Background:

Obesity increases the risk of 13 cancer types. Given the long process of carcinogenesis, it is important to determine the impact of patterns of body mass over time.

Methods:

Using data from 30,377 participants in the Melbourne Collaborative Cohort Study, we identified body mass index (BMI) trajectories across adulthood and examined their association with the risk of obesity-related cancer. Participants completed interviews and questionnaires at baseline (1990–1994, age 40–69 years), follow-up 1 (1995–1998), and follow-up 2 (2003–2005). Body mass was recalled for age 18 to 21 years, measured at baseline, self-reported at follow-up 1, and measured at follow-up 2. Height was measured at baseline. Cancer diagnoses were ascertained from the Victorian Cancer Registry and the Australian Cancer Database. A latent class trajectory model was used to identify BMI trajectories that were not defined a priori. Cox regression was used to estimate HRs and 95% confidence intervals (CI) of obesity-related cancer risks by BMI trajectory.

Results:

Six distinct BMI trajectories were identified. Compared with people who maintained lower normal BMI, higher risks of developing obesity-related cancer were observed for participants who transitioned from normal to overweight (HR, 1.29; 95% CI, 1.13–1.47), normal to class I obesity (HR, 1.50; 95% CI, 1.28–1.75), or from overweight to class II obesity (HR, 1.66; 95% CI, 1.32–2.08).

Conclusions:

Our findings suggest that maintaining a healthy BMI across the adult lifespan is important for cancer prevention.

Impact:

Categorization of BMI by trajectory allowed us to identify specific risk groups to target with public health interventions.

In 2016, an International Agency for Research on Cancer (IARC) expert working group concluded that body fatness [operationalized as overweight or obesity; body mass index (BMI) ≥25 kg/m2] increases the risk of 13 cancer types: esophagus, postmenopausal breast, liver, gallbladder, kidney, colorectal, multiple myeloma, meningioma, thyroid, gastric cardia, pancreatic, ovary, and corpus uteri (hereafter uterus; ref. 1). Between 1982 and 2017, the incidence of these cancers in Australia increased, especially for thyroid, liver, and kidney cancer. In women, increases have also been observed for postmenopausal breast cancer and uterine cancer, while men have also experienced increasing rates of esophageal cancer (2). In 2013, overweight and obesity were estimated to cause 4% of incident cancer cases and 5% of cancer deaths in Australia (3); these proportions are likely to increase if obesity continues to rise. Cross-sectional data collected between 1995 and 2015 in Australia indicate that, across most ages, more recent birth cohorts are experiencing higher prevalence of overweight and obesity than cohorts born 20 years earlier (4). Recent simulations of interventions have estimated that, over a 25-year period (2013–2037), 10% to 13% of obesity-related cancers in men and 7% to 11% in women could be avoided if all Australians had a BMI <25 kg/m2 (5).

Given that the carcinogenic processes usually take many years, it is important to determine the possible impact of patterns of body size across the lifepath on cancer risk (6). Adults currently included in cohort studies of body size and cancer risk may not have experienced overweight and obesity in childhood, making patterns of weight over adulthood of particular interest. Some studies have assessed changes in body mass using predefined categories of change (e.g., increasing BMI, stable BMI, decreasing BMI; ref. 7). This approach may miss more complex patterns of change in body mass (8).

We have previously used data from the Melbourne Collaborative Cohort Study (MCCS) to identify groups of people whose BMI followed specific trajectories across adulthood, and examined the associations of these trajectories with mortality (9). Here, we applied the same method to examine the association of BMI trajectories with the risk of obesity-related cancer.

