Skip to content

Advertisement

  • Research article
  • Open Access
  • Open Peer Review

Physical activity, obesity and sedentary behaviour and the risks of colon and rectal cancers in the 45 and up study

  • 1, 2Email author,
  • 1, 2, 3,
  • 1,
  • 2 and
  • 2
BMC Public HealthBMC series – open, inclusive and trusted201818:325

https://doi.org/10.1186/s12889-018-5225-z

  • Received: 8 January 2018
  • Accepted: 27 February 2018
  • Published:
Open Peer Review reports

Abstract

Background

Obesity and physical activity (PA) are predictors of colon (CC) and rectal (RC) cancers. Prolonged sitting is also emerging as a potential predictor for these cancers. Little knowledge exists about the interactive effects of obesity, PA and prolonged sitting on cancer risk. This analysis assessed independent and interactive effects of PA, body mass index (BMI) and sitting time on CC and RC risks.

Methods

This analysis used data from a prospective study of 226,584 participants aged 45 years and over in New South Wales (NSW), Australia, who joined the 45 and Up study between 2006 and 2009. Baseline data were linked with data relating to mortality, cancer registration, hospital admission and Department of Human Services to December 2010. Multivariable Cox regression was used to estimate adjusted hazard ratios (referred to as relative risks, RRs) and 95% confidence intervals (Cis). Statistical significance was defined as p < 0.05.

Results

There were 846 and 369 ascertained cases of CC and RC. BMI was positively associated with CC risk (p = 0.003, P-trend = 0.0006) but not with RC. CC risk was increased in participants in the highest BMI quartile (≥29.4-≤50 kg/m2) compared to the lowest (15- < 23.6 kg/m2), (RR = 1.32, 95% CI:1.08–1.63). PA was associated with CC risk (p = 0.02) but not with RC. Specifically, CC risk was lower in individuals partaking in any amount of vigorous activity (time/week) compared to participants with no engagement (RR = 0.78, 95% CI:0.65–0.93). Sitting time was not associated with CC or RC. We found no evidence of interactive effects of PA, BMI and prolonged sitting on cancer risk.

Conclusion

This evidence suggests that a healthy weight and vigorous activity are essential to reduce CC risk since these factors may be independent of each other.

Keywords

  • Physical activity
  • Obesity
  • Sedentary behaviour
  • Colorectal cancer
  • Prospective study

Background

Colorectal cancer (CRC), defined as cancer of the colon or rectum, is the third most common cancer globally, accounting for almost 1.4 million new cases annually. Considerable variation in its incidence is observed across world regions [1]; the highest reported incidence occurred in Australia and New Zealand (age-standardised rates of 44.8 and 32.2 per 100,000 in men and women respectively) and the lowest in Western Africa (4.5 and 3.8 per 100,000) in 2012 [1]. The burden of CRC increases with the embracement of unhealthy lifestyle choices. Therefore, a substantial number of cases could be preventable by changing these behaviours [2].

High body fatness, defined as having a body mass index (BMI) greater than 25 kg/m2, has been classified as a confirmed predictor associated with increased risk of colon and rectal cancers [3, 4]. In Australia, the prevalence of overweight (BMI 25 to 29.9 kg/m2) and obesity (BMI ≥30 kg/m2) has increased over time from 56% in 1995 to 63% in 2015, being higher in men (70%) than in women (56%) [5]. Conversely, physical activity (PA) has also been established as a predictor associated with a reduced risk of colon cancer (CC), but its association with rectal cancer (RC) remains inconclusive [3]. Physical inactivity, defined as insufficient levels to meet physical activity guidelines http://www.who.int/mediacentre/factsheets/fs385/en/, was estimated to be around 44.5% amongst the adult Australian population during the period 2014–15 [5]. Despite the well-known association of PA with CC, the dose of activity required to diminish risk has not been determined. Australia’s Physical Activity Guidelines for adults currently recommend at least 150 min of moderate activity or 75 min of vigorous activity weekly http://www.health.gov.au/internet/main/publishing.nsf/Content/health-pubhlth-strateg-phys-act-guidelines.

Moreover, interactions between these specific risk factors and cancer risk have been less studied. [6] Some epidemiological evidence suggest that PA has the potential to counteract the detrimental effects of obesity; in those studies, physically active individuals had a reduced risk of cardio-metabolic outcomes regardless of BMI [7]. Two population based case-control studies observed that commuting or leisure time physical activity (LTPA) significantly altered the risk related to a high body fatness on CC although neither study assessed adherence to PA guidelines in negating this effect on cancer risk [8, 9].

Sedentary behavior is not synonymous with physical inactivity [10]. It mostly represents prolonged sitting which is ubiquitous in present-day societies, particularly in high income countries [11]. In recent years, this behaviour has emerged as an additional potential risk factor associated with adverse cardio-metabolic profile, premature mortality and various types of cancer including CRC [12]. The increased cancer risk associated with sedentary behaviour has been found to be independent of PA where the deleterious health consequences of too much sitting persist even after adjusting for the possible confounding effects of PA [10].

The current analysis examined the independent effects of BMI, as a proxy for body fatness, intensity of PA according to guidelines and sitting time on CC and RC risks. We also assessed the interactions between (i) BMI and PA, (ii) sitting time and PA; and (iii) BMI and sitting time, on the risks of developing CC and RC in The Sax Institute’s 45 and Up Study.

