16th December 2016, Volume 129 Number 1447

Ann Richardson, James Hayes, Chris Frampton, John Potter

Colorectal cancer accounts for almost 10% of cancer incidence worldwide,1 and New Zealand is among the countries with the highest incidence rates in the world.2 Each year there are about 3,000 registrations and 1,100 deaths from colorectal cancer in New Zealand.3 Incidence and mortality rates are lower for Māori than non-Māori, but rates for Māori and non-Māori females are starting to converge.3 Colorectal cancer incidence rates vary more than 20-fold worldwide, and changes occurring among populations migrating from low to high incidence areas within one or two generations suggest that lifestyle factors are important in determining risk.4 Substantial reductions in the incidence of colorectal cancer (more than 10% in the decade 2006 to 2016) have been projected for New Zealand, even in the absence of a national screening programme; however, the number of people diagnosed with colorectal cancer each year will not decline, because of the increasing size and age of the population.5

This project was designed to identify modifiable lifestyle factors that could reduce the future incidence of colorectal cancer among New Zealanders. Ten years ago, Cox and Sneyd estimated the contribution to the control of colorectal cancer achievable from primary prevention, screening, early diagnosis and treatment in New Zealand.6 Since then, new estimates of the effects of certain modifiable lifestyle risk factors for colorectal cancer have been published.1,7

The population attributable fraction (also known as the population attributable risk percent, or population etiological fraction) is the proportion of disease in the population due to a specific exposure that could be prevented if that exposure were eliminated.8 Population attributable fractions (PAFs) help to identify the exposures that have the greatest impact on the health of a population. By comparing PAFs for different risk factors for a disease, the relative importance of the risk factors can be assessed. For a modifiable risk factor, the PAF also provides an indication of the maximum possible impact of primary preventive strategies.

The purpose of this research was to identify modifiable lifestyle factors for colorectal cancer among New Zealanders that could be amenable to primary preventive strategies. The results of this research can provide New Zealanders and their health providers with information to allow people to reduce their risk of colorectal cancer.

Method

Interpretation of a population attributable fraction (PAF) assumes that the association between the exposure and the disease is causal. Causal lifestyle factors for colorectal cancer have been identified; convincing evidence has been found for obesity, lack of physical activity, consumption of red meat, consumption of processed meat, alcohol and smoking.1,7 This project was not intended to provide a comprehensive literature review of these modifiable risk factors for colorectal cancer, but rather to identify robust estimates of relative risks with 95% confidence intervals, that could be used to calculate PAFs. Priority was given to statistically significant relative risks reported from well-designed systematic reviews, intervention studies or cohort studies, ahead of odds ratios reported from case-control studies. All the relative risks used in the PAF calculations were adjusted; we did not use crude relative risks.

Prevalence of lifestyle risk factors in New Zealand

Once the relative risk estimates for each lifestyle risk factor had been obtained from the literature, information on the prevalence of the risk factors in New Zealand was collected. The most recent information on the prevalence of each risk factor in New Zealand was obtained from New Zealand Ministry of Health publications and online data. Ministry of Health publications report the results of population-based surveys such as the 2013/14 New Zealand Health Survey,9 the New Zealand Alcohol and Drug Use Survey10 and the 2008/09 New Zealand Adult Nutrition Survey.11

It was not always possible to obtain prevalence estimates that aligned exactly with the exposure categories used to calculate the relative risks. Where it was not possible to obtain prevalence estimates that aligned exactly, we took a conservative approach in selecting prevalence estimates to avoid overestimating PAFs.

Crude prevalence estimates (rather than age-standardised prevalence estimates) were used to calculate PAFs, as these provided the estimates of the actual proportion of the population or population subgroup exposed to each risk factor, which are required for calculating PAFs. Prevalence estimates for males and females and Māori, Pacific, Asian and European/other New Zealanders were used to calculate sex-specific and ethnic group-specific PAFs.

Calculation of population attributable fractions

Information about the prevalence of risk factors in New Zealand, and the estimates of relative risks abstracted from the literature, were used to calculate the population attributable fraction (PAF).

 

c

 

To calculate 95% confidence intervals for the PAFs, we simulated normal or log-normal distributions for RR and Pe based on the CI calculated or provided from published sources. Pairs of RR and Pe values were then randomly selected from these distributions and a PAF calculated for each random selection.12 From 10,000 calculated PAF values, the empirical distribution of PAF was then used to estimate a two-sided 95% CI. SPSS v22 (SPSS Inc., Armonk, NY, USA) was used for these simulations and calculations.

