Anthropometric traits and risk of multiple myeloma: a pooled prospective analysis

Abstract

Background

Obesity is a risk factor for multiple myeloma (MM), yet results of prior studies have been mixed regarding the importance of early and/or later adult obesity; other measures of body composition have been less well studied.

Methods

We evaluated associations of early adult (ages 18–21) and usual adult body mass index (BMI), waist circumference, and predicted fat mass with MM by pooling data from six U.S. prospective cohort studies comprising 544,016 individuals and 2756 incident diagnoses over 20–37 years of follow-up. We used Cox proportional hazards models to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) for associations, adjusted for age and other risk factors.

Results

Each 5 kg/m² increase in usual adult BMI was associated with a 10% increased risk of MM (HR: 1.10; 95% CI: 1.05–1.15). Positive associations were also noted for early adult BMI (HR per 5 kg/m²: 1.14; 95% CI: 1.04–1.25), height (HR per 10 cm: 1.28; 95% CI: 1.20–1.37), waist circumference (HR per 15 cm: 1.09; 95% CI: 1.00–1.19), and predicted fat mass (HR per 5 kg: 1.06; 95% CI: 1.01–1.11).

Conclusions

These findings highlight the importance of avoidance of overweight/obesity and excess adiposity throughout adulthood as a potential MM risk-reduction strategy.

Background

Obesity is an established risk factor for multiple myeloma (MM) [1], a rare cancer of the plasma cells with poor survival [2]. Incidence rates of MM in the U.S. increased by 0.8% per year during 1975–2006 and by 1.9% per year during 2006–2016 [3], paralleling trends in obesity prevalence among adults [4]. In a 2019 meta-analysis of 23 prospective cohort studies comprising more than 7800 cases, the risk of MM increased by 6% per 5 kg/m² of adult body mass index (BMI) (hazard ratio (HR): 1.06; 95% confidence interval (CI): 1.03, 1.10)); for the majority of studies included in the meta-analysis, BMI was assessed at a single point during adulthood [5]. BMI earlier in life [6–9] and other measures of body composition, such as central adiposity [10–14], have been less well studied. Some previous studies reported that overweight and obesity as assessed by BMI in both young and middle adulthood were associated with increased MM risk [7–9]; however, others found no evidence of an association between young adult BMI and MM [6, 12, 15]. To date, obesity remains the only known modifiable risk factor for MM.

In this study, we aimed to prospectively evaluate associations of several anthropometric traits with MM, including young and middle adult BMI and several less-studied metrics by leveraging centrally harmonised, time-varying anthropometric measure and covariate data from U.S. cohort studies comprising more than 540,000 men and women.

Methods

Study populations

We pooled data from six U.S. prospective cohort studies with repeated measures of anthropometric traits and follow-up over a mean of 20 years, including the California Teachers’ Study (CTS), the Cancer Prevention Study-II (CPS-II) Nutrition Cohort, Health Professionals Follow-up Study (HPFS), and the Nurses’ Health Studies (NHS and NHSII), as well as electronic medical record (EMR) data from a case–control study nested within the Kaiser Permanente Southern California (KPSC) member cohort. Years of enrolment ranged from 1976 (NHS) to 2006–2014 (KPSC). See Supplementary Methods for more details about the collaborating cohorts.

In each of the cohorts, participants provided information on their personal medical history as well as behavioural and other risk factors for disease via baseline and follow-up questionnaires. For KPSC, data were obtained from the EMR. Exposures and covariates were harmonised across studies for data analysis. For the current analysis, we excluded men and women who had a diagnosis of cancer (except non-melanoma skin cancer) prior to study enrolment, or who had missing or implausible values of anthropometric variables (defined below). The analytic cohort included 544,016 men and women with more than 11 million person-years of follow-up.

This study was approved by the Institutional Review Boards of City of Hope, Brigham and Women’s Hospital, Harvard T.H. Chan School of Public Health, Kaiser Permanente Southern California, and those participating cancer registries as required. Informed consent was implied by the return of the baseline questionnaire.