Study sample

The MCCS is a prospective cohort study of 41,513 residents of the Melbourne metropolitan area, with a mean age of 55 years (99% ages between 40 and 69 years) at recruitment from 1990 to 1994. Details of the cohort have been presented elsewhere (10). Southern European migrants were oversampled to broaden the range of lifestyle exposures. Baseline data were collected by interviewer-administered questionnaires on demographics, lifestyle and medical conditions, physical measurements, and a self-administered food frequency questionnaire (11). Further collections occurred between 1995 and 1998 (88% responded by telephone or mail) and between 2003 and 2007 (68% attended a study center). The research protocol was approved by Cancer Council Victoria's Human Research Ethics Committee.

Participants selected for this analysis

For this analysis, we excluded participants if they were ages <40 or >70 years at baseline (n = 298), were underweight (BMI <18.5 kg/m2) at baseline or had extreme BMI values (BMI <15 or >45 kg/m2; n = 616), were diagnosed with invasive or metastatic cancer other than basal cell or squamous cell skin cancers (i.e., keratinocyte skin cancer) before their last BMI measurement (n = 5,647), had missing data for any confounding variables (n = 496), or had BMI measured or reported for fewer than three time points to allow estimation of potentially nonlinear trajectories (n = 4,079). These exclusions left 30,377 eligible participants. A flowchart of the participant selection is presented in Fig. 1.

BMI assessment

Each participant's body mass was recorded at up to four time points: at age 18 to 21 years, baseline (age 40–69 years), follow-up 1 (age 43–77 years), and follow-up 2 (age 48–85 years). At baseline, participants were asked to recall their body mass when ages 18 to 21 years. Height and body mass at baseline were measured by study staff following standard protocols (12). Body mass at follow-up 1 was self-reported. It was measured by study staff at follow-up 2 (12). To adjust for measurement error arising from having body mass self-reported at follow-up 1, a correction was applied to data based on sex- and age-specific comparison of self-reported and measured body mass in the Australian National Nutrition Survey (13, 14). Recalled body mass at age 18 to 21 years was not adjusted because the correction formula may not be applicable to it. For each time point, BMI was calculated as body mass (kg) divided by baseline height squared (m2).

Obesity-related cancer assessment

Incident cancers were ascertained from the Victorian Cancer Registry and the Australian Cancer Database and defined as the earliest diagnosis of an invasive or metastatic primary cancer (we excluded in situ and benign tumors). Obesity-related cancers were those defined in the recent evaluation carried out by the IARC Handbooks of Cancer Prevention program: esophagus adenocarcinoma, postmenopausal breast, liver, gallbladder, renal cell, colon and rectum, multiple myeloma, meningioma, thyroid, gastric cardia, pancreas, ovary, and corpus uteri (Supplementary Table S1; ref. 1).

Statistical analysis

BMI trajectories

The statistical methods, which have been described previously, followed the recommendations of Nagin and colleagues (8, 9). In brief, we used a latent class trajectory model to identify latent homogeneity in BMI trajectories from age 18 to 21 years to late adulthood. Longitudinal BMI data were fitted as a mixture of several latent trajectories in a censored normal model, with a polynomial function for age (15). Because other studies of BMI trajectory groups have used between four and six groups (16–21), we first compared up to six groups. Assuming all had quadratic patterns, we used the Bayesian information criterion (BIC) and the log Bayes factor as a guide to select the number of groups. We then compared linear and quadratic patterns of the chosen number of trajectories, also using the BIC and the log Bayes factor to select optimal shapes. Participants were assigned to the group that had the highest posterior predicted probability calculated from the final model. We evaluated the adequacy of the final model using recommended diagnostic measures: an average posterior probability of assignment for each group of 0.7 or higher; odds of correct classification of 5.0 or higher; the proportion of a sample assigned to a certain group close to the proportion estimated from the model; and a reasonably narrow confidence interval around each trajectory (22).