Methods

Study design, setting and subjects

The Sax Institute’s 45 and Up Study is a prospective, population-based cohort study in the state of New South Wales (NSW), Australia. The study was established to investigate different relationships between a wide range of exposures and health outcomes in the ageing population; details of the study design, sampling method and baseline data collection have been published elsewhere [13]. Eligible participants were randomly sampled from the general population of NSW through the Department of Human Services (formerly Medicare Australia) enrolment database, which provides near complete coverage of the population. A sex-specific baseline questionnaire was mailed to potential study participants who joined the study by completing the questionnaire and signing a consent for routine linkage of their health records to administrative databases. 267,014 men and women aged 45 years and over were recruited between January 2006 and December 2009 [13]. For the purpose of this analysis, we used 45 and Up baseline data and record linkage data from the NSW Cancer Registry (NSWCR), Admitted Patient Data Collection (APDC) which are records of patients’ services provided by hospitals in NSW, Medicare Benefits Schedule (MBS) a list of the Medicare services subsidised by the Australian government, Pharmaceutical Benefits Scheme (PBS) and Registry of Birth Deaths Marriages (RBDM). The 45 and Up Study was approved by the University of New South Wales Human Research Ethics Committee and this analysis was approved by the New South Wales Population and Health Services Research Ethics Committee (HREC/14/CIPHS/54). The use of MBS and PBS data was approved by the Department of Health and Ageing Departmental Ethics Committee.

Identification of cases

Linkage of the 45 and Up cohort data to the MBS and PBS data was conducted by the Sax Institute, using a unique identifier that was supplied to the Department of Human Services for the acquisition of the respective data. Incident cases of CC and RC and dates of diagnoses were obtained through probabilistic linkage from the NSWCR by the New South Wales Centre for Health Record Linkage (CHeRel) for all cancer registrations until the 31st of December 2010. Notification of new cancer cases is required under the Public Health Act 2010 by pathology laboratories, hospitals, radiotherapy and medical oncology departments. The International Classification of Diseases for Oncology 3rd edition was used to classify incident cases of CC (C18) and RC which included cancers of the rectosigmoid junction (C19–20) [14].

Data collection

Baseline questionnaire collected self-reported information on age, height, weight, educational attainment, country of birth, medical history, parental history of cancer and personal health behaviours, including: PA, weekly alcohol intake, smoking status, diet and daily time spent sitting.

Exposure variables

Assessment of body mass index

Body mass index, expressed in kg/m2, was derived from self-reported weigh (kg) and height (m) at baseline. These questions were phrased as “how tall are you without shoes” and “about how much do you weight”. Consistent with established methods, participants who reported extreme values for height being shorter than 100 cm, taller than 240 cm, weighing ≤35 kg or ≥270 kg, or with a calculated BMI of < 15 kg/m2 were excluded from the analysis due to the increased probability of measurement error [15]. All remaining participants were categorised according to baseline BMI quartiles since BMI risk may be more fine grained that the broad World Health Organization (WHO) categories [15]; we examined actual distribution of BMI in the sample as: 15 to < 23.6 (reference), ≥23.6 to < 26.2, ≥26.2 to < 29.4 and ≥29.4 to 50 kg/m2. Participants, for whom BMI was not possible to be estimated due to missing values of weight or height, were categorised as “unknown”.

Assessment of physical activity

In the baseline questionnaire, PA was measured with items from the Active Australia Survey (AAS) [16], which has been shown to have acceptable reliability and validity [17]. All participants were asked the weekly frequency and time spent on (i) walking, (ii) moderate and (iii) vigorous activity that lasted at least 10 min. Moderate activity included activities such as gentle swimming, social tennis, vigorous gardening or work around the house; while vigorous activity comprised activities that made participants breathe harder or puff and pant such as jogging, cycling, aerobics, competitive tennis, but not household chores or gardening. Time spent on walking and moderate activity were combined as the former is a form of the latter to derive groups according to PA guidelines; zero activity (reference), > 0 to < 150 and ≥150 min/week. Time spent on vigorous activity was categorised as zero activity (reference), > 0 to < 75 and ≥75 min/week. Participants who did not report their activity level were classified as “unknown”.

Assessment of sitting time

Sitting time was based on participants’ response to the question “About how many hours in each 24-hour day do you usually spend sitting”. Responses were categorised according to quartiles of the number of hours spent sitting in a 24-h period. The respective quartiles were 0 to < 3 (reference), 3 to < 5, 5 to < 8 or ≥ 8 h/day. Those participants who did not answer this question were classified as “unknown”.

Confounders

Potential confounders of CC and RC risks were selected on the basis of published evidence [3] which included: sex (female or male), birth cohorts (1920s, 1930s, 1940s, 1950s or 1960s), educational attainment (no school certificate or other qualifications, school or intermediate certificate, higher school or leaving certificate, trade-apprenticeship, certificate-diploma or university degree-higher), region of birth (Australia or overseas), smoking status (never, former-those who quit smoking 5 year prior baseline or current), weekly alcohol consumption acquired from quartiles of drinkers (0, 1–3, 4–6, 7–13 or ≥ 14 drinks/week), adherence to the Australian guidelines of fruit and vegetable defined as consuming more than 2 serves of fruit and 5 serves of vegetable a day (yes or no), weekly consumption of processed meat based on tertiles (0, 1, or ≥ 2 servings/week), quartiles of weekly intake of red meat (< 2, 2, 3 or ≥ 4 servings/week), weekly consumption of fibre derived from frequency of brown, wholemeal bread and cereal intake (< 7, 7–13, 14–20 or ≥ 21 servings/week), prevalence of diabetes mellitus (yes or no) identified by using diagnostic codes in APDC data, claims for glycosylated haemoglobin (HbA1c) in MBS data and claims for diabetes medication in PBS data (Anatomical Therapeutic Chemical Classification System (ATC) A10) as described in the validation of self-report diabetes and linked data in this cohort [18]; aspirin use (yes or no) outlined as having taken this medication for most of the last 4 weeks, history of colorectal testing (tested less than three 3 years ago, tested more than 3 years ago or never) [19] and parental history of CRC (yes or no).

Statistical analysis

Multivariable Cox regression was used to estimate adjusted hazard ratios (referred to as relative risks, RRs) and 95% confidence intervals (Cis). Statistical significance was defined as p < 0.05 for BMI, PA or sitting time, using age as the underlying time variable. Participants were censored if they died, were diagnosed with other cancers or were alive at the end of follow up period (31 of December 2010), whichever came first. RRs were estimated for both men and women combined since there was no evidence of effect modification by sex.