Results

Risk factors

Alcohol

Alcohol increases the risk of colorectal cancer, and there is a dose-response effect, with the risk increasing with the amount of alcohol consumed.1,13 The most recent estimate of the effect of alcohol on colorectal cancer comes from a pooled analysis of 66 epidemiological studies, where the relative risk for heavy drinking (>50g or >5 drinks per day, as a standard drink in New Zealand is 10g alcohol) is 1.44 (1.25–1.65).13

Diet

Consumption of red meat increases the risk of colorectal cancer, with a dose-response relationship shown in cohort studies.1 A meta-analysis of 23 studies found an increased risk of colorectal cancer with consumption of red meat. The relative risk for the highest quartile of consumption of red meat compared with the lowest was 1.35 (1.21–1.51).14 Consumption of processed meat also increases the risk of colorectal cancer. The relative risk for the highest quartile of consumption of processed meat compared with the lowest was 1.31 (95% CI 1.13–1.51).14

Obesity

Obesity increases the risk of colorectal cancer; both general obesity (measured as BMI) or central obesity (measured as waist circumference).15 A recent systematic review compared the risk of colorectal cancer for the obese (BMI 30kg/m2 or higher) vs normal (BMI less than 25kg/m2) category of BMI and reported a summary relative risk of 1.33 (1.25–1.42).15

Physical activity

Lack of physical activity is associated with an increased risk of colorectal cancer. Physical activity is associated with energy intake and partly determines BMI. A review on the effect of physical inactivity found that the relative risk of colon cancer for physical inactivity was 1.32 (1.23–1.39).16

Tobacco

Tobacco smoking is associated with increased colorectal cancer risk,1 although this is restricted to a molecularly defined subset of colorectal cancer.17 A systematic review and meta-analysis of prospective studies of cigarette smoking and colorectal cancer18 found a summary relative risk of 1.15 (1.00–1.32) for current versus never-smokers.

Information on the prevalence of risk factors in New Zealanders

Alcohol

In New Zealand 16.1% (15.2–17.0) of adults have a hazardous drinking pattern, as indicated by an Alcohol Use Disorders Identification Test (AUDIT) score of eight or more,9 the closest estimate available for New Zealanders to the relevant exposure category used in epidemiological studies of alcohol consumption and colorectal cancer.

Physical inactivity

The most recent estimates among New Zealanders of the prevalence of sedentary behaviour (defined as less than 30 minutes of physical activity in the last week) are derived from the most recent New Zealand Health Survey of adults aged 15 years and over.9 In the total population, 14.3% (13.3–15.5) of adults are physically inactive.

Obesity

Body mass index (BMI) is the most commonly used measure of body size. It is calculated by dividing weight (in kilograms) by height squared (in metres). Obesity is defined as a BMI of 30 or more.9 In New Zealand, 29.9% (28.9–30.9) of adults are obese.

Smoking

In New Zealand, 17.2% (16.4–18.1) of adults are current smokers. The prevalence of current smoking is defined as ‘smoked more than 100 cigarettes in a lifetime and currently smoking at least once a month’.9

Red meat

Most of the total population aged 15 years and over (94.5%) reported eating red meat in the past four weeks, with 14.4% (12.7–16.1) eating red meat five or more times per week.11 Estimates for consumption five or more times per week were not reported by ethnic group.

Processed meat

Processed meat was eaten in the past four weeks by 87.3% of the total population aged 15 years and over, with 8.6% (7.3–9.8) eating processed meat five or more times per week.11 Estimates for consumption five or more times per week were not reported by ethnic group.

The PAF results for the total New Zealand population are shown in Table 1 and the ethnic group-specific results for New Zealand men and women are shown in Tables 2 and 3 respectively. In all three tables, the relative risks and prevalence estimates for each risk factor are as follows: for alcohol, the RR is for heavy drinking (>50g or >5 drinks/day) and prevalence estimates are for AUDIT scores of 8 and over; for smoking, the RR is for current vs never-smokers, and the prevalence estimates are for current smokers (smoked more than 100 cigarettes in lifetime and smoking at least once a month); for obesity, the RR is for obese (BMI 30kg/m2 or higher) vs normal BMI (less than 25kg/m2), and the prevalence estimates are for obesity (BMI 30kg/m2 or higher); for physical inactivity, the RR is for an activity level insufficient to meet the WHO recommendations for adults aged 18–64 years of at least 75 minutes of vigorous-intensity aerobic physical activity throughout the week,19 and the prevalence estimates are for <30 minutes physical activity in the last seven days; for consumption of red meat, the RR is for the highest vs the lowest quartile of consumption and the prevalence estimates are for consumption of red meat five or more times per week; for consumption of processed meat, the RR is for the highest vs the lowest quartile of consumption and the prevalence estimates are for consumption of processed meat five or more times per week.