Case ascertainment

Incident (first primary) MM diagnoses (ICD-8 code 203; ICD-O-2 code 9732; ICD-O-3 code 9732) were self-reported or identified through death records or linkage to state cancer registries. All cases were confirmed by review of medical records or cancer registries. After baseline exclusions, we confirmed 2756 incident MM diagnoses across the six studies.

Exposure assessment

Except for KPSC, for which EMR data were queried, measures of body composition and covariates were self-reported on baseline and follow-up questionnaires. Anthropometry traits of interest included adult height; “recent” weight (e.g., based on the most recent update; see below); baseline, recent, and usual adult BMI (see below); weight and BMI in early adulthood (age 18 or 21) (CTS, CPS-II, HPFS, NHS and NHSII only); weight change from early adulthood (CTS, CPS-II, HPFS, NHS and NHSII only); waist circumference (CTS, CPS-II, HPFS, NHS and NHSII only); and predicted fat mass (CTS, CPS-II, HPFS, NHS and NHSII only). Weight was ascertained approximately every two years throughout follow-up; waist circumference was updated when available (except CPS-II). We excluded data from individuals who had implausible values of weight or height at baseline [<0.5th percentile or >99.5th percentile of the distribution in the National Health and Examination Survey (NHANES) III [16]] or implausibly low values of waist circumference (<29 inches for men; <20 inches for women). BMI was calculated as weight in kilograms divided by height in metres squared (kg/m²). For each follow-up cycle, we calculated the cumulative average middle adult BMI as the mean of all available information on BMI from baseline to the beginning of the given follow-up cycle; we term this metric “usual adult” BMI. Weight change was defined as recent weight minus weight at age 18 or 21. Predicted fat mass was calculated for each follow-up cycle using sex-specific formulas based on age, height, weight, waist circumference and race/ethnicity; these formulas were derived and validated using data from NHANES [17].

Statistical analyses

Given no evidence of between-cohort heterogeneity based on a random-effects meta-analysis [18], we pooled data across all six studies. Person-time of follow-up was calculated for each participant from the return date of the baseline questionnaire (or enrolment year for KPSC) to the date of cancer diagnosis, death, or the end of follow-up (2014 or 2015), whichever occurred first. Sampling weights (inverse of the sampling probability) were applied to the controls in KPSC such that the KPSC study sample could be analysed as a cohort design in statistical analyses [19, 20]. We fit multivariable Cox proportional hazards models, for the total pooled study population and for men and women separately, stratified by cohort, age andfollow-up year and adjusted for the race (Black, white, other), educational attainment (<12 years, 12–15 years, 16 or more years), smoking history (never, past, current), alcohol consumption (current, not current), and height (when BMI was main exposure of interest) to estimate HRs and 95% CIs for risk of MM associated with each anthropometric variable. Where applicable, anthropometric and other variables were treated astime-varying in models. Although we did not observe meaningful confounding by covariates, we present the fully adjusted models for comparison with prior literature. For analyses among men and women combined, sex was included as an additional stratification variable. Height, waist circumference, and weight were modelled as continuous variables. BMI variables were categorised as <18.5, 18.5–22.9 (reference), 23–24.9, 25–29.9, 30–34.9, 35–39.9 and ≥40 kg/m² for recent and usual adult BMI and <18.5, 18.5–22.9 (reference), 23–24.9, 25–29.9 and ≥30 kg/m² for early adult BMI; linear tests for trend in BMI associations were based on models that included the continuous variable. Weight change was modelled in the following categories: loss ≥10 kg, loss 4.5–<10 kg, loss 2–<4.5 kg, stable (+/−2 kg) (reference), gain 2–<4.5 kg, gain 4.5–<10 kg, gain 10– <20 kg, gain 20–<30 kg and gain ≥30 kg. The predicted fat mass was analysed in quartiles. To assess the relative contribution of different measures of body size (e.g., BMI in early adulthood vs. usual adult BMI), we fit multivariable models that included mutual adjustment for the anthropometric trait of interest. We alsocross-classified categories of early and usual adult BMI and tested for statistical interaction based on a Wald test of the cross-product term. Tests for statistical interaction of anthropometric variables with sex and race were also based on a Wald test of the cross-product term. Finally, we evaluated possible effect modification by race. Missing indicator categories were used to account for missing values for exposures and covariates.