We examined potential two-way interactions between BMI, sitting time and PA on CC and RC risks by adding appropriate interaction terms to the models. The proportional hazard assumptions of the Cox regression models were assessed by Wald tests of covariates through log-time interactions. Sensitivity analyses were performed by excluding: (i) participants with BMI < 18.5 kg/m2 from the lowest BMI category because weight loss and therefore very low BMI may be due to an undiagnosed cancer; and (ii) removing participants who self-reported poor health status at baseline or those who were diagnosed during the first 6 months to reduce the potential impact of reverse causality. Statistical analyses were performed using SAS software, version 9.4; SAS institute Inc., Cary, NC.

Results

Of 267,014 baseline questionnaire respondents, 40,430 were excluded from these analyses due to data linkage errors (n = 58), having a prevalent cancer other than non-melanocytic skin cancer prior to recruitment (n = 24,167) or extreme values for height, weight and/or calculated BMI (n = 16,205) [20]. After exclusions, a total of 226,584 participants remained for these analyses. There were 846 incident cases of CC and 369 RC diagnosed between 2006 and December 2010. Mean follow-up was 2.7 years, ranging from 0.0 to 5.5 years. The mean age at recruitment was 62.0 years (age range 45.0 to 106.2 years).

After exclusions, 61.5% of 226,584 reported having a BMI greater than 25 kg/m2 with a higher prevalence in men than in women. Of these, 26.1% did not meet guidelines of moderate activity with women being more likely to adhere to guidelines; and 59.0% did not meet guidelines of vigorous activity with men being more likely to achieve 75 min/week. Men were more likely to spend more than 8 h a day on sedentary pursuits than women. Compared to the lowest BMI quartile, the highest quartile was, on average, younger, more socially disadvantaged, Australian born, less likely to be current smoker and more likely to report less healthy eating habits (more likely to consume processed and red meat; and a low level of dietary fibre) and to have been diagnosed as diabetic. Compared to participants who reported sitting < 3 h/day, those who sat 8+ hours/day were younger, least socially disadvantaged, did not adhere to fruit and vegetable guidelines and consumed more processed meat. Participants who adhered to guidelines of moderate PA (≥150 min/week) or vigorous activity (≥75 min/week) compared to those who did not were younger, least socially disadvantaged, less likely to be current smokers and had lower prevalence of diabetes (Tables 1 and 2).
Table 1

Characteristics of study participants according to BMI and sitting time

 

BMI Kg/m2 (%)

Sitting time hours a day (%)

Characteristic

15- < 23.6

≥23.6- < 26.2

≥26.2- < 29.4

≥29.4-≤50

Unknown

0- < 3

3- < 5

5- < 8

8+

Unknown

total

56,275

54,765

57,130

57,205

1209

31,458

60,710

64,980

52,927

16,509

Male

18,437 (33%)

27,633 (50%)

31,662 (55%)

26,372 (46%)

551 (46%)

13,357 (42.5%)

27,054 (44.6%)

29,959 (46.1%)

27,654 (52.3%)

6631 (40.2%)

Mean Age

65.3

63.6

62.4

61.1

64.9

62.7

63.6

63.6

60.6

67.3

Education

 University degree or higher

5507 (30.5%)

8255 (30.3%)

7991 (25.6%)

5196 (20.0%)

124 (22.9%)

2586 (19.7%)

5302 (19.9%)

7697 (26.0%)

10,437 (38.2%)

1051 (16.5%)

Region of birth

 Australia

12,372 (67.1%)

19,562 (70.8%)

23,510 (74.3%)

20,258 (76.8%)

357 (64.8%)

9779 (73.2%)

19,907 (73.6%)

21,901 (73.1%)

19,725 (71.3%)

4747 (71.6%)

Smoking status

 Current

3187 (17.4)

3480 (12.7%)

3959 (12.6%)

3847 (14.7%)

74 (13.5%)

1937 (14.6%)

3807 (14.2%)

3976 (13.4%)

3746 (13.6%)

1081 (16.4%)

Alcohol consumption (drinks/week)

 14 or more

4467 (24.7%)

7737 (28.4%)

9565 (30.6%)

7734 (29.8%)

131 (24.8%)

3882 (29.6%)

7917 (29.7%)

8534 (28.8%)

7703 (28.1%)

1598 (25.3%)

Adherence to fruit and vegetable guidelines

 No

14,880 (82.1%)

22,133 (81.2%)

25,475 (81.6%)

20,960 (80.7%)

456 (83.4%)

10,586 (80.4%)

21,432 (80.2%)

23,843 (80.6%)

22,956 (84.1%)

5087 (79.4%)

Consumption of processed meat (servings/week)

 2 or more

6412 (34.8%)

10,220 (37.0%)

12,952 (40.9%)

12,144 (46.1%)

217 (39.4%)

5105 (38.2%)

10,722 (39.6%)

12,239 (40.9%)

11,569 (41.8%)

2310 (34.8%)

Intake of red meat (servings/week)

 4 or more

7162 (40.3%)

11,248 (41.9%)

13,770 (44.7%)

12,164 (47.4%)

227 (42.9%)

5549 (42.8%)

11,611 (44.1%)

13,062 (44.7%)

11,713 (43.4%)

2636 (43.2%)

Consumption of fibre (servings/week)

 21 or more

6369 (35.7%)

8818 (32.7%)

8894 (28.8%)

6911 (26.9%)

145 (27.1%)

3902 (30.0%)

8440 (32.0%)

9331 (31.8%)

7700 (28.5%)

1764 (28.7%)

Diabetes Mellitus

 Yes

1546 (8.4%)

2681 (9.7%)

3993 (12.6%)

5956 (22.6%)

134 (24.3%)

1694 (12.7%)

3627 (13.4%)

4141 (13.8%)

3578 (12.9%)

1270 (19.2%)

Aspirin Use

 Yes

4047 (21.9%)

6340 (22.9%)

7861 (24.8%)

7326 (27.8%)

157 (28.5%)

3151 (23.6%)

