 (18)	32 (10)
Controls	162.6 (125–180)	62.6 (38–118)	23.8 (14.8–44.6)	1.67 (1.23–2.30)				26 (11)	133 (57)	48 (21)	25 (11)
Zanetti et al. (1990)39
Cases	161.4 (143–173)	60.8 (43–89)	23.3 (16.8–33.3)	1.65 (1.32–2.04)				13 (7)	121 (66)	39 (21)	11 (6)
Controls	160.6 (145–178)	62.4 (41–105)	24.2 (16.5–43.7)	1.66 (1.31–2.27)				6 (3)	119 (61)	53 (27)	18 (9)

Table III presents the pORs for melanoma in relation to anthropometric measures for all women combined, and stratified by age at diagnosis of melanoma (<50, ≥50). We observed an increased pOR for risk of melanoma in the highest quartile of height compared with the lowest quartile for women overall (pOR 1.3, 95% CI 1.1–1.6). When stratified by age the association appeared limited to women <50 yr of age (pOR 1.4; 95% CI 1.1–1.9). A forest plot (Fig. 1) displays the study-specific odds ratios (ORs) for the highest quartile of height compared with the lowest for all women, and those aged <50 and ≥50 yr of age at diagnosis, together with the pooled ORs. Women who gained >2 kg since young adulthood were at an increased risk of melanoma (pOR 1.5; 95% CI 1.1–2.0), and again the association was observed among women <50 yr of age (pOR 1.5; 95% CI 1.1–2.2). A forest plot of this association for all women combined, and those <50 yr and ≥50 yr of age at diagnosis shows the study-specific ORs together with the pooled OR (Fig. 2). Weight gain was further categorized (2–4.9 kg, 5–9.9 kg, 10 or more kilograms), and although the small sample size in some groups may have decreased our power to detect statistically significant variation, the results were similar. Increased risks were observed for all categories of weight gain for women overall: pOR 1.5 (95% CI 1.0–2.4) for weight gain of 2–4.9 kg; 1.4 (95% CI 0.8–2.4) for weight gain of 5–9.9; and 1.2 (95% CI 0.7–2.0) for weight gain of 10 more kilograms. Again, the effects were most notable for women <50 yr of age (gain of 2–4.9 kg, pOR 1.5 (95% CI 1.0–2.6) vs. 1.2 (95% CI 0.7–2.3) for women ≥50 yr of age; gain of 5–9.9 kg pOR 1.6 (95% CI 0.9–2.7) vs. 1.1 (0.6–2.0); and gain of 10 or more kilograms pOR 1.2 (95% CI 0.6–2.1) vs. 1.1 (95% CI 0.7–2.0). Melanoma risk was unrelated to weight itself or BMI, whether recent or in young adulthood. Similarly, there was no association with adult BSA. An analysis of BSA in young adulthood yielded similar results (data not presented).

Figure 1. Forest plot of the association between height and melanoma for all women, women aged <50 yr, and women aged ≥50 yr. Each line represents an individual study result with the width of the horizontal line indicating the 95% CI, the position of the box representing the point estimate and the size of the box being proportional to the statistical weight of the study. Pooled ORs are represented by diamonds.

Figure 2. Forest plot of the association between weight gain from early adulthood (>2 kg) and melanoma for all women, women aged <50 yr, and women aged ≥50 yr. Each line represents an individual study result with the width of the horizontal line indicating the 95% CI, the position of the box representing the point estimate and the size of the box being proportional to the statistical weight of the study. Pooled ORs are represented by diamonds.

