Adherence to 2018 WCRF/AICR Cancer Prevention Recommendations and Risk of Cancer: The Melbourne Collaborative Cohort Study

Cancer prevention is becoming increasingly challenging due to the systematic differences that exist in cancer occurrence between and within different regions of the world (1). Globally, just under 20 million new cancers and almost 10 million cancer deaths were estimated to have occurred in 2020 (2). The worldwide cancer burden is projected to rise further, to approximately 28 million new cases in 2040 due to population growth, aging, and increasing prevalence of cancer risk factors (2). According to the World Cancer Research Fund (WCRF) and the American Institute for Cancer Research (AICR), a large proportion of cancer cases, and hence deaths (∼51% of deaths in men and 36% in women), are likely preventable through modulation of key lifestyle and behavioral factors including smoking, alcohol use, and obesity (3). The vast majority of the studies underpinning these findings were conducted in high-income countries (3).

In 2007, the WCRF and the AICR proposed evidence-based recommendations on body size, physical activity, alcohol intake, and diet aimed at preventing cancer, globally (4). In the ensuing decade, the body of epidemiologic evidence on cancer risk factors expanded, and according to the latest umbrella review of the WCRF/AICR meta-analyses, causal links with cancer risk potentially exist for even more diet and lifestyle-related factors than currently identified (5). The focus also shifted to studying the effects of patterns of diet and behavior and their interactions leading to the acquisition of the genetic and epigenetic alterations underlying carcinogenesis (6). Consequently, the WCRF/AICR reformulated their cancer prevention recommendations in 2018, but the emphasis on core elements such as healthy weight, physical activity, alcohol intake, and dietary factors remained consistent (6).

The 2018 WCRF/AICR Expert Report estimated a 30% to 50% reduction in cancer risk and mortality through adherence to cancer prevention recommendations that aim to limit individual high-risk diet and lifestyle-related behaviors (6). The quantification of these guidelines (e.g., with high/low-risk cutoffs) and their operationalization in different populations were recognized by WCRF/AICR experts as key to translating existing evidence from observational studies to effective public health strategies. Several cohort studies had examined cancer risk in relation to the 2007 WCRF/AICR cancer prevention recommendations, but the recommendations had not been uniformly operationalized making comparisons across studies challenging (7). In 2019, researchers from the U.S. NCI, AICR, and WCRF International, jointly with WCRF/AICR Continuous Update Project (CUP) Expert Panel and other researchers, developed a standard scoring system to assess cancer risk in observational studies according to the degree of adherence to the 2018 WCRF/AICR cancer prevention recommendations (8, 9).

The proposed 2018 WCRF/AICR scoring system combined eight of the 10 2018 WCRF/AICR cancer prevention recommendations (10). Subsequent studies have proposed variations to the weights assigned to the individual recommendations to give prominence to some of the components deemed as more important lifestyle-related cancer risk factors (e.g., obesity over individual dietary factors; refs. 11, 12). Studies have rarely incorporated the additional 2018 subrecommendations such as avoiding weight gain over time and limiting sedentary behavior, due to a lack of consensus on how to operationalize these subcomponents (12). The 2018 guidelines do not include abstinence from smoking as a specific recommendation but emphasizes that as a general principle to prevent cancer (6) because WCRF's focus is on diet, body size, and physical activity (8). Studies that have evaluated the role of smoking as a subcomponent are scarce (13).

In the current study, we examined how concordance with different adaptations of the 2018 WCRF/AICR cancer prevention recommendations relates to risk of total, alcohol-related, obesity-related, smoking-related, digestive tract, and site-specific cancer in a prospective study.