Risk of obesity-related cancer

Cox regression with age as the time scale was used to estimate HRs and 95% confidence intervals (CI) for the risk of any obesity-related cancer in relation to trajectory. Follow-up began at the last BMI assessment and ended at the first of: date of obesity-related cancer diagnosis; other cancer with invasive, metastatic, or uncertain behavior; date of death; or May 31, 2017, which was the date of the most recent complete cancer data linkage. Models included potential confounders (selected using a causal diagram presented in Supplementary Fig. S1): year of birth (<1930, 1930–39, 1940–49, ≥1950); height (cm); sex (male, female); country of birth (Australia/New Zealand/Northern Europe, Southern Europe); an area-based measure of socioeconomic position (SEIFA: socio-economic indexes for areas) from the Australian Census (quintiles); highest education attained (primary school or lower, some high or technical school, completed high or technical school, completed tertiary education); smoking status at last BMI assessment (never, former, current); alcohol drinking status at baseline [lifetime abstainer, former, current low (1–39 g/day for male; 1–19 g/day for female), current high (≥40 g/day for male; ≥20 g/day for female)]; physical activity score at baseline (none, low, moderate, high; computed from frequency of walking, vigorous and less vigorous exercise over the last 6 months; ref. 23); and Mediterranean diet score (0–3, 4–6, 7–9) details of which are described elsewhere (24). Because smoking can be a strong confounder, we performed a further analysis restricted to never smokers.

Tests based on Schoenfeld residuals were used to examine the proportional hazards assumption and the Cox models were stratified on covariates that did not satisfy the proportional hazards assumption.

We also examined associations between trajectories of BMI and each of colorectal and postmenopausal breast cancer as these were the two types for which enough cases had accrued to be examined separately. In addition, we undertook sensitivity analyses for all models: (i) we excluded the first 2 years of follow-up to account for potential reverse causation; (ii) we excluded participants without data from follow-up 2 (n = 8,008), so that the final data points contributing to the trajectories were not extrapolated and the beginning of follow-up were the same for all participants; (iii) we compared our findings with the HRs obtained from a one time point assessment of BMI (last measured BMI); and (iv) finally, we modeled sex-specific trajectories and examined their associations with risk of obesity-related cancer for men and women separately, as suggested by a reviewer.

Statistical analyses were performed using Stata version 14.2 (Stata Corp.).

Six trajectories of BMI across adulthood were identified (Fig. 2). Almost all participants (94%) had normal BMI at age 18 to 21 years (TR1, TR2, TR3, and TR5). For two thirds, BMI increased later in life to higher normal (TR2) or overweight (TR3). A smaller proportion (19%) maintained a lower normal BMI (TR1) throughout adulthood. Thirteen percent of participants became obese in midlife (TR5 and TR6) and 3% were consistently borderline obese over the adult years considered (TR4). Model adequacy diagnostics are presented in Supplementary Table S2.

The characteristics of participants within each trajectory were essentially the same as reported in our analysis of mortality (9) and are presented in Table 1. Mediterranean diet score, physical activity, and alcohol consumption decreased across TR1 to TR6. A higher proportion of participants in the borderline obesity trajectory (TR4) compared with other trajectories had prevalent diabetes and heart disease at baseline.

Table 2 presents the HRs for BMI trajectories in relation to risk of obesity-related cancer in all participants and in never-smokers at last BMI assessment, using stable lower normal BMI trajectory (TR1) as the reference category. We observed no meaningfully increased risk for the stable higher-normal BMI trajectory (TR2: HR, 1.04; 95% CI, 0.92–1.18). Increased risk was observed for all other trajectories, although the HR for chronic borderline obesity (TR4) had a wide CI. As the higher normal trajectory (TR2) showed evidence of nonproportional hazards for never-smokers (P = 0.02), we fitted an interaction between this trajectory and attained age and presented HRs for age 67, 74, and 80 (25th, 50th, and 75th percentiles of age at diagnosis). For this trajectory, the HR was slightly lower at younger attained age than at older attained age. In all participants, while the proportional hazards assumption was not violated, the same pattern (i.e., larger HRs with increasing age) was seen for TR2. The HRs for BMI trajectories were similar for never-smokers and the whole cohort.