Table III. Adjusted1 Pooled Odds Ratios (95% Confidence Intervals) for Melanoma in Women in Relation to Anthropometric Variables and Weight Change in Adult Life, Stratified by Age <50/≥50

Anthropometric measure	All women			Age < 50			Age ≥ 50
	Cases	Controls	Pooled OR (95% CI)	Cases	Controls	Pooled OR (95% CI)	Cases	Controls	Pooled OR (95% CI)
BSA, body surface area. 1 Adjusted for age; additional adjustment. 2 Hair color. 3 Freckling. 4 Nevi. 5 Eye color. 6 Education. 7 Exclusions based on incomplete collection of anthropometric measurements. 8 Significant heterogeneity, random effects model used (see text).
Height (cm)
Q1 (<160)	468	664	1.0 (ref)	202	345	1.0 (ref)	266	319	1.0 (ref)2
Q2 (160–162.6)	277	358	1.2 (0.9–1.5)	132	182	1.0 (0.7–1.5)	145	176	1.0 (0.7–1.5)
Q3 (162.7–167.6)	546	783	1.2 (0.9–1.4)	283	467	1.1 (0.8–1.5)	263	316	1.1 (0.8–1.5)
Q4 (>167.6)	670	854	1.3 (1.1–1.6)	418	543	1.4 (1.1–1.9)	252	311	1.1 (0.8–1.5)
Excluded7	8	11		3	4		5	7
Weight (kg)
Q1 (≤55)	445	606	1.0 (ref)	279	400	1.0 (ref)	166	206	1.0 (ref)
Q2 (55.1–61.0)	494	658	1.0 (0.8–1.4)	269	382	1.0 (0.6–1.6)	225	276	1.1 (0.6–1.7)
Q3 (61.1–68.2)	550	755	1.0 (0.7–1.6)	290	452	1.0 (0.6–1.7)	260	303	1.0 (0.6–2.1)
Q4 (>68.2)	580	743	1.1 (0.7–1.7)8	258	361	0.9 (0.5–1.9)8	322	382	1.1 (0.6–2.1)
Excluded7	15	22		5	11		10	11
Weight in young adulthood (kg)
Q1 (≤52.3)	319	479	1.0 (ref)	151	220	1.0 (ref)34	168	259	1.0 (ref)
Q2 (52.4–55.9)	276	443	1.0 (0.7–1.4)	142	246	0.8 (0.5–1.3)	134	197	1.1 (0.7–1.7)
Q3 (56.0–60.0)	297	432	1.2 (0.8–1.7)	179	263	1.2 (0.8–1.8)	118	169	1.1 (0.7–1.8)
Q4 (>60.0)	306	479	1.0 (0.8–1.4)	180	265	0.7 (0.2–3.2)8	126	210	1.0 (0.6–1.5)
Excluded7	160	297		131	258		39	29
Weight change since early adulthood
Loss of 2 or more kilograms	109	201	1.2 (0.7–2.7)	59	113	1.1 (0.5–2.0)45	50	88	1.2 (0.5–2.8)6
Gain/loss of <2 kilograms	208	432	1.0 (ref)	154	334	1.0 (ref)	54	98	1.0 (ref)
Gain of 2 kg or more	852	1,156	1.5 (1.1–2.0)	438	542	1.5 (1.1–2.2)	414	614	1.2 (0.6–2.2)
Excluded7	165	302		132	261		33	41
BMI (Adult) (kg m^−2)
<18.5	102	118	1.1 (0.8–1.6)	69	72	1.4 (0.9–2.1)	33	44	0.8 (0.4–1.6)24
18.5–24.9	1,304	1,811	1.0 (ref)	467	730	1.0 (ref)	547	678	1.0 (ref)
25–29.9	389	504	0.9 (0.8–1.1)	139	232	0.8 (0.6–1.1)	250	272	1.0 (0.8–1.4)
30+	157	210	0.9 (0.7–1.2)	67	93	0.9 (0.6–1.4)	90	117	0.9 (0.6–1.3)
Excluded7	16	25		6	11		10	14
BMI in young adulthood (kg m^−2)
<18.5	150	205	1.0 (0.7–1.5)	78	101	1.0 (0.5–2.1)	373	596	1.0 (ref)
18.5–24.9	849	1,342	1.0 (ref)	467	730	1.0 (ref)	69	99	1.0 (0.6–1.8)
25–29.9	50	106	0.7 (0.4–1.3)	26	57	0.7 (0.2–2.2)	24	49	0.8 (0.3–1.8)
30+	13	28	0.8 (0.2–3.5)	9	16	0.9 (0.1–314.4)	4	6	1.0 (0.1–23.6)
Excluded7	158	295		128	254		30	41
BSA (Adult) (m^2)
<1.58	465	659	1.0 (ref)	271	424	1.0 (ref)	194	235	1.0 (ref)
1.58–1.67	473	662	1.1 (0.9–1.3)	250	398	1.0 (0.8–1.4)	223	264	1.1 (0.8–1.5)
1.67–1.77	497	656	1.1 (0.9–1.3)	273	374	1.1 (0.8–1.5)	224	282	1.0 (0.7–1.4)
>1.77	517	666	1.0 (0.6–1.6)8	238	334	0.9 (0.5–1.6)8	279	332	1.0 (0.7–1.4)
Excluded7	16	25		6	11		10	14