The Melbourne Collaborative Cohort Study (MCCS) includes 41,513 participants (58.9% women; 99.2% ages 40–69 years) recruited during 1990 to 1994 in Melbourne, Australia. Details of the cohort have been published previously (14). The study protocol was approved by the Cancer Council Victoria Human Research Ethics Committee. Participants gave written informed consent to participate and for investigators to obtain access to their medical records. For this analysis, we used data for 26,158 MCCS participants for whom detailed information on physical activity, as well as body weight, diet (including alcohol) and smoking habits, was ascertained at the second follow-up assessment in 2003 to 2007 (Supplementary Fig. S1). We excluded participants with a confirmed cancer diagnosis before second follow-up assessment (n = 2,890), and those who reported extreme values of total energy intake (<1st percentile or >99th percentile, n = 437), for whom the WCRF/AICR scores could not be derived because of missing data on diet, body size, or physical activity (n = 2,580) or had missing data on the covariates modeled (n = 250), leaving 20,001 participants [48.2% of the original cohort; 12,264 (61.3%) women] for analysis (Supplementary Fig. S2).

At study entry in 1990 to 1994, structured questionnaires were used to obtain information on age, sex, country of birth, education, previous medical conditions, and lifestyle behaviors (14). A 121-item food frequency questionnaire (FFQ) was used to collect dietary information (15). Height was measured to 1 mm with a stadiometer and weight measured to 100 g. At the follow-up assessment in 2003 to 2007, structured interview schedules were used to obtain information on diet, alcohol intake, physical activity, and smoking (14). Participants completed a 144-item validated self- or interviewer-administered semiquantitative FFQ including food and beverage groups (grain-based foods, dairy foods and fats, meat, fish and seafood, fruit, vegetables, sugar-sweetened beverages and miscellaneous), with photos to help estimate portion size (16). Alcohol intake during the previous year was assessed for non-abstainers as frequency and quantity of intake per drinking occasion, for beer, wine, and spirits. Detailed information on the method of calculating alcohol intake in grams per day in the MCCS has been reported previously (17). Self-reported physical activity was collected by trained interviewers using a slightly modified version of the International Physical Activity Questionnaire long form (IPAQ-long; available at www.ipaq.ki.se; ref. 18). The total amount of physical activity performed for transport and leisure was calculated as per IPAQ-short guidelines and expressed in continuous form as weighted minutes per week (19). The total time spent watching television on week and weekend days was used to calculate the average number of hours of sedentary behavior per day. More details are available elsewhere (20, 21). Participants who were smokers at study entry were asked at the second follow-up if they were still smoking, and, if so, the amount smoked. Weight was measured at second follow-up assessment to 100 g using digital electronic scales. Body mass index (BMI) was calculated as weight in kilograms divided by the square of height (measured at baseline) in meters (kg/m²). Waist circumference was measured to 1 mm using a 2-m metal anthropometric tape (22).

We derived two scores characterizing adherence to 2018 WCRF/AICR cancer prevention recommendations (diet score and standardized lifestyle score; refs. 6, 8, 9) using information collected at follow-up assessment in 2003 to 2007 (baseline height was used). The WCRF/AICR diet score included five components: (i) whole grains, vegetables, fruit, and beans; (ii) “fast foods” and other processed foods high in fat, starches, or sugars; (iii) red meat and processed meat; (iv) sugar-sweetened beverages; and (v) alcohol. Each component was assigned a score of 0 (nonadherence), 0.5 (partial adherence), or 1 (adherence). The whole grains, vegetables, fruit, and beans component included two subcomponents each contributing a maximum of 0.5 points. The diet score was calculated by adding the five dietary component scores and ranged from 0 to 5 (Table 1). The WCRF/AICR standardized lifestyle score included the components on diet, healthy weight, and physical activity. This score follows the standard scoring system developed by the NCI, WCRF, and AICR expert panel (8, 9). Healthy weight and physical activity were also assigned scores of 0 (nonadherence), 0.5 (partial adherence), or 1 (adherence). The standardized lifestyle score was calculated by adding the seven component scores (range 0 to 7; Table 1). The healthy weight component included two subcomponents related to BMI and waist circumference each contributing a maximum of 0.5 points. Two additional versions of the lifestyle score were operationalized: an exploratory lifestyle score that included weight change between baseline and second follow-up data collections (as a third subcomponent to the healthy weight component) and sedentary behavior (as a second subcomponent with equal weight to physical activity; expanded lifestyle score; range 0 to 7) and another with smoking at second follow-up data collection (standardized lifestyle score + smoking; range 0 to 8), given its importance as a modifiable lifestyle risk factor for cancer (Table 1). See Supplementary Table S1 for a more detailed description.