6828 (25.2%)

7732 (25.8%)

6185 (22.4%)

1835 (27.7%)

History of colorectal testing

 Never

8850 (49.3%)

12,611 (46.7%)

14,467 (46.6%)

12,878 (49.9%)

281 (52.0%)

6432 (49.2%)

12,653 (47.8%)

13,452 (45.8%)

13,296 (49.0%)

3254 (51.7%)

Parental history of CRC

 Yes

1821 (9.9%)

2800 (10.1%)

3280 (10.4%)

2647 (10.0%)

48 (8.7%)

1327 (9.9%)

2698 (10.0%)

3131 (10.5%)

2858 (10.3%)

582 (8.8%)

Table 2

Characteristics of study participants according to time spent on moderate and vigorous activity

 

Moderate (%)

Vigorous (%)

Characteristic

0

> 0–149 min

≥ 150 min

Unknown

0

> 0- < 74 min

≥ 75 min

Unknown

Total

11,727

47,304

159,715

7838

98,308

35,293

53,120

39,863

Male

5559 (47.4%)

23,621 (49.9%)

72,004 (45.1%)

3471 (44.3%)

42,526 (43.3%)

18,201 (51.6%)

26,637 (50.1%)

17,291 (43.4%)

Mean Age

64.8

62.2

62.9

65.0

64.0

60.4

59.5

68.2

Education

 University degree or higher

913 (16.9%)

6601 (28.4%)

19,053 (26.8%)

506 (15.3%)

9762 (23.3%)

5820 (32.3%)

8961 (34.0%)

2530 (15.1%)

Region of birth

 Australia

4035 (72.6%)

16,787 (71.1%)

52,792 (73.3%)

2445 (70.4%)

30,882 (72.6%)

13,472 (74.0%)

19,593 (73.6%)

12,112 (70.1%)

Smoking status

 Current

1118 (20.3%)

3498 (14.9%)

9353 (13.1%)

578 (16.8%)

6676 (15.8%)

2273 (12.6%)

3068 (11.6%)

2530 (14,7%)

Alcohol consumption (drinks/week)

 14 or more

1336 (24.6%)

5695 (24.5%)

21,750 (30.5%)

853 (26.2%)

11,703 (27.8%)

5046 (28.0%)

7873 (29.8%)

5012 (30.1%)

Adherence to fruit and vegetable guidelines

 No

4638 (85.0%)

19,632 (84.6%)

56,917 (80.0%)

2717 (81.0%)

34,749 (82.8%)

14,809 (82.5%)

20,929 (79.6%)

13,417 (79.2%)

Consumption of processed meat (servings/week)

 2 or more

2270 (40.8%)

9284 (39.3%)

29,172 (40.5%)

1219 (35.1%)

17,793 (41.8%)

7253 (39.9%)

10,333 (38.8%)

6566 (38.0%)

Intake of red meat (servings/week)

 4 or more

2389 (45.0%)

9583 (41.9%)

31,206 (44.4%)

1393 (43.3%)

18,841 (45.5%)

7709 (43.4%)

10,544 (40.5%)

7477 (45.5%)

Consumption of fibre (servings/week)

 21 or more

1227 (22.9%)

5872 (25.6%)

23,162 (32.9%)

876 (27.3%)

12,375 (29.8%)

5270 (29.6%)

8361 (32.0%)

5131 (31.2%)

Diabetes Mellitus

 Yes

1192 (21.4%)

3572 (15.1%)

9009 (12.5%)

537 (15.5%)

6973 (16.4%)

1878 (10.3%)

2358 (8.9%)

3101 (17.9%)

Aspirin Use

 Yes

1507 (27.1%)

5632 (23.8%)

17,795 (24.7%)

797 (23.0%)

11,351 (26.7%)

3944 (21.7%)

5223 (19.6%)

5213 (30.2%)

History of colorectal testing

 Never

2974 (54.9%)

11,718 (50.9%)

32,586 (46.1%)

1809 (54.8%)

20,141 (48.3%)

8489 (47.5%)

12,515 (47.8%)

7942 (47.5%)

Parental history of CRC

 Yes

524 (9.4%)

2328 (9.8%)

7460 (10.4%)

284 (8.2%)

4171 (9.8%)

1964 (10.8%)

2893 (10.9%)

1568 (9.1%)

Increasing BMI was associated with increased CC risk (p = 0.003 and p-trend = 0.0006) (Table 3); participants with a BMI ≥29.4-≤50 kg/m2 had a 32% increased CC risk compared to participants with a BMI 15- < 23.6 kg/m2 (RR = 1.32, 95% CI: 1.08–1.63). Although diabetes might be an intermediate outcome on the causal pathway between BMI and disease risk [21], further adjustments with diabetes had little influence on the final outcome, thus it was not included in these analyses (data not shown). Sensitivity analysis excluding underweight participants from the lowest BMI group did not appreciably change any effects of relative risk of neither CC nor RC.
Table 3

Cox proportional hazards of incident CC according to BMI, sedentary behaviour and types of physical activity

Variable

Events

No person-years

Age-adjusted HR

Multivariable HR(1)

BMI Kg/m2

 15- < 23.6

195

154,782

1.00

1.00

  ≥ 23.6- < 26.2

188

150,904

0.99 (0.81–1.21)

0.90 (0.73–1.12)

  ≥ 26.2- < 29.4

223

156,854

1.13 (0.93–1.37)

1.12 (0.91–1.38)

  ≥ 29.4-≤50

232

156,109

1.18 (0.98–1.43)

1.32 (1.08–1.63)

 Unknown

8

3273

1.94 (0.96–3.94)

2.21 (1.09–4.49)

P-value(a)

   

0.003

 P-trend(b)

   

0.0006

Sitting hours/day

 0- < 3

105

87,149

1.00

1.00

 3- < 5

228

167,892

1.13 (0.90–1.42)

0.98 (0.77–1.25)

 5- < 8

267

178,262

1.24 (0.99–1.56)

1.12 (0.88–1.42)