We next examined odds ratios stratified by histologic subtype of melanoma. In this analysis, we found an elevated odds ratios in the highest quartile of height for SSM (pOR 1.3; 95% CI 1.0–1.7), and to a lesser extent NM (pOR 1.2; 95% CI 0.7–2.1), whereas the association was null for LMM (pOR 1.0, 95% CI 0.3–3.1) (Table IV). The association with weight gain of 2 kg or more was present for SSM (pOR 1.3, 95% CI 1.0–1.8). Odds ratios for NM were elevated for both weight loss and weight gain of 2 or more kilograms, but with wide CIs and lacked statistical significance (pOR 1.9, 95% CI 0.6–5.8 for weight loss and 1.8, 95% CI 0.8–4.2 for weight gain) and as mentioned, we were unable to analyze LMM.

Table IV. Adjusted1 Pooled Odds Ratios (95% Confidence Intervals) for Melanoma in Women in Relation to Anthropometric Variables and Weight Change in Adult Life, Stratified by Histologic Subtype

Anthropometric measure	SMM			NM			LMM
	Cases	Controls	Pooled OR (95% CI)	Cases	Controls	Pooled OR (95% CI)	Cases	Controls	Pooled OR (95% CI)
SSM, superficial spreading melanoma; NM, nodular melanoma; LM, lentigo maligna melanoma. 1 Adjusted for age; additional adjustment. 2 Hair color. 3 Ethnicity. 4 Family history of melanoma in a first degree relative. 5 Skin color. 6 Eye color. 7 Education. 8 Exclusions based on incomplete collection of anthropometric measurements. 9 Significant heterogeneity, random effects model used (see text).
Height (cm)
Q1 (<160)	258	569	1.0 (ref)	51	547	1.0 (ref)	18	351	1.0 (ref)2345
Q2 (160–162.6)	140	298	1.1 (0.8–1.5)	34	280	1.3 (0.6–2.5)	16	143	2.8 (0.9–8.5)
Q3 (162.7–167.6)	329	724	1.1 (0.8–1.4)	56	682	0.9 (0.5–1.7)	24	406	1.5 (0.6–4.3)
Q4 (>167.6)	464	820	1.3 (1.0–1.7)	83	757	1.2 (0.7–2.1)	16	469	1.0 (0.2–4.9)
Excluded8	1	11		0	11		2	11
Weight (kg)
Q1 (≤55)	323	621	1.0 (ref)	47	603	1.0 (ref)	18	479	1.0 (ref)2356
Q2 (55.1–61.0)	346	683	1.0 (0.8–1.3)	58	642	1.1 (0.7–1.8)	38	416	2.1 (0.9–4.9)
Q3 (61.1–68.2)	306	642	1.0 (0.7–1.2)	68	605	1.3 (0.7–2.1)	21	367	1.2 (0.5–3.1)
Q4 (>68.2)	330	628	1.0 (0.8–1.3)	65	579	1.2 (0.7–2.1)	24	338	1.2 (0.4–3.0)
Excluded8	2	22		3	22		2	22
Weight in young adulthood (kg)
Q1 (≤52.3)	231	479	1.0 (ref)	45	448	1.0 (ref)
Q2 (52.4–55.9)	200	443	0.9 (0.7–1.3)	37	398	1.0 (0.5–2.3)
Q3 (56.0–60.0)	227	432	1.2 (0.8–1.7)	37	397	1.2 (0.5–2.7)
Q4 (>60.0)	225	475	1.0 (0.7–1.4)	45	441	1.1 (0.5–2.3)
Excluded8	107	297		32	297
Weight change since early adulthood
Loss of 2 or more kilograms	76	201	1.0 (0.5–1.7)6	20	188	1.9 (0.6–5.8)37
Gain/loss of <2 kg	172	432	1.0 (ref)	19	415	1.0 (ref)
Gain of 2 kg or more	621	1,156	1.3 (1.0–1.8)	116	1,041	1.8 (0.8–4.2)
Excluded8	109	302		34	302
BMI (Adult) (kg m^−2)
<18.5	70	114	1.2 (0.4–3.6)79	12	111	1.3 (0.6–2.9)237
18.5–24.9	836	1,661	1.0 (ref)	147	1,563	1.0 (ref)
25–29.9	197	430	0.8 (0.7–1.1)	44	399	1.0 (0.7–1.6)
30+	86	190	0.8 (0.6–1.3)	18	159	1.3 (0.7–2.6)
Excluded8	2	25		3	25