For colorectal cancer, no strong associations were observed for any trajectory except overweight to class II obesity (TR6), for which the HR was 1.47 (95% CI, 0.98–2.20); however, only 30 people in this trajectory were diagnosed with colorectal cancer (Table 3). For postmenopausal breast cancer, the HRs for each trajectory followed a similar pattern to that observed for all obesity-related cancers. The largest HRs were observed for women whose BMI changed from normal to overweight (TR3: HR, 1.31; 95% CI, 1.05–1.62) and from normal BMI to class I obesity (TR5: HR, 1.55; 95% CI, 1.20–2.00; Table 3).

The results from sensitivity analyses (excluding the first 2 years of follow-up; and excluding participants who did not attend follow-up 2) were not materially different from the main results (see Supplementary Tables S3 and S4). Models including BMI trajectories were almost identical to those with last measured BMI in terms of the BIC and Akaike information criterion (Supplementary Table S5). The HRs estimated for the last measured BMI were, in general, somewhat lower than the HRs derived for trajectories ending at the equivalent category of BMI. For example, for all obesity-related cancers, TR3 normal to overweight HR, 1.29 (95% CI, 1.13–1.47); last measured BMI overweight HR, 1.14 (95% CI, 1.00–1.30). See Supplementary Table S6 for further details.

In the sex-specific analyses, we identified similar numbers and shapes of BMI trajectories in women and men (Supplementary Fig. S2). The risks of obesity-related cancer in men and in women were similar to the risk in the main analysis. For the “Overweight to class II obesity” trajectory, the HR was higher for men than for women. However, the HR was based on a small number of cases and had a wide CI (Supplementary Table S7).

In all trajectories modeled in our study, BMI increased over adulthood and tended to decrease at older ages, but only trajectories that moved beyond a normal BMI were associated with an increased risk of obesity-related cancer, compared with a lifelong healthy BMI. BMI increasing from normal to overweight was associated with a 29% increased risk, whereas BMI increasing from normal to class I obesity was associated with a 50% increased risk. The trajectory with the greatest risk increase (66%) was overweight to class II obesity.

A key strength of this study is that we could derive trajectories of BMI across adulthood because body mass had been measured or reported at multiple times. Furthermore, we avoided bias by not grouping participants into categories of change between two time points (8). Our study had several limitations that are important to consider, including assessing weight at age 18 to 21 years based on recall at baseline (20–50 years later), which could be inaccurate. However, previous studies have typically reported moderate to strong correlation between body mass measured in young adulthood and recalled decades later (25, 26). Another limitation is the long period of time between age 18 to 21 years and baseline, which means that there may have been different patterns of BMI change over early and mid-adulthood that we could not identify. Information about potential confounders at age 18 to 21 years was missing. Creating trajectories using the latent class trajectory model required excluding participants who attended the baseline visit only (i.e., excluding those with no subsequent data due to nonattendance or death). In one study, this type of selection bias resulted in underestimation of rate ratios for higher BMI categories in the case of all-cause mortality (27). Some trajectories in the analyses of colorectal cancer and breast cancer had few cases, and the estimated HRs were imprecise as indicated by wide CIs.

Only one study has examined the association of body shape trajectories with total cancer and obesity-related cancer. Song and colleagues used data from the Nurses' Health Study and the Health Professionals Follow-up Study to model trajectories using pictograms for early life and BMI for later life (28). Five trajectories were identified: for women, lean-moderate increase, lean-marked increase, and heavy-stable/increase trajectories were associated with a higher risk of total and obesity-related cancers compared with the lean-stable pattern. For men, all trajectories were associated with an increased risk of total and obesity-related cancer compared with the lean-stable group.