 8+

165

143,987

0.95 (0.75–1.22)

1.02 (0.79–1.32)

 Unknown

81

44,633

1.51 (1.13–2.02)

0.92 (0.67–1.28)

P-value(a)

   

0.55

 P-trend(b)

   

0.88

Moderate activity minutes/week

 None

52

30,194

1.00

1.00

  > 0–149

175

137,275

0.73 (0.54–1.00)

1.08 (0.76–1.54)

 150 or more

583

433,966

0.78 (0.58–1.03)

1.25 (0.90–1.74)

 Unknown

36

20,487

1.01 (0.66–1.55)

1.09 (0.65–1.82)

P-value(a)

   

0.17

 P-trend(b)

   

0.76

Vigorous activity minutes/week

None

423

270,374

1.00

1.00

Any amount:(*)

220

248,354

0.56 (0.48–0.66)

0.78 (0.65–0.93)

  > 0–74

97

102,607

0.60 (0.48–0.75)

0.78 (0.61–0.98)

 75 or more

123

145,747

0.54 (0.44–0.66)

0.78 (0.63–0.97)

 Unknown

203

103,194

1.26 (1.07–1.49)

0.98 (0.81–1.18)

P-value(a)

   

0.02

 P-trend(b)

   

0.11

(1)Multivariable models adjusted for: birth cohort, sex, education, BMI, sitting time, time spent on moderate and vigorous activity, smoking, alcohol, country of birth, guidelines of fruit and vegetables, weekly intake of processed food, red meat and fibre, aspirin, parental history of CRC and history of colorectal testing

(a)P-value’ for each variable corresponds to a test of whether all HRs = 1

(b)‘P-trend’ is for test of linear association in the log hazard scale and obtained by substituting the categorical versions of covariates in the Cox model with continuous or ordinal versions where appropriate

‘Unknown’ category was excluded from the estimation of P-values and P-trends.

(*)The “any amount” category was derived by combining the “> 0–74” and “75 or more” and the HR for “any amount” was estimated by fitting a separate multivariate model

Of the 2 types of activities, only vigorous activity was associated with CC (Table 3). Participants who engaged in any amount of vigorous activity/week had 22% lower risk of developing CC compared to participants who did not perform this type of activity. The respective RR was 0.78 95% CI 0.65–0.93 with no evidence of a dose response relationship.

Risk of CC was not associated with sitting time (p = 0.55) or moderate activity (p = 0.17) (Table 3). Additionally, RC risk was not associated with BMI (p = 0.20), sitting time (p = 0.65), moderate activity (p = 0.77) or vigorous activity (p = 0.11) (Table 4). Sensitivity analysis excluding participants who self-reported poor health status at baseline or those participants who were diagnosed within the first 6 months did not substantially change any effects of relative risks for either CC or RC risk (data not shown). No significant interactions between PA and BMI, sitting time and PA or BMI and sitting time were evident on CC or RC risks in these analyses (p-values greater than 0.10) (Figs. 1 and 2).
Table 4

Cox proportional hazards of incident rectal cancer according to BMI, sedentary behaviour and types of physical activity

Variable

Events

No person-years

Age-adjusted HR

Multivariable HR(1)

BMI Kg/m2

 15- < 23.6

93

154,782

1.00

1.00

  ≥ 23.6- < 26.2

93

150,905

1.03 (0.77–1.37)

0.84 (0.62–1.14)

  ≥ 26.2- < 29.4

87

156,855

0.92 (0.69–1.24)

0.73 (0.54–1.00)

  ≥ 29.4-≤50

93

156,110

0.99 (0.74–1.32)

0.76 (0.55–1.04)

 Unknown

n.p

3273

1.52 (0.48–4.81)

0.89 (0.22–3.62)

P-value(a)

   

0.20

 P-trend(b)

   

0.12

Sitting hours/day

 0- < 3

50

87,149

1.00

1.00

 3- < 5

100

167,893

1.04 (0.74–1.46)

1.03 (0.72–1.48)

 5- < 8

111

178,263

1.09 (0.78–1.52)

1.10 (0.77–1.57)

 8+

72

143,988

0.87 (0.61–1.25)

0.90 (0.61–1.32)

 Unknown

36

44,633

1.41 (0.92–2.16)

1.07 (0.66–1.74)

P-value(a)

   

0.65

 P-trend(b)

   

0.40

Moderate activity minutes/week

 None

20

30,195

1.00

1.00

  > 0–149

77

102,607

0.85 (0.52–1.39)

0.85 (0.51–1.43)

 150 or more

252

433,967

0.88 (0.56–1.38)

0.93 (0.57–1.51)

 Unknown

20

20,487

1.47 (0.79–2.73)

1.17 (0.57–2.38)

 P-value(a)

   

0.77

 P-trend(b)

   

0.21

Vigorous activity minutes/week

 None

163

270,375

1.00

1.00

  > 0–74

53

102,607

0.86 (0.63–1.17)

0.98 (0.71–1.36)

 75 or more

58

145,748

0.66 (0.49–0.89)

0.70 (0.50–0.98)

 Unknown

95

103,194

1.52 (1.18–1.96)

1.34 (1.01–1.79)

P-value(a)

   

0.11

 P-trend(b)

   

0.78

(1)Multivariable models adjusted for: birth cohort, sex, education, BMI, sitting time, time spent on moderate and vigorous activity, smoking, alcohol, country of birth, guidelines of fruit and vegetables, weekly intake of processed food, red meat and fibre, aspirin use, parental history of CRC and history of colorectal testing

(a)P-value’ for each variable corresponds to a test of whether all HRs = 1

(b)‘P-trend’ is for test of linear association in the log hazard scale and obtained by substituting the categorical versions of covariates in the Cox model with continuous or ordinal versions where appropriate

‘Unknown’ category was excluded from the estimation of P-values and P-trends

n.p Not publishable because of small number, confidentiality or ethical concerns about the data

Fig. 1
Fig. 1

Adjusted hazard ratios and 95% CI for the interaction between BMI-PA, Sitting Time-PA; and BMI-Sitting Time on CC Risk

Fig. 2
Fig. 2

Adjusted hazard ratios and 95% CI for the interaction between BMI-PA, Sitting Time-PA; and BMI-Sitting Time on RC Risk

Discussion

This is the first analysis to our knowledge that examined the independent and interactive effects of PA based on the Australian recommended guidelines, BMI and sitting time on CC and RC risks. In this Australian cohort, BMI and vigorous activity were independent predictors for CC risk where individuals with a BMI ≥29.4 kg/m2 were at higher risk of developing CC than those in the lowest. Also, individuals who engaged in any amount of vigorous activity were at lower risk of developing CC than those who did not partake in this activity. We found no evidence of interactions between any of the study variables assessed and CC or RC risk.