Incident diagnoses of cancer and deaths are identified by at least annual record linkage to the Victorian Cancer Registry and the Victorian Registry of Births, Deaths, and Marriages, both of which are considered complete. Notification to the Victorian Cancer Registry of all cancers diagnosed in Victoria is mandated by legislation since 1981. These linkages are complemented by at least yearly record linkage to the National Death Index and the Australian Cancer Database to capture cancers and deaths for participants who may have moved interstate. Incident primary invasive cancers were coded following the 3rd Revision of the International Classification of Diseases for Oncology (ICD-O-3). The primary outcomes included total cancer (all types of invasive cancer), and alcohol-related cancer (23), obesity-related cancer (24), smoking-related cancer (25), and digestive tract cancer (as defined below). The 10 most common cancers in MCCS were assessed individually as secondary outcomes.

Alcohol-related cancers (23): oral cavity, pharynx, larynx and esophagus (squamous cell carcinoma) [collectively referred to as upper aerodigestive tract (UADT) cancer] (C01-C06, C09-C10, C13-C15, C32), breast (C50), liver (C22), and colorectum (C18.0, C18.2-18.9, C19.9, C20.9).

Obesity-related cancers (24): esophageal adenocarcinoma (C15), postmenopausal breast (C50), liver (C22), gallbladder (C23), kidney (C64), colorectum (C18.0, C18.2-18.9, C19.9, C20.9), multiple myeloma (C42), meningioma (C70), thyroid (C73), gastric cardia (C16.0), pancreas (C25), ovary (C56), and endometrium (C54.1, C55).

Smoking-related cancers (25): oral cavity and pharynx (C01-C06, C09-C14), nasal cavity (C30) and sinuses (C31), larynx (C32), esophagus (both adenocarcinoma and squamous cell carcinoma) (C15), lung (C34), stomach (C16), pancreas (C25), colorectum (C18.0, C18.2-18.9, C19.9, C20.9), liver (C22), kidney (C64; both body and pelvis), ureter (C66), bladder (C67), and acute myeloid leukemia (ICD-O morphology codes 9805, 9840, 9861, 9865-9869, 9871-9874, 9891, 9895-9897, 9910-9911, 9920, 9930-9931).

Digestive tract cancers: mouth (C01-06), oropharynx (C10), esophagus (C15), stomach (C16), small intestine (C17), colorectum (C18.0, C18.2-18.9, C19.9, C20.9), liver (C22), gallbladder and biliary tract (C23, C24), and pancreas (C25 excluding C25.4).

We used Cox regression (26) with age as the time scale to estimate HRs and 95% confidence intervals (CI) for associations between diet and lifestyle scores and risk of total, alcohol-related, obesity-related, smoking-related, digestive tract, and site-specific cancer. Follow-up started at second follow-up assessment and ended at the earliest of date of diagnosis of invasive cancer, death or end of follow-up (June 30, 2021). The diet and lifestyle scores were modeled using categories (five approximate groups/quintiles) or continuously (per unit increment in score). To estimate HRs for the 10 most common cancers, Cox regression models were fit using a competing risks method (27). We fit interaction terms to test for differences in associations by sex, smoking status, and attained age (by splitting the data at the median age at diagnosis = 75 years). All models were adjusted for sex, educational attainment (primary school, technical school, secondary school, university), any family history of cancer (yes, no), smoking status (never, former, current; except with lifestyle score which included smoking), and total energy intake from food and stratified by birth cohort (year of birth <1935, 1935–1944, ≥1945) and country of birth (Australia/New Zealand/United Kingdom or Italy/Greece). HRs for prostate, breast, and colorectal cancers were adjusted for prostate-specific antigen testing (yes, no), history of mammography (yes, no), and history of colonoscopy or sigmoidoscopy (yes, no), respectively.