Studies of BMI trajectories (or body shape using diagrams) across a range of ages have examined associations with risk of specific cancers, including colorectal cancer and postmenopausal breast cancer. The Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial estimated BMI trajectories using data from ages 20, 50, and study entry at a mean age of 63 years. Four trajectories were identified (BMI remaining normal; normal to overweight; normal to obese; and overweight to obese). Compared with participants having stable normal BMI, those in the normal with obese trajectory had increased colorectal cancer risk (16). The French E3N study used self-reported body shape at age 8, at menarche, at 20 to 25, and at 35 to 40 years to define six body size trajectories. Women with a large body shape at menarche, irrespective of their shape before or after, tended to have reduced risk of estrogen receptor–positive or progesterone receptor–positive postmenopausal breast cancer, compared with women who were consistently lean (17). This is consistent with findings from the Nurses' Health Study, in which adiposity at age 10 was inversely associated with risk of postmenopausal breast cancer (29).

Although the IARC review considered evidence for an association between overweight or obesity and risk of prostate cancer to be limited (1), analysis of the NIH-AARP Diet and Health Study reported that a trajectory increasing from normal body mass at age 18 years to obesity at study entry (mean age 63 years) was associated with a higher risk of fatal prostate cancer compared with stable normal weight (18). Combining data from the PLCO and NIH-AARP studies, Petrick and colleagues examined trajectories based on BMI at ages 20, 50, and study baseline (around age 63 years) in relation to esophageal adenocarcinoma and gastric cardia adenocarcinoma. Compared with a stable normal BMI, trajectories that reflected BMI increases from normal to overweight or obese, or from overweight to obese, were associated with an increased risk of both cancers, the risk being over three times higher for the latter trajectory (19). In a Canadian case–control study of pancreatic cancer, five BMI trajectories based on BMI in adolescence, 20s, 30s to 40s, 50s to 60s were identified; persistent overweight and progressive obesity were associated with an increased risk relative to stable normal weight (20). Data from the Nurses' Health Study and the Health Professionals Follow-up Study were used to determine body shape trajectories based on somatotype pictograms for age 5, 10, 20, 30, and 40 years, and self-reported height and weight to calculate BMI at age 45, 50, 55, and 60 years in relation to multiple myeloma risk. Four trajectories were identified, and the pattern that started out at a medium body size and increased, compared with remaining lean, was associated with an increased risk of multiple myeloma (21).

In the MCCS, trajectories gave slightly higher HRs than single BMI measures for associations with obesity-related cancers, although model fits were not notably better with the additional information provided by the trajectories. This comparison was based on BMI at the final measure so that the follow-up was the same for the trajectory and single time point analyses. Had the comparison been made on a single BMI measure at another time point, the results may have been different. Zheng and colleagues showed that the effect sizes of BMI at a single time point were smaller than the corresponding estimates using BMI trajectories in relation to mortality (30). Kelly and colleagues have provided data for BMI at ages 20, 50, and baseline (mean age 63 years) and prostate cancer, as well as associations for BMI trajectories combining these data points; at these ages the HR for prostate cancer mortality were 1.69 (95% CI, 0.75–3.82), 1.55 (1.05–2.29), and 1.46 (1.02–2.09), respectively, while the HRs for trajectories normal to obese, or overweight to obese were 1.95 (1.21–3.12) and 2.65 (1.35–5.18), respectively (31). These figures like our data show somewhat higher HRs for trajectories than single time points, and some evidence that obesity at average age 63 years was differently associated with prostate cancer mortality depending on BMI at age 20 years. We did not have any trajectories finishing at the same BMI but starting at very different BMIs. Participants in the MCCS mostly reported low BMI at age 18 to 21 years which is consistent with other Australian data showing that children ages 5 to 15 years had prevalence of overweight and obesity less than 10% up to the 1970s when the prevalence increased rapidly (32), and birth cohort data showing that the prevalence of overweight and obesity in people ages 18 to 21 years has increased for cohorts born between 1974 and 1977 and 1994 and 1997 (4). However, as people become overweight and obese at earlier ages, the potential extra information from using trajectories may become more important.