We further conducted the analyses stratified by anatomical site of melanoma (Table V). Increased risks associated with the highest quartile of height were observed for melanomas of the head, lower limbs and upper limbs (pORs 1.5, 1.3 and 1.4, respectively), although trends were not evident for some sites and CIs overlapped unity. A weight gain of 2 kg or more was associated with risk of melanoma of the lower limbs (pOR 1.8; 95% CI 1.0–3.2), less so of the upper limbs and trunk and not of the head and neck.

Table V. Adjusted1 Pooled Odds Ratios (95% Confidence Intervals) for Melanoma in Women in Relation to Anthropometric Variables and Weight Change in Adult Life, Stratified by Anatomical Site of Melanoma

Anthropometric measure	Head and neck			Trunk			Lower limbs			Upper limbs
	Cases	Controls	p OR (95% CI)	Cases	Controls	p OR (95% CI)	Cases	Controls	p OR (95% CI)	Cases	Controls	p OR (95% CI)
1 Adjusted for age; additional adjustment. 2 Hair color. 3 Ethnicity. 4 Eye color. 5 Skin color. 6 Sun exposure. 7 Education. 8 Exclusions based on incomplete collection of anthropometric measurements. 9 Significant heterogeneity, random effects model used (see text).
Height (cm)
Q1 (<160)	54	642	1.0 (ref)2	165	642	1.0 (ref)	134	642	1.0 (ref)2	83	642	1.0 (ref)3
Q2 (160–162.6)	44	340	1.4 (0.8–2.5)	92	340	1.1 (0.7–1.6)	84	340	1.0 (0.6–1.5)	31	340	0.8 (0.4–1.5)
Q3 (162.7–167.6)	64	741	1.3 (0.8–2.3)	153	741	1.0 (0.7–1.3)	181	741	1.2 (0.9–1.7)	97	741	0.9 (0.6–1.4)
Q4 (>167.6)	66	741	1.5 (0.9–2.5)	185	791	1.1 (0.8–1.6)	207	791	1.3 (0.9–1.9)	138	791	1.4 (0.9–2.2)
Excluded8	1	11		1	11		0	11		2	9
Weight (kg)
Q1 (≤55)	50	669	1.0 (ref)	145	669	1.0 (ref)	160	669	1.0 (ref)4	90	669	1.0 (ref)
Q2 (55.1–61.0)	62	642	1.3 (0.8–2.2)	162	642	1.2 (0.9–1.6)	168	642	1.0 (0.8–1.4)	89	642	1.0 (0.7–1.6)
Q3 (61.1–68.2)	57	659	1.1 (0.7–1.9)	161	659	1.1 (0.8–1.6)	150	659	0.9 (0.6–1.2)	78	659	0.9 (0.6–1.4)
Q4 (>68.2)	83	647	1.4 (0.9–3.1)	144	647	0.9 (0.7–1.3)	172	647	1.0 (0.7–1.2)	120	647	1.3 (0.9–2.0)
Excluded8	3	22		1	22		2	22		1	20
Weight in young adulthood (kg)
Q1 (≤52.3)	28	448	1.0 (ref)	68	448	1.0 (ref)	110	448	1.0 (ref)	81	448	1.0 (ref)
Q2 (52.4–55.9)	23	398	1.0 (0.4–2.9)	83	398	1.4 (0.8–2.5)	89	398	1.0 (0.6–1.8)	45	398	0.6 (0.3–1.2)
Q3 (56.0–60.0)	30	397	1.6 (0.6–4.0)	87	397	1.5 (0.8–2.7)	87	397	1.2 (0.7–2.1)	62	397	0.9 (0.5–1.6)
Q4 (>60.0)	36	441	1.4 (0.5–3.4)	62	441	0.9 (0.5–1.8)	104	441	0.9 (0.2–4.0)9	68	441	0.9 (0.5–1.6)
Excluded8	9	297		49	297		79	297		23	297
Weight change since early adulthood
Loss of 2 or more kilograms	16	188	1.1 (0.3–3.4)25	26	188	1.0 (0.4–2.4)4	38	188	1.4 (0.6–3.1)6	22	188	0.9 (0.4–2.7)34
Gain/loss of <2 kg	25	415	1.0 (ref)	61	415	1.0 (ref)	50	415	1.0 (ref)	54	415	1.0 (ref)
Gain of 2 kg or more	74	1,041	0.8 (0.4–1.9)	206	1,041	1.3 (0.8–2.2)	289	1,041	1.8 (1.0–3.2)	174	1,041	1.3 (0.7–2.3)
Excluded8	10	302		50	302		81	302		23	302
BMI (Adult) (kg m^−2)
<18.5	14	111	2.1 (0.8–5.5)23	22	111	0.8 (0.4–1.7)3	40	111	1.3 (0.7–2.3)37	23	111	1.4 (0.7–2.9)37
18.5–24.9	133	1,713	1.0 (ref)	409	1,713	1.0 (ref)	412	1,713	1.0 (ref)	228	1,713	1.0 (ref)
25–29.9	55	473	1.1 (0.7–1.7)	114	473	0.8 (0.6–1.1)	111	473	0.9 (0.6–1.2)	71	473	1.1 (0.7–1.6)
30+	24	197	1.5 (0.8–2.7)	46	197	1.0 (0.6–1.7)	45	197	0.9 (0.5–1.4)	27	197	1.0 (0.5–1.7)
Excluded8	3	25		1	25		2	25		2	25