Sensitivity analyses were performed excluding the first 2 years of follow-up to account for potential reverse causation and adjusting for more detailed smoking habits (smoking intensity: never, former ≥20 years since quitting, former <20 years since quitting, current <20 cigarettes/day, current ≥20 cigarettes/day) to limit residual confounding by smoking where relevant. A secondary analysis assessed associations between each individual component of the diet and lifestyle scores and total cancer risk in a model mutually adjusted for the other components. The shape of the associations between the scores and total cancer risk was examined by comparing models that included the score as a linear term and with restricted cubic splines (with three knots at the 10th, 50th, and 90th percentile; ref. 28). Each model was examined for outliers and influential points (29). Nested models were compared using the likelihood ratio test (30). Tests based on Schoenfeld residuals and graphical methods were used to examine the proportional hazards assumption (31). Cox models were stratified on sex (for alcohol-related and smoking-related cancers) and family history of cancer (for digestive tract cancer except with expanded lifestyle score which included weight change and sedentary behavior) as there was evidence of nonproportional hazards for these variables. All statistical tests were two sided, and statistical analyses were performed using Stata 16.1 (StataCorp).

Statistical code is available from the corresponding author. The MCCS data can be made available on request to [email protected].

The median scores were 2.75 (interquartile range = 2.25–3.25) and 4 (3.25–4.75) for the WCRF/AICR diet and standardized lifestyle scores, respectively (Supplementary Fig. S3). The Spearman correlation coefficient between the WCRF/AICR standardized lifestyle score and the WCRF/AICR diet score, expanded lifestyle score and standardized lifestyle score + smoking were 0.80, 0.96, and 0.96, respectively. Only 32 participants (0.16% of the study sample) obtained the maximum score for any of the lifestyle scores. The individual cancer prevention recommendation in the standardized lifestyle score that was least likely to be met was limiting the consumption of red and processed meat (only 8% of participants were “adherent”) while 31%–34% of participants met the criteria for other dietary recommendations except sugar-sweetened beverages (55% “adherent”) and 25% and 72% met the criteria for healthy weight and physical activity, respectively.

Characteristics of the study sample, overall and stratified by approximate quintiles of the WCRF/AICR standardized lifestyle score (quintile 1 = “least adherent” and quintile 5 = “most adherent”), are shown in Table 2. The mean age at study recruitment was 65.3 (SD = 8.7) years. Women were more likely to be adherent to the recommendations [e.g., median standardized lifestyle score = 4.25 (interquartile range = 3.5–5) compared with 3.75 (3.25–4.5) for men]. A higher level of educational attainment and lower proportion of smoking were evident with increasing lifestyle score (e.g., 32.1% with tertiary education in quintile 5 compared with 24.2% in quintile 1 and 67.2% never smokers in quintile 5 compared with 54.5% in quintile 1). During a mean follow-up of 16 years, 4,710 incident cancers were diagnosed. These included 1,307 alcohol-related, 1,812 obesity-related, 1,444 smoking-related, and 974 digestive tract cancers.