In a recent assessment of observational studies on body fatness and cancer risk, the International Agency for Research on Cancer reported a 30% increased likelihood of developing CRC in those individuals in the highest evaluated BMI compared to the lowest. [4] However, dissociation between colon and rectal cancers was not considered. Some evidence indicate that the aetiology of these two cancers may differ [14]. For instance, a recent meta-analysis of prospective studies observed a 47% increased risk of CC and only 15% for RC when contrasting the highest category of BMI against the lowest [22]. In our analysis, we found that the highest BMI group was similarly associated with an increased risk of CC but not with RC. The lack of association between BMI and RC in our analysis is concordant with a previous prospective Australian study [23].

There is convincing evidence to support a protective role of PA in CC risk. A recent meta-analysis of prospective studies reported a 23% risk reduction for CC when comparing the most active to the least active in the PA spectrum [24]. However, this protective effect of PA on CC risk was based on a combined measure of moderate to vigorous activity (MVPA) without considering the individual type of activity [2530]. This grouping makes it difficult to determine if intensity and duration are relevant for reducing cancer risk [31]. Very few observational studies have assessed the effects of different types of PA on CC risk [3235]. A prospective study among Japanese reported a significant risk reduction in CC in men who walked more than 1 h a day, but not in women [35]. Moreover, a prospective cohort documented that vigorous activity decreased CC risk in men but not in women [34]. On the other hand, we found that any amount of vigorous activity reduced CC risk irrespective of sex. This discrepancy in findings might have occurred due to the low number of female cases accrued (fewer than 50) in the studies by Takahashi et al. [35] or Lee et al. [34], while our study had a relatively large number of CC cases for both genders. Also, vigorous physical activity is different from walking as there is a much higher energy expenditure in vigorous activity than walking.

The role of PA on rectal cancer risk is less clear as no association has consistently been observed [32]. A recent meta-analysis of prospective studies reported no change in risk [26]. To our knowledge, there are only three studies that have examined this association by types of PA with RC risk [32, 34, 36]. Of them, only a population based case-control study observed a decreased risk with lifetime vigorous activity [32]. We found no evidence for a relationship between PA and RC risk.

Sedentary behaviour has been proposed as an independent risk factor in colorectal carcinogenesis [10, 37]. A recent meta-analysis of prospective studies reported a 27% increased risk for CC and 6% for RC when comparing the highest amount of sitting to the lowest [37]. However, moderate heterogeneity was observed, reflecting the variability between studies in measuring and categorising this complex behaviour. The different domains of sedentary behaviour include recreation (TV viewing and computer use), workplace sitting and commuting [12]. We did not find any association between total sitting time and the risks of CC or RC nor did we find evidence of two-way interactions with PA or BMI.

The elucidation of the interaction between BMI and PA in terms of cancer risk is of public health interest since these risk factors tend to be related, and have been shown to interact in the context of cardiovascular disease risk [7, 38]. Emerging evidence from case-control studies proposes that PA might offset CC risk related to a high BMI [8, 9]. Only four studies have reported the assessment of the interaction between BMI and PA on CC risk. Of them, two prospective studies reported no significant interactions [14, 27] while two case-control studies detected significant interactions between high levels of lifetime commuting activity or long-term vigorous activity and BMI [8, 9]. We did not observe a significant interaction between PA and BMI on either CC or RC risk. Additionally, our measurement of PA is not comparable to those used in the case-control studies as the questionnaire only assessed current and not lifetime activity. Consistent with our results, we recently reported no evidence of interaction between BMI and current PA on CRC risk in a case-control study [39]. The nature of this interaction does not depend on the scale used as presence on one scale (additive or multiplicative) will also be present on the other scale [40].

Major strengths of this analysis are the prospective nature of the study design, the large cohort sample size of 226,584 participants that provided reasonable statistical power to detect an effect of the exposure variables on cancer risk; and the linkages of the questionnaire data to deaths records, cancer registry and administrative data. A limitation is the relatively short-term follow-up (mean 2.7 years) for cancer incidence which can, depending on the outcome, lead to a low number of cases and imprecise effect estimates. In this study, however, a large number of incident cases for both CC (n = 846) and RC (n = 369) were accrued during the follow-up period. Another potential limitation is that exposure variables were ascertained by self-report. Nevertheless, BMI and PA have been validated. For instance, a subsample of the 45 and UP cohort showed a strong correlation between self-reported and measured BMI (r = 0.95) [20] and the AAS questionnaire possesses a reliability which ranges from 0.56 to 0.64 and validity estimated around 0.52 [41]. Furthermore, vigorous activity tends to be better reported by participants than other categories of recreational PA [41]. Any misclassification of the exposure variables collected before the diagnosis of cancer would most likely have resulted in attenuated estimates of effects [27]. Finally, while we did not find evidence of two-way interactions, confidence intervals for interaction variable categories within were wide, perhaps suggesting limited statistical power to detect such interactions [42].

Conclusion

This analysis supports the importance of independently adhering to vigorous guidelines of physical activity as well as achieving and maintaining a healthy BMI. Future research using other cohort studies is needed to confirm the absence of interaction between PA and BMI.