Sensitivity analysis excluding the single study that used hospital-based controls36 explained the heterogeneity in the pORs for weight and BSA (all women), but did not change the results. Excluding the single study that employed a telephone interview (rather than an in-person interview)11 did not explain the heterogeneity of any of the study results and also had no effect on the pORs.

Discussion

In this pooled analysis, we found that both height and weight gain were associated with an increased risk of melanoma particularly in women <50 yr of age, irrespective of BSA. This implies that a stable body weight may be more favorable with respect to risk of melanoma than a positive weight change. We found no clear association between weight or BMI and risk of any specific anatomic site or histologic subtype of melanoma.

Our results generally agree with the published literature including data from studies that were not analyzed in our pooled analyses. Our findings in relation to height and risk of melanoma are consistent with 35, 7, 18 of 5 other studies5, 7, 10, 18, 20 that have examined this relationship for women but did not meet the criteria for inclusion in our pooled analyses. We cannot exclude the possibility that our findings are due to differences in average height by country, although this is unlikely as the majority of the studies included in the pooled analyses were conducted in North America and studies in other parts of the world have also observed associations with height.5, 7, 18

The lack of association between BMI or weight and melanoma risk in our pooled analyses is consistent with the findings of 55, 10, 12, 15, 16 of 8 other case–control studies,5–7, 10, 12, 13, 15, 16 and 717–23 of 8 cohort studies.17–24 To our knowledge there are no previous studies that have examined the association between weight gain and risk of melanoma among women. Our pooled analyses excluded men because the original study aims of the analysis focused on sex steroid-related factors in women.27, 34 In men, height has been inconsistently associated with risk of melanoma.7, 10, 16, 19, 20 In agreement with the published literature for women, the majority of studies that have examined the relationship between weight or BMI and melanoma in men have reported no associations; however, 27, 10 of 37, 10, 16 case–control studies and 318, 45, 46 of 917–24, 45, 46 cohort studies reported positive associations with BMI but not with weight gain.46