An estimated 18% lower risk of total cancer was observed for the highest quintile (“most adherent”) of the WCRF/AICR standardized lifestyle score (HR = 0.82; 95% CI: 0.74–0.92; P_trend <0.001), compared with the lowest quintile (“least adherent”) (Table 3). The corresponding risk reduction for the “most adherent” quintile of the WCRF/AICR standardized lifestyle score was higher for alcohol-related (HR = 0.65; 95% CI: 0.53–0.79; P_trend <0.001), obesity-related (HR = 0.64; 95% CI: 0.54–0.76; P_trend <0.001), smoking-related (HR = 0.71; 95% CI: 0.58–0.86; P_trend <0.001) and digestive tract (HR = 0.73; 95% CI: 0.57–0.92; P_trend = 0.002) cancers. The associations for the highest quintile, compared with lowest quintile, were slightly weakened when the score was limited to dietary factors (e.g., HR = 0.91; 95% CI: 0.82–1.02; P_trend = 0.046 for total cancer; HR = 0.74; 95% CI: 0.60–0.91; P_trend = 0.049 for alcohol-related cancer; HR = 0.77; 95% CI: 0.64–0.91; P_trend = 0.02 for obesity-related cancer) or when weight change and sedentary behavior were incorporated into the score (e.g., HR = 0.88; 95% CI: 0.80–0.98; P_trend <0.001 for total cancer; Table 3). In contrast, for the lifestyle score which included smoking, the association was either stronger (HR = 0.74; 95% CI: 0.67–0.81; P_trend <0.001 for total cancer; HR = 0.55; 95% CI: 0.46–0.66; P_trend <0.001 for smoking-related cancer; HR = 0.64; 95% CI: 0.51–0.79; P_trend <0.001 for digestive tract cancer) or did not change appreciably (for alcohol-related and obesity-related cancers) for the highest quintile, compared with the lowest quintile (Table 3). The overall pattern of these associations remained stable when the first 2 years of follow-up were excluded (Supplementary Table S2) or when smoking intensity was included in the models instead of smoking status (Supplementary Table S3).

No evidence of departure from a linear dose–response relationship was observed between diet and lifestyle scores and total cancer risk (P_nonlinearity > 0.05; Supplementary Fig. S4). A one-unit increment in lifestyle scores was associated with a 6% to 9% lower risk of total cancer (e.g., HR = 0.94; 95% CI: 0.91–0.96 for WCRF/AICR standardized lifestyle score; Table 3). Lower cancer risk of approximately similar or stronger magnitude was observed per one-unit increment in lifestyle scores for alcohol-related, obesity-related, smoking-related, and digestive tract cancers (Table 3). Consistent with findings using quintiles, lower risk of digestive tract cancer per one-unit increment in score was only observed for the WCRF/AICR standardized lifestyle score and the lifestyle score which included smoking. A one-unit increment in the WCRF/AICR diet score was associated with lower risk of total (HR = 0.96; 95% CI: 0.93–1.00), alcohol-related (HR = 0.92; 95% CI: 0.85–0.99), and obesity-related (HR = 0.93; 95% CI: 0.87–0.99) cancers; the associations were weak for smoking-related and digestive tract cancers (Table 3). There was no evidence that sex (Supplementary Tables S4–S5; Supplementary Fig. S5), smoking or age (Supplementary Fig. S6) modified these associations (P_interaction > 0.05).

Figure 1 depicts HRs per one-unit increment in diet and lifestyle scores for the 10 most common cancers. Higher diet and lifestyle scores were consistently associated with lower risk of breast cancer (e.g., HR = 0.88 per one-unit increment; 95% CI: 0.81–0.95 for WCRF/AICR standardized lifestyle score) and prostate cancer (e.g., HR = 0.93; 95% CI: 0.86–0.99 for WCRF/AICR standardized lifestyle score). A higher lifestyle score that included smoking was associated with lower risk of lung (HR = 0.64, 95% CI: 0.57–0.72), stomach (HR = 0.78; 95% CI: 0.63–0.97), rectal (HR = 0.83; 95% CI: 0.71–0.97) and pancreatic (HR = 0.85; 95% CI: 0.71–1.01) cancers. The observed associations for colon, bladder, kidney, and UADT cancers were weaker.