All statistical tests were two-sided, and P values <0.05 were considered significant.

Results

The majority (78%) of the combined study population was female, and 8% of the study population was non-white. Mean age at baseline was 50 years, and mean BMI at baseline was 24.9 kg/m². Other descriptive characteristics of the study population are shown in Table 1.

Table 1.

Pooled cohort characteristics.

Cohort	Total	NHS	NHSII	HPFS	CPS-II	CTS	KPSC
Cohort size	544,016	116,784	114,064	44,172	150,706	109,940	8350
Baseline year		1976	1989	1986	1992/1993	1995	2006–2014
Mean (range) follow-up time (years)	21.5 (0–37)	30.5 (0–37)	23.6 (0–25)	19.0 (0–25)	16.8 (0–23)	18.5 (0–20)	5.4 (0–7)
MM cases	2756	330	47	182	563	196	1438
Mean (range) baseline age (years)	50 (18−>90)	43 (30–56)	35 (25–44)	54 (39–79)	63 (40−>90)	52 (21−>90)	64 (18−>90)
Mean baseline BMI (kg/m2)	24.9	23.8	24.1	25.5	26.0	24.8	28.6
n (%) non-white	43,084 (8%)	7262 (6%)	8587 (8%)	4184 (9%)	3985 (3%)	14,707 (13%)	4359 (52%)
% Female	78%	100%	100%	0%	53%	100%	45%
% Not current alcohol consumption at baseline	37%	33%	43%	24%	42%	34%	N/A
% Never smoker at baseline	52%	43%	65%	45%	44%	66%	N/A
% College educated	71%	100%	100%	100%	38%	50%	N/A

MM multiple myeloma, BMI body mass index, NHS Nurses’ Health Study, HPFS Health Professionals Follow-up Study, CPS Cancer Prevention Study, CTS California Teachers Study, KPSC Kaiser Permanente Southern California, N/A not available at baseline.

We observed consistent positive associations of greater body size and risk of MM across all metrics evaluated. In models of recent BMI, in which BMI was updated at each follow-up cycle, the HR for risk of MM associated with each 5 kg/m² increase in BMI was 1.02 (95% CI: 0.98–1.07). The corresponding HR for baseline BMI was 1.12 (95% CI: 1.07–1.17). Considering usual adult BMI (i.e., the time-varying cumulative average BMI over follow-up), each 5 kg/m² increase was associated with a 10% increased risk of MM (HR: 1.10; 95% CI: 1.05–1.15) (Table 2). Results for usual adult BMI were similar among men (HR per 5 kg/m²: 1.06; 95% CI: 0.98–1.15) and women (HR per 5 kg/m²: 1.11; 95% CI: 1.05–1.17; P-interaction = 0.38). Increased risk of MM was most evident among men and women with usual adult BMI ≥ 30 kg/m², with nearly twice the risk of MM among individuals with usual adult BMI ≥ 40 kg/m² compared to those with BMI 18.5–22.9 kg/m² (HR: 1.77; 95% CI: 1.30–2.42). Early adult BMI was also associated with an increased risk of MM (HR per 5 kg/m²: 1.14; 95% CI: 1.04–1.25) among men and women combined. Results were similar in analyses that mutually adjusted for the two BMI metrics (Table 2). Cohort-specific results are presented in Supplementary Table 1.

Table 2.

Hazard ratios (95% confidence intervals) for usual and early adult BMI in relation to multiple myeloma risk, overall and by sex.