Height (tallness) has been consistently associated with increased risk of breast47, 48 and colon cancer49, 50 in women. A number of hypotheses have been put forward to explain these associations. The first is that height is directly correlated with the number of cells that can undergo malignant transformation.51 For melanoma, this could be related to an increased area of skin and number of melanocytes at risk. We found no association, however, between BSA and risk of melanoma, and thus increased skin surface area is unlikely to be driving the association with height. A further hypothesis is that adult height may be an indicator of nutritional status or energy balance during childhood or adolescence. Animal studies have found that reduced caloric intake during development reduces the future risk of malignancy.52, 53 One cohort study that examined childhood energy intake supports the hypothesis of effects on overall cancer mortality.54 Another possibility is an endocrine etiology, since height is correlated with serum hormones that may play a role in carcinogenesis.55, 56 It also may be possible that a behavioral correlate of body height is related to sun exposure, such as participation in outdoor sporting activities,57 that increases the likelihood of developing melanoma. However, our adjustment for sun exposure-related factors generally had little to no effect on our results.

Weight gain alters the metabolism of endogenous hormones including the sex steroids, insulin and insulin-like growth factors, that may influence carcinogenesis through changes in the normal balance between cell proliferation, differentiation and apoptosis.55, 56, 58 Accumulating evidence suggests that adult weight gain may be a more relevant variable to assess the effect of weight on health than body weight or BMI.59 We cannot explain why the association with weight gain was stronger for melanomas on the lower limbs. The lack of significant heterogeneity by anatomical site suggests that this finding may have occurred by chance; nonetheless, this may warrant further investigation. Cutaneous melanoma displays different anatomic distributions in men and in women,18 and since 1950, the greatest increase in cutaneous melanoma incidence in White women has been on the limbs, and largely the lower limbs.60–62 Thune et al.18 suggested that this sex difference may be related to differences in hormonal milieu as well as differences in sun-related behavior. This hypothesis could be explored further in cohort studies where anthropometric factors and other risk factor data are available along with hormone levels.

Strengths of our study include the large number of cases and controls made possible by pooling data from 8 individual case–control studies. Pooling these data increased our statistical power to examine anthropometric measures in relation to melanoma using a standard methodological approach and well-defined measures of obesity. It also allowed sub-group analyses to examine the effects by age, histologic subtype and body-site distribution. Still, several limitations should be considered in interpreting the results. The studies contributing to the summary estimates are vulnerable to various types of bias. Only 1 of the 8 studies included in the pooled analysis used hospital versus population-based controls.36 The prevalence of obesity is known to be higher in hospitalized patients63; many diagnoses for which people are admitted to hospital are obesity related (e.g., cardiovascular disease and diabetes). This will result in an estimate of effect that is biased toward the null hypothesis of no association. Excluding this study, however, had no material effect on the study results. The response rates of the individual studies ranged from 61 to 97% in cases and 48 to 95% for controls. Seven of 8 studies, however, had case response rates over 75%, and 5 of 8 studies had control response rates over 75% and it is therefore unlikely that nonresponse could have resulted in appreciable bias.

All studies included in the pooled analyses relied on retrospective self-reported height, weight, and weight history information. Recall of body weight from different periods of time, particularly during earlier life is subject to recall bias.64 Women with higher BMI are more likely to underestimate weight, whereas underweight women are more likely to overestimate body weight.64–67 This could lead to nondifferential misclassification and attenuation of the true association between obesity and melanoma if there is one. We also cannot exclude the possibility of differential misclassification, i.e., that under-reporting of weight may occur unequally between cases and controls. There also is the possibility of selection bias due to self-selection into the control groups of more health conscious women, who are less likely to be overweight or obese.

Despite these limitations, the present study provides new evidence that height and weight gain may be risk factors for cutaneous melanoma in women, especially among women aged <50 yr. Future research that includes additional studies and data in men should be considered to clarify the potential role of anthropometric factors in melanoma etiology.