We also examined the independent association for each individual component used in the diet and lifestyle scores with total cancer risk (Fig. 2). Evidence for reduction in total cancer risk was strongest for abstention from alcohol, healthy waist circumference and never smoking. Weak evidence of association was also observed for being physically active and low red and processed meat consumption.

Greater adherence to 2018 WCRF/AICR cancer prevention recommendations on diet, alcohol consumption, body size, and physical activity (8) translated to substantial reductions in cancer risk including overall, alcohol-related, obesity-related, and smoking-related cancer. The inclusion of abstinence from smoking, the single most important behavior related to cancer risk (25), amplified the cancer risk reduction effect in our data, the lifestyle score with smoking being associated with 26% lower risk for cancer overall for the most adherent group, up from 18% lower without it. Greater adherence to the 2018 guidelines was also associated with lower risk of breast and prostate cancers. Adding abstinence from smoking extended the range of cancers “protected” against to include several smoking-related cancers, namely lung, stomach, rectal, and pancreatic cancers. Adherence to the dietary advice alone led to a smaller risk reduction across all cancers although it was still notable for alcohol-related and obesity-related cancer.

The current study is a comprehensive assessment of a broad group of key modifiable cancer-related risk factors using various adaptations of WCRF/AICR recommendations and numerous cancer endpoints. We believe our findings will help improve the current scoring system during its operationalization and application in research. The few published prospective studies on the 2018 WCRF/AICR cancer prevention recommendations predominantly assessed cancer risk related to the WCRF/AICR standardized lifestyle score (10–13, 32–35) and rarely derived scores including smoking (13, 33), weight change (12, 32), sedentary behavior (12), and diet components alone (35), and frequently lacked participant data on waist circumference (10, 32, 34, 35), fast foods (13, 32) and sugar-sweetened beverages (13, 32). These studies commonly examined either the risk of total cancer (11) or a single cancer type (32–35) except three that assessed risks for broad cancer groupings and multiple cancer sites (10, 12, 13). In contrast, we derived four diet and lifestyle scores based on seven of the 2018 WCRF/AICR cancer prevention recommendations including components on fast foods, sugar-sweetened beverages and waist circumference, and subcomponents on weight change and sedentary behavior; and investigated a lifestyle score that also included smoking as a component; and assessed risk of total, alcohol-related, obesity-related, smoking-related, digestive tract, and site-specific cancer. We reaffirm that potentially modifiable diet and lifestyle-related behaviors proposed in the 2018 WCRF/AICR cancer prevention recommendations are idealized states. Identifying evidence-based guidelines that could translate into better informed cancer prevention strategies is important, nonetheless. Even modest reductions in population exposure to potentially modifiable causal agents would likely translate to sizeable reductions in cancer risk and premature deaths from cancer. Other strengths of the current study include the prospective design over a 16-year follow-up period, availability of detailed information on diet, body size, alcohol intake, smoking and physical activity, and potential confounders, a large number of incident cancer cases, and complete follow-up through linkage with state and national registries and databases. A key strength of the 2018 WCRF/AICR cancer prevention recommendations themselves is that they closely align with corresponding guidelines promoting overall health. For instance, the Australian National Health and Medical Research Council recognizes risky drinking, lack of fruit and vegetable intake, greater processed meat intake and BMI ≥30 kg/m² as threats to optimum human health (36–38).