Overall, the research to date suggests that body mass trajectories that reflect increasing adiposity from young adulthood are associated with an increased risk of several obesity-related cancers in later life. Our study adds to this body of evidence and highlights the importance of maintaining a normal BMI into older ages to reduce the risk of cancer. To reduce the cancer burden associated with obesity, primary prevention targeting young adults to prevent weight gain over adulthood is vital.

B.M. Lynch reports grants from Victorian Cancer Agency during the conduct of the study. J.K. Bassett reports grants from Australian National Health and Medical Research Council during the conduct of the study. G.G. Giles reports grants from Australian National Health and Medical Research Council during the conduct of the study. R.L. Milne reports grants from Australian National Health and Medical Research Council during the conduct of the study. A.M. Hodge reports grants from Australian National Health and Medical Research Council during the conduct of the study. No disclosures were reported by the other authors.

Y. Yang: Formal analysis, methodology. B.M. Lynch: Conceptualization, supervision, writing–original draft, writing–review and editing. P.-A. Dugué: Supervision, writing–review and editing. A. Karahalios: Writing–review and editing. R.J. MacInnis: Writing–review and editing. J.K. Bassett: Writing–review and editing. A. McAleese: Conceptualization. C. Sinclair: Conceptualization. G.G. Giles: Conceptualization, resources. R.L. Milne: Conceptualization, resources, writing–review and editing. A.M. Hodge: Conceptualization, supervision, writing–original draft, writing–review and editing. D.R. English: Conceptualization, supervision, writing–review and editing.

The authors thank all participants in the MCCS for their valuable contributions to the study. Y. Yang is supported by a Melbourne Research Scholarship from the University of Melbourne. B.M. Lynch is supported by a fellowship from the Victorian Cancer Agency (MCRF18005). Cohort recruitment of the MCCS was funded by Cancer Council Victoria (http://www.cancervic.org.au/) and VicHealth (http://www.vichealth.vic.gov.au/). The MCCS was further supported by grants 209057, 251553, 396414, and 1074383 from the Australian National Health and Medical Research Council (NHMRC; http://www.nhmrc.gov.au/), and ongoing follow-up and data management has been funded by Cancer Council Victoria since 1995.