Abbreviations

AAS: 

Active Australia Survey

APDC: 

Admitted Patient Data Collection

ATC: 

Anatomical Therapeutic Chemical Classification System

BMI: 

body mass index

CC: 

colon cancer

CHeReL: 

Centre for Health Record Linkage

CRC: 

colorectal cancer

LTPA: 

leisure time physical activity

MBS: 

Medicare Benefits Schedule

NSW: 

New South Wales

NSWCR: 

NSW Cancer Registry

PA: 

physical activity

PBS: 

Pharmaceutical Benefits Scheme

RBDM: 

Registry of Birth Deaths Marriages

RC: 

rectal cancer

Declarations

Acknowledgements

This research was completed using data collected through the 45 and Up Study (www.saxinstitute.org.au). The 45 and Up Study is managed by the Sax Institute in collaboration with major partner Cancer Council NSW; and partners: The National Heart Foundation of Australia (NSW Division); NSW Ministry of Health; NSW Government Family & Community Services – Carers, Ageing and Disability Inclusion; and the Australian Red Cross Blood Service. We thank all those who participated in the 45 and Up Study. In addition, we acknowledge The NSW Ministry of Health and Cancer Institute NSW for provision of cancer data, The Department of Human Services (DHS) for supplying data from Medicare Benefits Schedule (MBS) and Pharmaceutical Benefits Scheme (PBS).

Funding

This analysis is part of Mr. Carlos Nunez’s PhD project at Cancer Council NSW with thesis to be submitted through the University of Sydney, School of Public Health. He was the recipient of the Outrun Cancer Scholarship. Outrun Cancer Corporate Treadmill Marathon is a not-for-profit organisation that encourages people to take action against cancer and adopt a healthier lifestyle. The funding agency played no role in the study design, data collection, data analysis, data interpretation or writing of the report.

Availability of data and materials

Due to confidentiality restrictions in the ethics approval of this study, de-identified information may be released. The authors were only allowed to publish results from the data due to confidentiality conditions.

Authors’ contributions

CN conducted the analysis, data interpretation, drafting of the manuscript and critically reviewed and revised the intellectual content of the manuscript. VS, SE, FS and AB participated in the interpretation of data and critically reviewed and revised the intellectual content of the manuscript. Additionally, SE provided statistical support. All of the authors have read and approved the version of the manuscript.

Ethics approval and consent to participate

The study was approved by the University of New South Wales Human Research Ethics Committee and the analysis was consented to by the New South Wales Population and Health Services Research Ethics.

Consent for publication

Not applicable

Competing interest

The authors declare that they have no competing interests.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Authors’ Affiliations

(1)
Cancer Research Division, Cancer Council NSW, 153 Dowling St, Woolloomooloo, Sydney, NSW, 2011, Australia
(2)
Sydney School of Public Health, the University of Sydney, Camperdown, Sydney, NSW, 2006, Australia
(3)
Faculty of Medicine and Health Science, Macquarie University, Sydney, NSW, 2109, Australia