Table 1: Colorectal cancer modifiable lifestyle risk factors, prevalence (%) and population attributable fractions (PAF) (%) for the total New Zealand population.

Risk factor

Relative risk

(95% CI)

Prevalence of risk

factor (95% CI)

PAF (95% CI)

Alcohol

1.44 (1.25–1.65)

16.1 (15.2–17.0)

6.6 (3.6–9.6)

Smoking

1.15 (1.00–1.32)

17.2 (16.4–18.1)

2.5 (0.0–5.2)

Obesity

1.33 (1.25–1.42)

29.9 (28.9–30.9)

9.0 (6.7–11.2)

Physical inactivity

1.32 (1.23–1.39)

14.3 (13.3–15.5)

4.4 (3.1–5.6)

Red meat

1.35 (1.21–1.51)

14.4 (12.7–16.1)

4.8 (2.6–7.0)

Processed meat

1.31 (1.13–1.51)

8.6 (7.3–9.8)

2.6 (0.9–4.3) 

Table 2: Colorectal cancer modifiable lifestyle risk factors, prevalence (%) and population attributable fractions (PAF) (%) for New Zealand men.

Risk factor

Relative risk

(95% CI)

Prevalence of risk factor (95% CI)1

PAF (95% CI)

Alcohol

Māori

Pacific

Asian

European/other

1.44 (1.25–1.65)

 

36.7 (32.7–41.0)

26.7 (20.7–33.6)

5.3 (3.7–7.5)

22.7 (21.2–24.4)

 

13.8 (7.9–19.7)

10.4 (5.3–15.5)

2.3 (0.9–3.7)

9.0 (5.0–13.1)

Smoking

Māori

Pacific

Asian

European/other

1.15 (1.00–1.32)

 

39.5 (35.1–44.0)

28.3 (21.0–36.9)

14.4 (11.1–18.5)

16.2 (14.9–17.7)

 

5.6 (0.0–11.3)

4.0 (0.0–8.5)

2.1 (0.0–4.5)

2.4 (0.0–4.9)

Obesity

Māori

Pacific

Asian

European/other

1.33 (1.25–1.42)

 

 

42.6 (39.0–46.2)

62.3 (55.8–68.4)

14.6 (10.8–19.5)

28.3 (26.5–30.1)

 

12.3 (9.2–15.5)

17.0 (12.8–21.1)

4.6 (2.7–6.5)

8.5 (6.3–10.7)

Physical inactivity

Māori

Pacific

Asian

European/other

1.32 (1.23–1.39)

 

13.9 (11.1–17.2)

15.1 (11.1–20.3)

15.4 (11.4–20.6)

11.0 (9.8–12.4)

 

4.2 (2.8–5.7)

4.6 (2.6–6.5)

4.7 (2.7–6.7)

3.4 (2.4–4.4)

Red meat

1.35 (1.21–1.51)

15.7 (13.6–17.8)

5.2 (2.8–7.5)

Processed meat

1.31 (1.13–1.51)

12.0 (9.7–14.3)

3.6 (1.3–5.9)

Source: Ministry of Health 2016 online tables http://www.health.govt.nz/publication/annual-update-key-results-2013-14-new-zealand-health-survey

Table 3: Colorectal cancer modifiable lifestyle risk factors, prevalence (%) and population attributable fractions (PAF) (%) for New Zealand women.