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

1.
Lauby-Secretan
B
,
Scoccianti
C
,
Loomis
D
,
Grosse
Y
,
Bianchini
F
,
Straif
K
;
International Agency for Research on Cancer Handbook Working Group
. 
Body fatness and cancer–viewpoint of the IARC working group
.
N Engl J Med
2016
;
375
:
794
8
.
2.
Australian Institute of Health and Welfare
.
Cancer in Australia 2017
.
Canberra (Australia)
:
AIHW
; 
2017
.
Cancer series no. 101. Cat. no.: CAN 100. Available from
: https://www.aihw.gov.au/reports/cancer/cancer-in-australia-2017/contents/table-of-contents.
3.
Wilson
LF
,
Antonsson
A
,
Green
AC
,
Jordan
SJ
,
Kendall
BJ
,
Nagle
CM
, et al
How many cancer cases and deaths are potentially preventable? Estimates for Australia in 2013
.
Int J Cancer
2018
;
142
:
691
701
.
4.
Australian Institute of Health and Welfare
.
Overweight and obesity: an interactive insight
.
Canberra (Australia)
:
AIHW
; 
2019
.
Cat. no.: PHE 251. Available from
: https://www.aihw.gov.au/reports/overweight-obesity/overweight-and-obesity-an-interactive-insight/contents/what-is-overweight-and-obesity.
5.
Wilson
LF
,
Baade
PD
,
Green
AC
,
Jordan
SJ
,
Kendall
BJ
,
Neale
RE
, et al
The impact of changing the prevalence of overweight/obesity and physical inactivity in Australia: an estimate of the proportion of potentially avoidable cancers 2013–2037
.
Int J Cancer
2019
;
144
:
2088
98
.
6.
Ferreira
I
,
Stehouwer
CD
. 
Obesity paradox or inappropriate study designs? Time for life-course epidemiology
.
J Hypertens
2012
;
30
:
2271
5
.
7.
Karahalios
A
,
English
DR
,
Simpson
JA
. 
Change in body size and mortality: a systematic review and meta-analysis
.
Int J Epidemiol
2017
;
46
:
526
46
.
8.
Nagin
D
.
Group-based modeling of development
.
Cambridge (MA)
:
Harvard University Press
; 
2005
.
9.
Yang
Y
,
Dugue
PA
,
Lynch
BM
,
Hodge
AM
,
Karahalios
A
,
MacInnis
RJ
, et al
Trajectories of body mass index in adulthood and all-cause and cause-specific mortality in the Melbourne Collaborative Cohort Study
.
BMJ Open
2019
;
9
:
e030078
.
10.
Milne
RL
,
Fletcher
AS
,
MacInnis
RJ
,
Hodge
AM
,
Hopkins
AH
,
Bassett
JK
, et al
Cohort profile: the Melbourne Collaborative Cohort Study (Health 2020)
.
Int J Epidemiol
2017
;
46
:
1757
1757i
.
11.
Ireland
P
,
Jolley
D
,
Giles
G
,
O'Dea
K
,
Powles
J
,
Rutishauser
I
, et al
Development of the Melbourne FFQ: a food frequency questionnaire for use in an Australian prospective study involving an ethnically diverse cohort
.
Asia Pac J Clin Nutr
1994
;
3
:
19
31
.
12.
Lohman
TG
,
Roche
AF
,
Martorell
R
, editors.
Anthropometric standardization reference manual
.
Champaign (IL)
:
Human Kinetics Books
; 
1988
.
13.
Ball
K
,
Crawford
D
,
Ireland
P
,
Hodge
A
. 
Patterns and demographic predictors of 5-year weight change in a multi-ethnic cohort of men and women in Australia
.
Public Health Nutr
2003
;
6
:
269
81
.
14.
McLennan
W
,
Podger
AS
.
National Nutrition Survey users' guide, 1995
.
Canberra (Australia)
:
Australian Bureau of Statistics, Department of Health and Family Services
; 
1998
.
Cat. No.: 4801.0. Available from
: https://www.abs.gov.au/AUSSTATS/abs@.nsf/Lookup/4801.0Main+Features11995?OpenDocument.
15.
Jones
BL
,
Nagin
DS
. 
A note on a Stata plugin for estimating group-based trajectory models
.
Sociol Methods Res
2013
;
42
:
608
13
.
16.
Zheng
R
,
Du
M
,
Zhang
B
,
Xin
J
,
Chu
H
,
Ni
M
, et al
Body mass index (BMI) trajectories and risk of colorectal cancer in the PLCO cohort
.
Br J Cancer
2018
;
119
:
130
2
.
17.
Fagherazzi
G
,
Guillas