This study has some limitations. There is likely to be measurement error and exposure misclassification as components other than weight were measured via self-report and were assumed to be constant over time (39); however, this is likely to attenuate the findings to the null. Furthermore, by using approximate quintiles, we minimized systematic error previously highlighted when a priori score-driven categories were used where participants with values closer to category cut-off points are either rewarded or penalized (10). Unmeasured residual confounding, particularly due to smoking habits (although we further adjusted for smoking intensity that accounted for the dose of smoking among current smokers and the time since quitting for former smokers), and sun exposure for melanoma, cannot be ruled out. Potential confounding due to screening and diagnostic intensity may lead to biased risk estimates for commonly screen-detected cancers such as prostate cancer (40). We minimized their impact on our results by controlling models assessing prostate, breast, and colorectal cancer risks for screening. Our findings using small numbers of cancers also need to be interpreted with caution. For example, the estimated effect sizes for colon, rectal, lung, pancreatic, kidney, stomach and UADT cancer with standardized lifestyle score were similar to or stronger than those for prostate cancer, but not statistically significant. Similarly, the estimated effect size for rectal cancer with the diet score was similar to that for breast and prostate cancers, but not statistically significant; there were four to six times more breast and prostate than rectal cancers and rectal cancer comprised only 16% of digestive tract cancer, which only showed a weak association per unit increment in the diet score. Furthermore, we did not include the recommendation on breast feeding (34); its inclusion is not mandatory as it only applies to a specific subpopulation (8). Generalizability of our results may also be limited because the study participants were relatively healthy middle-aged adults of European ancestry living in a high-income country (14). The main cancer risk factors affecting the Australian population (e.g., smoking, obesity, dietary risks; ref. 41) are different to those affecting people living in economically less developed settings (e.g., infection, air pollution, occupational toxins; ref. 3).

From a methodologic perspective, operationalizing a score with greater relative weights for cancer prevention recommendations on alcohol use, weight, and physical activity in comparison with other dietary recommendations could address the underestimation of the beneficial effect from adhering to cancer prevention guidelines (12). Our analysis using individual components suggested that alcohol use, waist circumference, and smoking may be driving associations with total cancer risk. Others have identified different drivers of cancer risk (10, 11) meaning that more work is needed to determine the optimal weighting between components, an aspect beyond the scope of the current study. The 2018 WCRF/AICR cancer prevention recommendations also included additional goals on avoiding weight gain in adulthood and limiting sedentary habits (including screen time; refs. 8, 9). Shams-White and colleagues (8, 9) did not operationalize these within their scoring system, given the challenges of operationalizing weight gain within the scope of the current scoring system and the lack of consensus on cut-off points for sedentary habits. They, however, encourage exploratory efforts that incorporate measures of weight change and sedentary habits to improve the utility of the standardized lifestyle score (6). In the current study, adding weight change in adulthood and time spent watching television as subcomponents to the WCRF/AICR standardized lifestyle score did not translate to a higher cancer risk reduction. This finding is consistent with those from the Nurses’ Health Study and the Health Professionals Follow-up Study (12). While including weight change and sedentary habits led to downweighting of the effects of waist circumference and physical activity, respectively, both important drivers of lower cancer risk in our data, limitations related to how weight change and sedentary habits were defined and their cut-off points used could have also contributed to the current findings. The current study, however, is limited by the use of combined lifestyle factors at only one timepoint, that is, at second follow-up assessment in 2003 to 2007 because detailed information on physical activity is not available from the study baseline (1990–1994) or the first follow-up (1995–1998). Looking at changes in risk factors may be important in studies with such data available.