	Usual adult BMI				Early adult BMI†
Overall
Category, kg/m2	Person-years	N	HR* (95% CI)	HR** (95% CI)	Person-years	N	HR* (95% CI)	HR** (95% CI)
<18.5	166,935	23	0.83 (0.51–1.35)	0.83 (0.51–1.35)	1,318,718	137	0.96 (0.80–1.15)	0.96 (0.80–1.16)
18.5–22.9	4,018,691	504	1.00 (ref)	1.00 (ref)	7,043,833	760	1.00 (ref)	1.00 (ref)
23–24.9	2,392,003	475	1.02 (0.89–1.17)	1.02 (0.88–1.17)	1,402,809	175	1.00 (0.84–1.18)	0.98 (0.82–1.16)
25–29.9	3,586,084	1034	1.02 (0.91–1.16)	1.01 (0.89–1.14)	984,698	142	1.19 (0.99–1.43)	1.13 (0.94–1.38)
30–34.9^	1,159,348	484	1.16 (1.00–1.35)	1.13 (0.97–1.32)	214,719	26	1.46 (0.98–2.16)	1.25 (0.83–1.89)
35–39.9	358,289	161	1.10 (0.88–1.37)	1.06 (0.85–1.33)
40+	142,308	75	1.77 (1.30–2.42)	1.69 (1.23–2.31)
HR per 5 kg/m2	11,823,658	2756	1.10 (1.05–1.15)	1.09 (1.04–1.14)	10,964,777	1240	1.14 (1.04–1.25)	1.09 (0.99–1.21)
Men
Category, kg/m2
<18.5	3938	3	∞	∞	133,840	32	0.87 (0.60–1.27)	0.89 (0.61–1.29)
18.5–22.9	289,395	140	1.00 (ref)	1.00 (ref)	985,092	247	1.00 (ref)	1.00 (ref)
23–24.9	477,107	222	0.98 (0.77–1.24)	0.96 (0.76–1.23)	426,790	100	1.03 (0.81–1.30)	1.02 (0.80–1.29)
25–29.9	928,484	573	1.00 (0.80–1.24)	0.96 (0.77–1.20)	310,703	82	1.20 (0.93–1.54)	1.12 (0.85–1.47)
30–34.9^	206,626	240	1.04 (0.81–1.34)	0.99 (0.76–1.28)	28,839	7	1.25 (0.58–2.67)	1.00 (0.45–2.20)
35–39.9	33,584	74	1.13 (0.79–1.62)	1.08 (0.75–1.55)
40+	5158	20	1.48 (0.80–2.76)	1.41 (0.76–2.63)
HR per 5 kg/m2	1,944,292	1272	1.06 (0.98–1.15)	1.05 (0.96–1.14)	1,885,264	468	1.17 (1.00–1.37)	1.10 (0.92– 1.31)
Women
Category, kg/m2
<18.5	162,996	20	1.05 (0.65–1.72)	1.06 (0.65–1.73)	1,184,878	105	0.98 (0.79–1.21)	0.98 (0.79–1.21)
18.5–22.9	3,729,296	364	1.00 (ref)	1.00 (ref)	6,058,742	513	1.00 (ref)	1.00 (ref)
23–24.9	1,914,896	253	1.03 (0.86–1.23)	1.03 (0.86–1.23)	976,020	75	0.95 (0.75–1.21)	0.93 (0.73–1.20)
25–29.9	2,657,600	461	1.01 (0.87–1.18)	1.01 (0.87–1.18)	673,995	60	1.17 (0.89–1.53)	1.12 (0.84–1.48)
30–34.9^	952,722	244	1.23 (1.02–1.49)	1.22 (1.00–1.48)	185,880	19	1.54 (0.97–2.45)	1.35 (0.83–2.18)
35–39.9	324,705	87	1.04 (0.77–1.39)	1.01 (0.75–1.36)
40+	137,150	55	1.83 (1.28–2.62)	1.75 (1.22–2.53)
HR per 5 kg/m2	9,879,366	1484	1.11 (1.05–1.17)	1.11 (1.04–1.17)	9,079,513	772	1.12 (1.00–1.26)	1.09 (0.96– 1.23)

HR hazard ratio, CI confidence interval, BMI body mass index.

^†Early adult BMI is BMI at ages 18 or 21; excludes Kaiser Permanente Southern California (KPSC).

*HRs adjusted for age, sex, cohort, race, education, smoking, alcohol, height.

**Usual adult BMI (i.e., cumulative average BMI) additionally adjusted for categorical early adult BMI; early adult BMI additionally adjusted for usual adult BMI.

∞HRs based on fewer than five cases not shown.

^Highest category is 30+ kg/m² for early adult BMI.

N represents the number of MM cases in each category.