G
,
Boutron-Ruault
MC
,
Clavel-Chapelon
F
,
Mesrine
S
. 
Body shape throughout life and the risk for breast cancer at adulthood in the French E3N cohort
.
Eur J Cancer Prev
2013
;
22
:
29
37
.
18.
Kelly
SP
,
Lennon
H
,
Sperrin
M
,
Matthews
C
,
Freedman
ND
,
Albanes
D
, et al
Body mass index trajectories across adulthood and smoking in relation to prostate cancer risks: the NIH-AARP Diet and Health Study
.
Int J Epidemiol
2019
;
48
:
464
73
.
19.
Petrick
JL
,
Kelly
SP
,
Liao
LM
,
Freedman
ND
,
Graubard
BI
,
Cook
MB
. 
Body weight trajectories and risk of oesophageal and gastric cardia adenocarcinomas: a pooled analysis of NIH-AARP and PLCO studies
.
Br J Cancer
2017
;
116
:
951
9
.
20.
De Rubeis
V
,
Cotterchio
M
,
Smith
BT
,
Griffith
LE
,
Borgida
A
,
Gallinger
S
, et al
Trajectories of body mass index, from adolescence to older adulthood, and pancreatic cancer risk; a population-based case-control study in Ontario, Canada
.
Cancer Causes Control
2019
;
30
:
955
66
.
21.
Marinac
CR
,
Suppan
CA
,
Giovannucci
E
,
Song
M
,
Kværner
AS
,
Townsend
MK
, et al
Elucidating under-studied aspects of the link between obesity and multiple myeloma: weight pattern, body shape trajectory, and body fat distribution
.
JNCI Cancer Spectrum
2019
;
3
:
pkz044
.
22.
Nagin
DS
,
Odgers
CL
. 
Group-based trajectory modeling in clinical research
.
Ann Rev Clin Psychol
2010
;
6
:
109
38
.
23.
Jayasekara
H
,
English
DR
,
Haydon
A
,
Hodge
AM
,
Lynch
BM
,
Rosty
C
, et al
Associations of alcohol intake, smoking, physical activity and obesity with survival following colorectal cancer diagnosis by stage, anatomic site and tumor molecular subtype
.
Int J Cancer
2018
;
142
:
238
50
.
24.
Hodge
A
,
English
D
,
Itsiopoulos
C
,
O'Dea
K
,
Giles
G
. 
Does a Mediterranean diet reduce the mortality risk associated with diabetes: evidence from the Melbourne Collaborative Cohort Study
.
Nutr Metab Cardiovasc Dis
2011
;
21
:
733
9
.
25.
Casey
VA
,
Dwyer
JT
,
Berkey
CS
,
Coleman
KA
,
Gardner
J
,
Valadian
I
. 
Long-term memory of body weight and past weight satisfaction: a longitudinal follow-up study
.
Am J Clin Nutr
1991
;
53
:
1493
8
.
26.
Must
A
,
Willett
WC
,
Dietz
WH
. 
Remote recall of childhood height, weight, and body build by elderly subjects
.
Am J Epidemiol
1993
;
138
:
56
64
.
27.
Banack
HR
,
Bea
JW
,
Kaufman
JS
,
Stokes
A
,
Kroenke
CH
,
Stefanick
ML
, et al
The effects of reverse causality and selective attrition on the relationship between body mass index and mortality in postmenopausal women
.
Am J Epidemiol
2019
;
188
:
1838
48
.
28.
Song
M
,
Willett
WC
,
Hu
FB
,
Spiegelman
D
,
Must
A
,
Wu
K
, et al
Trajectory of body shape across the lifespan and cancer risk
.
Int J Cancer
2016
;
138
:
2383
95
.
29.
Rosner
B
,
Eliassen
AH
,
Toriola
AT
,
Chen
WY
,
Hankinson
SE
,
Willett
WC
, et al
Weight and weight changes in early adulthood and later breast cancer risk
.
Int J Cancer
2017
;
140
:
2003
14
.
30.
Zheng
H
,
Tumin
D
,
Qian
Z
. 
Obesity and mortality risk: new findings from body mass index trajectories
.
Am J Epidemiol
2013
;
178
:
1591
9
.
31.
Kelly
SP
,
Graubard
BI
,
Andreotti
G
,
Younes
N
,
Cleary
SD
,
Cook
MB
. 
Prediagnostic body mass index trajectories in relation to prostate cancer incidence and mortality in the PLCO Cancer Screening Trial
.
J Natl Cancer Inst
2017
;
109
:
djw225
.
32.
Sabin
MA
,
Kao
KT
,
Juonala
M
,
Baur
LA
,
Wake
M
. 
Viewpoint article: childhood obesity–looking back over 50 years to begin to look forward
.
J Paediatr Child Health
2015
;
51
:
82
6
.