We observed a 6% lower risk of total cancer for each unit increment in the WCRF/AICR standardized lifestyle score, in comparison with 3% in the Cohort of Swedish Men and Swedish Mammography Cohort with contrasting dietary and lifestyle habits, adjustment in their models for additional covariates including aspirin use, dietary supplement use and history of diabetes mellitus, and more men than women (11). An identical 6% lower risk per interquartile range increment in the WCRF/AICR standardized lifestyle score for total cancer risk was reported in the Nurses’ Health Study (interquartile range: 3.13–4.42) and the Health Professionals Follow-up Study (interquartile range: 3.10–4.38; ref. 12). Their HRs comparing the highest versus lowest quintile of the WCRF/AICR standardized lifestyle score (women: HR = 0.89, 95% CI: 0.85–0.94; men: HR = 0.87, 95% CI: 0.81–0.94) and relatively stronger effect sizes for obesity-related, alcohol-related, smoking-related, and digestive tract cancers were also consistent with our findings (12). Our findings for total cancer are not directly comparable to published UK Biobank data (13) or U.S. data from the NIH-AARP Diet and Health Study (10) but are generally consistent with those findings. The UK Biobank data (13) used an increment in SD as the continuous exposure (HR = 0.89, 95% CI: 0.87–0.90) and considered diagnosis with any of the 13 cancers selected as the overall cancer endpoint. The NIH-AARP data (10) that found a 6%–11% lower risk for each unit increment, except for male current smokers for whom the association was weak, defined as their largest cancer endpoint any of the 17 cancers (which made up 77% of the total cancers in the current study) included in the 2018 WCRF/AICR Expert Report with evidence suggesting causal links to diet, alcohol use, body size, and physical activity (6). Significant inconsistencies in findings across different smoking strata in the NIH-AARP data (10) should encourage future studies to derive lifestyle scores similar to ours that incorporate abstinence from smoking, as has also been recommended by WCRF/AICR (8). Our findings for lung, stomach, rectal, and pancreatic cancer that recognize the additional benefits of abstinence from smoking, consistent with findings from the UK Biobank (13), further justify extending the range of core lifestyle behaviors to include abstinence from smoking in a more comprehensive lifestyle score. We also reaffirm previous evidence of the importance of adhering to WCRF/AICR cancer prevention recommendations to prevent breast cancer (10, 13).

In summary, adherence to the latest cancer prevention recommendations of the WCRF/AICR based on diet, alcohol use, body size, and physical activity was associated with lower risk of cancer. Considering that the majority of our study sample did not meet the individual cancer prevention recommendations and that only <0.2% were classified as fully adherent to any lifestyle score, the current study emphasizes the vast potential for preventing cancer had the recommendations been adhered to more closely. Future studies should assess the robustness of the weights ascribed to individual modifiable diet and lifestyle-related components that make up the WCRF/AICR scores.

B.M. Lynch reports grants from Victorian Cancer Agency and World Cancer Research Fund International, and non-financial support from Uppsala University outside the submitted work. G.G. Giles reports grants from National Health and Medical Research Council during the conduct of the study. No disclosures were reported by the other authors.

Y. Peng: Conceptualization, data curation, formal analysis, methodology, writing–original draft, writing–review and editing. J.K. Bassett: Conceptualization, data curation, methodology, writing–review and editing. A.M. Hodge: Conceptualization, methodology, writing–review and editing. Y.A. Melaku: Methodology, writing–review and editing. N. Afshar: Methodology, writing–review and editing. J.L. Hopper: Methodology, writing–review and editing. R.J. Maclnnis: Conceptualization, methodology, writing–review and editing. B.M. Lynch: Conceptualization, Methodology, writing–review and editing. S.A. Smith-Warner: Conceptualization, Methodology, writing–review and editing. G.G. Giles: Funding acquisition, methodology, writing–review and editing. R.L. Milne: Conceptualization, resources, methodology, writing–review and editing. H. Jayasekara: Conceptualization, resources, data curation, supervision, methodology, writing–original draft, project administration, writing–review and editing.

We thank the many thousands of Melbourne residents who continue to participate in the MCCS.

MCCS cohort recruitment was funded by VicHealth and Cancer Council Victoria. The MCCS was further augmented by Australian National Health and Medical Research Council grants 209057, 396414, and 1074383 and by infrastructure provided by Cancer Council Victoria. Cases and their vital status were ascertained through the Victorian Cancer Registry and the Australian Institute of Health and Welfare, including the Australian Cancer Database. Y. Peng and H. Jayasekara are supported by NHMRC grant GNT1163120. B.M. Lynch is supported by the Victorian Cancer Agency (MCRF18005).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.