Bold font indicates statistically significant results.

Other anthropometric measures were also associated with MM risk (Table 3). Each additional 10 cm in adult height was associated with a 28% increase in the risk of MM (HR: 1.28; 95% CI: 1.20–1.37). Statistically significant associations for height were observed among both men (HR: 1.32; 95% CI: 1.20–1.46) and women (HR: 1.24; 95% CI: 1.13–1.36). Each additional 15 kg of recent body weight was associated with a 7% increase in MM risk (HR: 1.07; 95% CI: 1.02–1.11); however, associations were attenuated upon additional adjustment for height, particularly among men (HR: 0.94; 95% CI: 0.87–1.02). Finally, waist circumference was positively associated with MM risk (HR per 15 cm increase: 1.09; 95% CI: 1.00–1.19), based on a subset of the pooled population with available measurements (n = 895 cases) (Table 3). Predicted fat mass was also positively associated with MM risk (HR per 5 kg: 1.06; 95% CI: 1.01–1.11; Table 3); relative to the first (bottom) quartile of predicted fat mass, HRs (95% CI) associated with the second, third and fourth (top) quartiles were 1.25 (95% CI: 1.02–1.54), 1.25 (95% CI: 1.01–1.54), and 1.13 (95% CI: 0.91–1.41), respectively.

Table 3.

Hazard ratios (95% confidence intervals) for height, weight, waist circumference, and predicted fat mass in relation to multiple myeloma risk, overall and by sex.^†

	Height (per 10 cm)		Weight (per 15 kg)			Waist circumference^ (per 15 cm)		Predicted fat mass (per 5 kg)
	N	HR* (95% CI)	N	HR* (95% CI)	HR** (95% CI)	N	HR* (95% CI)	N	HR* (95% CI)
Overall	2756	1.28 (1.20–1.37)	2674	1.07 (1.02–1.11)	1.01 (0.96–1.06)	895	1.09 (1.00–1.19)	780	1.06 (1.01–1.11)
Men	1272	1.32 (1.20–1.46)	1231	1.05 (0.98–1.12)	0.94 (0.87–1.02)	351	1.13 (0.95–1.33)	316	1.05 (0.95–1.16)
Women	1484	1.24 (1.13–1.36)	1443	1.08 (1.02–1.14)	1.04 (0.98–1.11)	544	1.08 (0.98–1.19)	464	1.06 (1.00–1.12)

HR hazard ratio, CI confidence interval.

^†Except for height, all exposure variables were treated as time-varying (i.e., modelled as recent exposure).

*HRs adjusted for age, sex, cohort, race, education, smoking, alcohol.

**Additionally adjusted for height.

^{^}Includes data from California Teachers Study, Cancer Prevention Study II, Nurses’ Health Studies, Health Professionals Follow-up Study.

N represents the number of MM cases in each category.

Bold font indicates statistically significant results.

We found little evidence that weight gain from early to later adulthood was a strong predictor of MM risk, with the possible exception of individuals who gained 30 kg or moreover their lifetime: compared to those who maintained a stable weight, these individuals had more than a 20% increased risk of MM (HR: 1.23; 95% CI: 0.93–1.64; Table 4). The association was stronger among individuals who were also overweight or obese at ages 18 or 21 (HR: 1.46; 95% CI: 0.65–3.29); however, there were very few individuals in this category (n = 15 cases), and the effect estimate was imprecise (Table 4). Individuals who were heavy in both time periods (early and later adulthood) had a higher risk of MM than those who were lean (i.e., BMI < 25 kg/m²) in both time periods (Supplementary Table 2), with HRs ranging from 1.13 (95% CI: 0.85–1.48) for individuals with BMI 25−<30 kg/m² to 1.77 (95% CI: 1.08–2.89) for individuals with BMI ≥ 30 kg/m² in both time periods (P-interaction = 0.16). There was no evidence of reduced risk for individuals whose BMI was ≥25 kg/m² in early adulthood but <25 kg/m² in later adulthood; however, there were relatively few individuals in this group (Supplementary Table 2).

Table 4.