Risk factor

Relative risk

(95% CI)

Prevalence of risk factor (95% CI)1

PAF (95% CI)

Alcohol

Māori

Pacific

Asian

European/other

1.44 (1.25–1.65)

 

24.8 (21.9–28.0)

12.3 (8.9–16.7)

2.2 (1.0–4.1)

10.5 (9.1–11.9)

 

9.8 (5.4–14.1)

5.1 (2.2–8.0)

1.0 (0.0–1.9)

4.4 (2.3–6.5)

Smoking

Māori

Pacific

Asian

European/other

1.15 (1.00–1.32)

 

41.6 (38.6–44.7)

22.2 (17.7–27.4)

 2.3 (1.3–3.9)

14.3 (13.2–15.4)

 

5.8 (0.2–11.9)

3.2 (0.3–6.7)

0.3 (0.1–0.8)

2.1 (0.1–4.3)

Obesity

Māori

Pacific

Asian

European/other

1.33 (1.25–1.42)

 

 

48.2 (44.9–51.5)

70.8 (64.6–76.3)

13.9 (11.0–17.4)

27.5 (25.9–29.1)

 

13.7 (10.4–17.0)

18.9 (14.5–23.3)

4.4 (2.8–5.9)

8.3 (6.2–10.4)

Physical inactivity

Māori

Pacific

Asian

European/other

1.32 (1.23–1.39)

 

17.7 (15.4–20.3)

26.1 (21.0–32.0)

25.0 (19.8–31.1)

13.3 (11.8–14.9)

 

5.4 (3.8–7.0)

7.7 (5.1–10.2)

7.4 (4.8–9.9)

4.1 (2.9–5.3)

Red meat

1.35 (1.21–1.51)

13.2 (11.1–15.3)

4.4 (2.3–6.5)

Processed meat

1.31 (1.13–1.51)

5.4 (4.1–6.6)

1.6 (0.5–2.8)

Source: Ministry of Health 2016 online tables http://www.health.govt.nz/publication/annual-update-key-results-2013-14-new-zealand-health-survey 

Because colorectal cancer is a common disease in New Zealand, with over 3,000 people diagnosed each year, even small changes in PAFs mediated by a reduction in the prevalence of risk factors could translate into appreciable numbers of New Zealanders avoiding a diagnosis of colorectal cancer. Table 4 shows the reduction in numbers of people diagnosed with colorectal cancer each year if the prevalence of modifiable lifestyle risk factors in New Zealand was halved.

Table 4: Effect of halving the prevalence of modifiable lifestyle risk factors on the number of people diagnosed with colorectal cancer in New Zealand each year.

Risk factor

Prevalence of risk factor (%)

Reduction in colorectal cancers diagnosed

(number of people per year)1

Alcohol

8.0

102

Smoking

8.6

39

Obesity

15.0

141

Physical inactivity

7.2

69

Red meat

7.2

75

Processed meat

4.3

39

PAFs cannot be summed, because in any individual there may be more than one contributing cause of colorectal cancer (see Discussion). 

Discussion

In this study, six modifiable lifestyle risk factors were identified for colorectal cancer. These risk factors were 9% for obesity, 7% for alcohol, 4% for insufficient physical activity, 3% for smoking, 5% for consumption of red meat and 3% for processed meat.

Similar findings were seen in papers reporting PAFs for colorectal cancer in Australia and the UK.20,21 The Australian estimates are 9% for overweight and obesity, 9% for alcohol, 5% for insufficient physical activity, 6% for smoking and 18% for red and processed meat combined.20 The UK estimates are 13% for overweight and obesity, 12% for alcohol, 3% for insufficient physical activity, 8% for smoking and 21% for red and processed meat combined.21 It is important to note that PAF estimates will differ across countries if the prevalence of modifiable risk factors differs, with the PAF increasing with the prevalence of the risk factor.

Fewer than half of New Zealand adults are physically active.9 The increasing prevalence of obesity in New Zealand has been attributed to changing dietary and physical activity patterns. Mean BMI in New Zealand adults increased since 1997, although this has levelled off recently, with more data required to confirm whether this represents a slowing in the increasing prevalence of obesity.9 There are large differences in the prevalence of obesity by ethnicity in New Zealanders (as shown in Tables 2 and 3).

Alcohol consumption is associated with an increased risk of colorectal cancer, and the prevalence of hazardous drinking has been declining in New Zealanders in recent years.9 Hazardous alcohol use markedly differs according to ethnic group (Tables 2 and 3). Among people who consume alcohol regularly, avoidance of hazardous drinking could have an important impact on colorectal cancer incidence. Because the prevalence estimate for alcohol was for hazardous drinking, our results will underestimate the true effect of alcohol on colorectal cancer (moderate to heavy drinking also increases the risk of colorectal cancer).