Hazard ratios (95% confidence intervals) for weight change from early adulthood to the most recent questionnaire in relation to multiple myeloma, overall and by early adult BMI.^†

	Overall			Early adult BMI < 25 kg/m2			Early adult BMI ≥ 25 kg/m2
Category	Person-years	N	HR* (95% CI)	Person-years	N	HR* (95% CI)	Person-years	N	HR* (95% CI)
Weight loss: ≥10 kg	213,893	21	0.84 (0.52–1.35)	54,389	5	0.66 (0.27–1.64)	159,504	16	0.93 (0.42–2.07)
Weight loss: 4.5–<10 kg	410,747	48	1.10 (0.77–1.57)	285,497	29	0.98 (0.64–1.51)	125,250	19	1.44 (0.67–3.12)
Weight loss: 2–<4.5 kg	333,765	30	0.87 (0.57–1.32)	281,981	23	0.84 (0.52–1.34)	51,783	7	1.23 (0.46–3.27)
Stable weight: +/− 2 kg	884,193	88	1.00 (ref)	796,439	75	1.00 (ref)	87,754	13	1.00 (ref)
Weight gain: 2–<4.5 kg	797,297	101	1.32 (0.99–1.77)	738,797	92	1.35 (0.99–1.83)	58,500	9	1.65 (0.66–4.12)
Weight gain: 4.5– <10 kg	2,316,223	226	0.96 (0.75–1.23)	2,146,846	203	0.96 (0.74–1.26)	169,377	23	0.99 (0.47–2.10)
Weight gain: 10 –<20 kg	2,913,831	360	1.05 (0.83–1.33)	2,707,584	326	1.03 (0.80–1.33)	206,247	34	1.63 (0.81–3.31)
Weight gain: 20 –<30 kg	1,507,570	188	1.04 (0.81–1.35)	1,373,537	171	1.03 (0.78–1.36)	134,033	17	1.11 (0.50–2.45)
Weight gain: ≥30 kg	888,436	113	1.23 (0.93–1.64)	770,964	98	1.17 (0.86–1.59)	117,472	15	1.46 (0.65–3.29)
Weight change per 5 kg	10,265,953	1175	1.02 (0.99– 1.04)	9,156,034	1022	1.02 (0.99–1.05)	1,109,919	153	1.03 (0.97–1.09)

HR hazard ratio, CI confidence interval, BMI body mass index.

^†Early adult BMI is BMI at ages 18 or 21; includes data from California Teachers Study, Cancer Prevention Study II, Health Professionals Follow-up Study, and Nurses Health Studies.

*HRs adjusted for age, sex, cohort, race, education, smoking, alcohol.

N represents the number of MM cases in each category.

There were some differences in associations by race, but for most anthropometric variables the P values for interaction by race were not statistically significant (Supplementary Table 3). The exception was for height, which was more strongly associated with MM in non-white race groups than in white individuals (P-interaction <0.01). Asian individuals had an increased risk of MM at lower BMI values than in the full population (Supplementary Table 4) and also had an elevated HR for usual adult BMI (HR per 5 kg/m²: 2.17; 95% CI: 1.13–4.17) compared to whites (corresponding HR: 1.09; 95% CI: 1.03–1.16) (Supplementary Table 3), but the test for interaction by race was not statistically significant (P = 0.15).

Discussion

In this large prospective study of men and women, we found that larger body size, measured at various times during life and in different ways, is positively associated with risk of MM. Our finding of a 2% increased risk of MM associated with each additional 5 kg/m² in BMI based on a simple update model is lower that reported in a recent meta-analysis of prospective cohort studies [5]. However, a stronger association of 10% increased risk per 5 kg/m² of BMI was apparent when we considered a time-varying cumulative average of BMI over follow-up as a measure of “usual adult” BMI, suggesting that cumulative exposure to adiposity may be more relevant for aetiology than recent BMI. Moreover, our findings suggest that both usual adult and young adult BMI (i.e., at ages 18–21) are independently associated with MM risk. For both usual and young adult BMI, higher risks of MM were most apparent for individuals with BMI ≥ 30 kg/m², with especially high risks among those with usual adult BMI ≥ 40 kg/m² and for individuals who were heavy in both young adulthood and later adulthood.