PAFs have limitations. PAF calculations require good estimates of the strength of association and the prevalence of risk factors, but it is not always possible to find population prevalence data that exactly reflect the risk factor measurements used in epidemiological studies. It is relevant that the PAFs we calculated for colorectal cancer in New Zealand are similar to those calculated for Australia and the UK.20,21

A second limitation is that the individual contributions of causal factors must be considered when interpreting PAFs.22 A disease may have more than one cause (for instance, smoking, high cholesterol, lack of exercise and obesity may all contribute to cardiovascular disease in the same individual). This means that potentially, this person’s illness could have been prevented if any one of these risk factors had been absent, and the sum of the PAFs for these risk factors therefore, may add to more than 100%, reflecting the multi-causal nature of disease.23 Thus, the PAF is best used as an indication of the relative importance of risk factors across a population and the possible impact of behaviour change.

Another limitation of PAFs is that confounding cannot be taken into account with the method for calculating PAF used in this paper; for instance, the association between obesity and colorectal cancer is plausibly confounded by lack of physical activity, but this cannot be adjusted for in the PAF calculation. A formula for calculating adjusted PAFs has been developed,24 but it was inappropriate for this paper, as the relative risks were obtained from many separate studies rather than from a single study, and the prevalence estimates were obtained from routinely collected data.

Provided these limitations are borne in mind, the PAF is useful as an indication of the relative importance of modifiable risk factors and the potential impact of primary preventive strategies in a population. It also provides a way to apply the results from international epidemiological studies to the New Zealand population. For example, halving the prevalence of modifiable risk factors for colorectal cancer would mean 141 fewer cases of colorectal cancer due to obesity, 102 fewer due to alcohol, 69 fewer due to physical inactivity, 39 fewer due to smoking, 75 fewer due to consumption of red meat and 39 fewer due to consumption of processed meat each year in New Zealand.

Conclusion

The most important modifiable lifestyle risk factors for colorectal cancer in New Zealand are obesity and alcohol consumption. These findings have considerable public health relevance since they suggest that it is possible to prevent an appreciable proportion of colorectal cancer by changing a few selected lifestyle factors. In addition to reducing the incidence of colorectal cancer, a reduction in obesity, alcohol consumption and smoking, and an increase in physical activity, would also reduce the incidence of other cancers, cardiovascular disease and diabetes in New Zealand.

Summary

Aspects of our lifestyle can increase the risk of bowel cancer. By combining the results of international research and the prevalence of risk factors in New Zealand, it is possible to estimate the impact of reducing our exposure to these risk factors. If obesity, alcohol consumption, smoking and consumption of red and processed meats could be reduced, and physical activity could be increased among New Zealanders, it would reduce the risk of bowel cancer considerably.

Abstract

Aim

To estimate population attributable fractions for modifiable lifestyle factors and colorectal cancer in New Zealand.

Method

Relative risks for lifestyle risk factors for colorectal cancer, and population data on the prevalence of exposure in New Zealand, were used to estimate the population attributable fraction (PAF) for each risk factor.

Results

Six modifiable lifestyle risk factors were identified. The PAFs for these risk factors were 9% for obesity, 7% for alcohol, 4% for physical inactivity, 3% for smoking, 5% for consumption of red meat and 3% for processed meat. PAFs differed by ethnic group and sex. In women, the highest PAFs were 19% for obesity in Pacific women, 14% for obesity in Māori women, 7% for physical inactivity in Asian women, and 8% for obesity in European/other women. In men, the highest PAFs were 17% for obesity in Pacific men, 14% for high alcohol consumption in Māori men, 5% for physical inactivity in Asian men and 9% for high alcohol consumption in European/other men.

Conclusion

If obesity, alcohol consumption, smoking and consumption of red and processed meats could be reduced, and physical activity could be increased among New Zealanders, it would reduce the risk of colorectal cancer considerably.

Author Information

Ann Richardson, Professor, Wayne Francis Cancer Epidemiology Research Group, School of Health Sciences, University of Canterbury, Christchurch; James Hayes, Senior Academic Lecturer, Medical Imaging, Department of Applied Science and Allied Health, Ara Institute of Canterbury, Christchurch; Chris Frampton, Professor, Biostatistics and Computational Biology Unit, University of Otago, Christchurch; John Potter, Chief Science Advisor, Ministry of Health, Wellington, Adjunct Professor, School of Health Sciences, University of Canterbury, Christchurch, Professorial Fellow, Centre for Public Health Research, Massey University, Wellington, Senior Advisor, Fred Hutchinson Cancer Research Center, Seattle, USA.

Acknowledgements

The Wayne Francis Cancer Epidemiology Research Group receives support from the Wayne Francis Charitable Trust.