Obesity has been established as a risk factor for MM [5, 7–9, 11, 13, 21–24]. In a recent meta-analysis of cohort studies, compared to BMI < 25 kg/m², the relative risk (RR) for MM associated with BMI 25–29.9 kg/m² was 1.12 (95% CI: 1.06, 1.09), and the RR for MM associated with BMI > 30 kg/m² was 1.17 (95% CI: 1.01, 1.34) [23]. The majority of prior studies relied on BMI assessed at study enrolment and did not have repeated measures of BMI available. In our pooled analyses, we found that modelling usual adult BMI yielded stronger associations than modelling recent BMI, suggesting only a small influence of reverse causation in the latter HRs. The findings for baseline BMI reflect a widely varying pre-diagnosis time period across the participating cohorts and thus do not implicate a discrete latency period; of note, those findings were similar to the findings for usual adult BMI, e.g., the measure based on the time-varying cumulative average BMI. We further showed that high BMI at ages 18 or 21 was an independent predictor of MM. Young adult BMI was also associated with an increased risk of incident MM in a pooled case–control analysis that included 2318 cases and 9609 controls (odds ratio per 5 kg/m²: 1.2; 95% CI: 1.1, 1.3) [7]. Results from large prospective cohort studies support these findings [8, 9], suggesting that individuals with a high BMI in both earlier and later adulthood are at increased risk of MM. Similarly, in a previous analysis of body shape trajectories in the NHS and HPFS, individuals who maintained a lean and stable weight throughout life had the lowest risk [14].

We noted somewhat stronger associations of usual and young adult BMI in relation to MM risk for women compared to men, consistent with results from a recent meta-analysis which showed a higher RR associated with BMI > 30 vs. <25 kg/m² among women (1.27; 95% CI: 1.15, 1.40) than men (1.17; 95% CI: 1.01, 1.34) [23]. While obesity is known to influence levels of sex hormones [25], the role sex hormones might play in MM risk has been largely unstudied [26, 27]. In addition, we observed stronger associations of usual adult BMI among Asian individuals compared to other race groups, with an elevated risk at a lower BMI than in the overall population. In a previous prospective study in a Japanese population, there was no association between weight and risk of MM; however, only 88 cases were included in that analysis [28]. More recently, higher BMI was associated with increased MM mortality in the Asia Cohort Consortium comprising more than 800,000 individuals and 428 deaths [29]. Additional large prospective studies of obesity and MM risk in Asian populations are warranted.

While BMI is a useful metric for overall adiposity, it does not distinguish body composition of adipose vs. lean tissue or distribution of fat mass [30]. Other measures of body size, such as waist circumference, may therefore better reflect central or abdominal adiposity. Similar to our findings for BMI, we also observed positive associations of waist circumference and predicted fat mass with MM risk; however, these results were based on about 30% of the total study population with available measures of waist circumference, and associations were somewhat weaker than those observed for usual adult BMI. In a recent meta-analysis of prospective cohort studies, no association was found between waist circumference and risk of MM [5]; however, only three studies were included in that analysis [10–12]. A recent record-based study in Spain found a somewhat weaker association for waist circumference compared to BMI [31]; again, statistical power was much lower for the former measure. In contrast, a stronger association of MM risk was observed for technician-measured waist circumference vs. BMI in the Malmö Diet and Cancer study [13]. No prior studies of percent fat mass (directly measured or predicted) were identified. Additional studies with direct measures of body composition or more frequent measures of waist circumference are needed to clarify the importance of central adiposity and fat mass to the association of obesity with MM risk.

Tall stature has been observed to be positively associated with the risk of many different cancer types [5, 32, 33]. Our finding of increased risk of MM among taller individuals is consistent with prior reports [34]. Attained adult height is strongly heritable [35] and is also influenced in part by nutrition and other environmental factors, such as infections, early in life [36]; these factors could also influence cancer risk. Height is positively correlated with the number of cells in the body, and increased cancer risk could therefore be attributed, at least in part, to the increased chance of malignant transformation [37, 38]. Higher circulating levels of growth hormones [39, 40], particularly during childhood and adolescence, are also hypothesised to increase cancer risk. Stronger associations of height with MM were noted for non-white individuals vs. white individuals; in the same population, we also found stronger associations of height with other lymphoid malignancies among non-white individuals compared to white individuals [41].

The biological mechanisms by which obesity may influence MM aetiology are not fully established but may include pathways involving interleukin-6 (IL-6), insulin-like growth factor 1 (IGF-1) and adiponectin [42–46]. Overweight and obesity are associated with high circulating levels of growth factors, adipokines and lipids. Obesity is also linked to heightened inflammation, deregulation of endogenous hormonal pathways that can influence cell-cycle control, insulin resistance and immune dysfunction [30, 47]. A recent prospective cohort study found that IL-6 and IGF-1, both established plasma cell growth factors [48], were associated with MM risk [43]. Our observation that MM risk is elevated for adiposity in both young adulthood and usual adulthood further suggests a critical role of sustained systemic physiologic dysregulation over the life course.

As the prevalence of obesity continues to increase in the U.S. and worldwide, the public health burden of MM is also expected to increase. In fact, age-adjusted incidence rates of MM in the U.S. have increased from 4.9 per 100,000 in 1975 to 7.3 per 100,000 in 2017; rates have risen about 1.8% per year over the last 10 years (P < 0.01) [49]. While the factors contributing to the rise in MM incidence are unknown, it is plausible that increasing prevalence of obesity, especially among younger adults, could be contributing.

A potential limitation of this study is exposure measurement error, given that most of the body size data were self-reported. Self-reported and technician-measured weight were highly correlated (r = 0.97) [50], and recalled weight at age 18 was highly correlated with measured weight in medical records (r = 0.87) [51]. Similarly, self-reported waist and hip circumference have been validated against technician measurements. The generalisability of findings is another concern. Except for height, we did not observe any evidence of heterogeneity in associations by race/ethnicity; however, we were limited in statistical power to detect associations in racial/ethnic minority groups as our population was over 90% white. We adjusted models for suspected risk factors for MM, but we lacked data on the family history of MM; therefore, residual confounding by this or other unmeasured covariates is a possibility.

A major strength of this study is the availability of data on early life BMI and repeated assessment of BMI over follow-up, which allowed us to evaluate the influence of usual adult adiposity on MM risk. We carefully harmonised exposure and covariate information across six cohorts to achieve excellent statistical power in models adjusted for potential confounders.

The accumulating epidemiologic evidence, including our current findings of increased risk of MM associated with higher BMI in young and middle adulthood, highlights the importance of weight maintenance throughout adulthood as a potential MM risk-reduction strategy. The increasing prevalence of obesity and increasing incidence of MM in the U.S. suggests that the burden of MM attributable to obesity could increase. Further studies are warranted to evaluate the role of adiposity in childhood as well as the impact of intentional weight loss on the risk of MM.

Author contributions

All authors contributed data from their respective cohort studies. KAB, LRT and BMB developed the analytic plan with significant contributions from SSW and CRC. ELD conducted the majority of the statistical analyses with contributions from KW and consultation from BAR. KAB drafted the manuscript with input and revisions from LRT, SSW and BMB. All other authors (ELD, CRC, BAR, KW and CZ) provided critical revisions to the writing and analyses. All authors approved the final version of this manuscript.

Funding

This work was supported by the National Institutes of Health (R01 CA202712, UM1 CA186107, P01 CA87969, U01 CA176726, U01 CA167552, U01 CA199277; P30 CA033572; P30 CA023100; UM1 CA164917; R01 CA077398; R01 CA207020). The Nurses’ Health Study II received additional funding from the Breast Cancer Research Foundation. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health.

Data availability

Data will be made available upon reasonable request to the senior author.

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

This study was approved by the Institutional Review Boards of City of Hope, Brigham and Women’s Hospital, Harvard T.H. Chan School of Public Health, Kaiser Permanente Southern California, and those of participating cancer registries as required. Informed consent was implied by the return of the baseline questionnaire. The study was performed in accordance with the Declaration of Helsinki.

Consent to publish

Not applicable.

Supplementary information

The online version contains supplementary material available at 10.1038/s41416-022-01907-2.