Correspondence

Ann Richardson, Professor, Wayne Francis Cancer Epidemiology Research Group, School of Health Sciences, University of Canterbury, Private Bag 4800, Christchurch 8140.

Correspondence Email

ann.richardson@canterbury.ac.nz

Competing Interests

Nil.

References

  1. Stewart BW, Wild CP. World Cancer Report 2014. Lyon: IARC, 2014.
  2. Torre LA, Bray F, Siegel RL, et al. Global cancer statistics, 2012. CA: Cancer Journal for Clinicians. 2015; 65 (2):87–108.
  3. Ministry of Health. Cancer: new registrations and deaths 2011. Wellington: Ministry of Health, 2014.
  4. Potter JD, Hunter D. Colorectal cancer. In: Adami HO, Hunter D, Trichopoulos D, eds. Textbook of cancer epidemiology. 2nd edition. New York: Oxford University Press, 2008.
  5. Ministry of Health. Cancer projections: incidence 2004–08 to 2014–18. Wellington: Ministry of Health, 2010.
  6. Cox B, Sneyd M. Prospects for cancer control: colorectal cancer. New Zealand Medical Journal. 2005; 118 (1221):U1632.
  7. World Cancer Research Fund, American Institute for Cancer Research. Food, Nutrition, Physical Activity, and the Prevention of Cancer: a Global Perspective. Washington DC: World Cancer Research Fund, 2007.
  8. Last JM. Dictionary of Epidemiology. Vol 4: New York: Oxford University Press, 2001.
  9. Ministry of Health. New Zealand Health Survey 2013/14. Wellington: Ministry of Health, 2015.
  10. Ministry of Health. Alcohol use in New Zealand: key results of the 2007/08 New Zealand Alcohol and Drug Use Survey. Wellington: Ministry of Health, 2009.
  11. University of Otago and Ministry of Health. A focus on nutrition: key findings of the 2008/2009 New Zealand adult nutrition survey. Wellington: Ministry of Health, 2011.
  12. Hayes J, Richardson A, Frampton C. Population attributable risks for modifiable lifestyle factors and breast cancer in New Zealand women. Internal Medicine Journal. 2013; 43 (11):1198–1204.
  13. Bagnardi V, Rota M, Botteri E, et al. Alcohol consumption and site-specific cancer risk: a comprehensive dose–response meta-analysis. British Journal of Cancer. 2014. DOI:10.1038/bjc.2014.579
  14. Norat T, Lukanova A, Ferrari P, Riboli E. Meat consumption and colorectal cancer risk: Dose-response meta-analysis of epidemiological studies. International Journal of Cancer. 2002; 98 (2):241–256.
  15. Ma Y, Yang Y, Wang F, et al. Obesity and risk of colorectal cancer: a systematic review of prospective studies. PloS One. 2013; 8 (1):e53916.
  16. Lee IM, Shiroma EJ, Lobelo F, et al. Effect of physical inactivity on major non-communicable diseases worldwide: an analysis of burden of disease and life expectancy. Lancet. 2012; 380 (9838):219–229.
  17. Potter JD. Colorectal cancer: molecules and populations. Journal of the National Cancer Institute. 1999; 91 (11):916–932.
  18. Liang PS, Chen TY, Giovannucci E. Cigarette smoking and colorectal cancer incidence and mortality: Systematic review and meta-analysis. International Journal of Cancer. 2009; 124 (10):2406–2415.
  19. World Health Organization. Global recommendations on physical activity for health. Geneva: World Health Organization, 2010.
  20. Whiteman DC, Webb PM, Green AC, et al. Cancers in Australia in 2010 attributable to modifiable factors: summary and conclusions. Australian and New Zealand Journal of Public Health. 2015; 39 (5):477–484.
  21. Parkin DM, Boyd L, Walker L. The fraction of cancer attributable to lifestyle and environmental factors in the UK in 2010. British Journal of Cancer. 2011; 105:S77–S81.
  22. Poole C. A history of the population attributable fraction and related measures. Annals of Epidemiology. 2015; 25 (3):147–154.
  23. Rowe AK, Powell KE, Flanders WD. Why population attributable fractions can sum to more than one. American Journal of Preventive Medicine. 2004; 26 (3):243–249.
  24. Bruzzi P, Green SB, Byar DP, et al. Estimating the population attributable risk for multiple risk factors using case-control data. American Journal of Epidemiology. 1985; 122 (5):904–914.