References

  1. Ferlay J, Soerjomataram I, Dikshit R, Eser S, Mathers C, Rebelo M, et al. Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012. Int J Cancer. 2015;136(5):E359–E86.View ArticlePubMedGoogle Scholar
  2. Keum N, Giovannucci EL. Epidemiology of colorectal cancer. Pathology and Epidemiology of Cancer: Springer; 2017. p. 391–407. https://doi.org/10.1007/978-3-319-35153-7.
  3. World Cancer Research Fund. Continuous update project report: diet, Nutrition, Physical Activity and Colorectal Cancer: WCRF; 2017. wcrf.org/colorectal-cancer-2017.
  4. Lauby-Secretan B, Scoccianti C, Loomis D, Grosse Y, Bianchini F, Straif K. Body fatness and cancer—viewpoint of the IARC working group. N Engl J Med. 2016;375(8):794–8.View ArticlePubMedGoogle Scholar
  5. Australian Bureau of Statistics. National Health Survey First Results. Australia 2014–15. 2015. http://www.abs.gov.au/ausstats/abs@.nsf/Lookup/by%20Subject/4364.0.55.001~201415~Main%20Features~Overweight%20and%20obesity~22.
  6. Schmid D, Leitzmann MF. Physical activity epidemiology. Handbook of Epidemiology: Springer; 2014. p. 1927–2002.Google Scholar
  7. Fogelholm M. Physical activity, fitness and fatness: relations to mortality, morbidity and disease risk factors. A systematic review. Obes Rev. 2010;11(3):202–21.View ArticlePubMedGoogle Scholar
  8. Hou L, Ji B-T, Blair A, Dai Q, Gao Y-T, Chow W-H. Commuting physical activity and risk of colon cancer in shanghai, China. Am J Epidemiol. 2004;160(9):860–7.View ArticlePubMedGoogle Scholar
  9. Slattery ML, Potter JD. Physical activity and colon cancer: confounding or interaction? Med Sci Sports Exerc. 2002;34(6):913–9.View ArticlePubMedGoogle Scholar
  10. Lynch BM. Sedentary behavior and cancer: a systematic review of the literature and proposed biological mechanisms. Cancer Epidemiol Biomark Prev. 2010;19(11):2691–709.View ArticleGoogle Scholar
  11. Ekelund U, Steene-Johannessen J, Brown WJ, Fagerland MW, Owen N, Powell KE, et al. Does physical activity attenuate, or even eliminate, the detrimental association of sitting time with mortality? A harmonised meta-analysis of data from more than 1 million men and women. Lancet. 2016;388(10051):1302–10.View ArticlePubMedGoogle Scholar
  12. Schmid D, Leitzmann MF. Television viewing and time spent sedentary in relation to cancer risk: a meta-analysis. J Natl Cancer Inst. 2014;106(7)Google Scholar
  13. 45 and Up Collaborators. Cohort profile: the 45 and up study. Int J Epidemiol. 2008;37(5):941–7.View ArticleGoogle Scholar
  14. Friedenreich C, Norat T, Steindorf K, Boutron-Ruault M-C, Pischon T, Mazuir M, et al. Physical activity and risk of colon and rectal cancers: the European prospective investigation into cancer and nutrition. Cancer Epidemiol Biomark Prev. 2006;15(12):2398–407.View ArticleGoogle Scholar
  15. Joshy G, Korda RJ, Bauman A, Van Der Ploeg HP, Chey T, Banks E. Investigation of methodological factors potentially underlying the apparently paradoxical findings on body mass index and all-cause mortality. PLoS One. 2014;9(2):e88641.View ArticlePubMedPubMed CentralGoogle Scholar
  16. The Active Australia Survey: A guide and manual for implementation, analysis and reporting. Canberra: Australian Institute of Health and Welfare. 2003.Google Scholar
  17. Brown WJ, Burton NW, Marshall AL, Miller YD. Reliability and validity of a modified self-administered version of the active Australia physical activity survey in a sample of mid-age women. Aust N Z J Public Health. 2008;32(6):535–41.View ArticlePubMedGoogle Scholar
  18. Comino EJ, Tran DT, Haas M, Flack J, Jalaludin B, Jorm L, et al. Validating self-report of diabetes use by participants in the 45 and up study: a record linkage study. BMC Health Serv Res. 2013;13(1):1.View ArticleGoogle Scholar
  19. Steffen A, Weber MF, Roder DM, Banks E. Colorectal cancer screening and subsequent incidence of colorectal cancer: results from the 45 and up study. Med J Aust. 2014;201(9):523–7.View ArticlePubMedGoogle Scholar
  20. Ng SP, Korda R, Clements M, Latz I, Bauman A, Bambrick H, et al. Validity of self-reported height and weight and derived body mass index in middle-aged and elderly individuals in Australia. Aust N Z J Public Health. 2011;35(6):557–63.View ArticlePubMedGoogle Scholar
  21. Doliba NM. Pancreatic islet adaptation and failure in obesity and diabetes. Metabolic Syndrome. 2016;461. https://doi.org/10.1007/978-3-319-11251-0.
  22. Ma Y, Yang Y, Wang F, Zhang P, Shi C, Zou Y, et al. Obesity and risk of colorectal cancer: a systematic review of prospective studies. PLoS ONE [electronic Resource]. 2013;8(1):e53916.View ArticleGoogle Scholar
  23. MacInnis RJ, English DR, Haydon AM, Hopper JL, Gertig DM, Giles GG. Body size and composition and risk of rectal cancer (Australia). Cancer Causes Control. 2006;17(10):1291.View ArticlePubMedGoogle Scholar
  24. Robsahm TE, Aagnes B, Hjartåker A, Langseth H, Bray FI, Larsen IK. Body mass index, physical activity, and colorectal cancer by anatomical subsites: a systematic review and meta-analysis of cohort studies. Eur J Cancer Prev. 2013;22(6):492–505.View ArticlePubMedGoogle Scholar
  25. Chao A, Connell CJ, Jacobs EJ, McCullough ML, Patel AV, Calle EE, et al. Amount, type, and timing of recreational physical activity in relation to colon and rectal cancer in older adults: the cancer prevention study II nutrition cohort. Cancer epidemiology. Biomarkers & Prevention. 2004;13(12):2187–95.Google Scholar
  26. Giovannucci E, Ascherio A, Rimm EB, Colditz GA, Stampfer MJ, Willett WC. Physical activity, obesity, and risk for colon cancer and adenoma in men. Ann Intern Med. 1995;122(5):327–34.View ArticlePubMedGoogle Scholar
  27. Larsson SC, Rutegard J, Bergkvist L, Wolk A. Physical activity, obesity, and risk of colon and rectal cancer in a cohort of Swedish men. Eur J Cancer. 2006;42(15):2590–7.View ArticlePubMedGoogle Scholar
  28. Nilsen TI, Romundstad PR, Petersen H, Gunnell D, Vatten LJ. Recreational physical activity and cancer risk in subsites of the colon (the Nord-Trondelag health study). Cancer Epidemiol Biomark Prev. 2008;17(1):183–8.View ArticleGoogle Scholar
  29. Wolin KY, Lee IM, Colditz GA, Glynn RJ, Fuchs C, Giovannucci E. Leisure-time physical activity patterns and risk of colon cancer in women. Int J Cancer. 2007;121(12):2776–81.View ArticlePubMedPubMed CentralGoogle Scholar
  30. Martinez ME, Giovannucci E, Spiegelman D, Hunter DJ, Willett WC, Colditz GA. Leisure-time physical activity, body size, and colon cancer in women. Nurses' Health Study Research Group. J Natl Cancer Inst. 1997;89(13):948–55.View ArticlePubMedGoogle Scholar
  31. Slattery ML. Physical activity and colorectal cancer. Sports Med. 2004;34(4):239–52.View ArticlePubMedGoogle Scholar
  32. Slattery M, Edwards S, Curtin K, Ma K, Edwards R, Holubkov R, et al. Physical activity and colorectal cancer. Am J Epidemiol. 2003;158(3):214–24.View ArticlePubMedGoogle Scholar
  33. Mai PL, Sullivan-Halley J, Ursin G, Stram DO, Deapen D, Villaluna D, et al. Physical activity and colon cancer risk among women in the California teachers study. Cancer Epidemiol Biomark Prev. 2007;16(3):517–25.View ArticleGoogle Scholar
  34. Lee KJ, Inoue M, Otani T, Iwasaki M, Sasazuki S, Tsugane S, et al. Physical activity and risk of colorectal cancer in Japanese men and women: the Japan public health center-based prospective study. Cancer Causes Control. 18(2):199–209.Google Scholar
  35. Takahashi H, Kuriyama S, Tsubono Y, Nakaya N, Fujita K, Nishino Y, et al. Time spent walking and risk of colorectal cancer in Japan: the Miyagi cohort study. Eur J Cancer Prev. 2007;16(5):403–8.View ArticlePubMedGoogle Scholar
  36. Boyle T, Heyworth J, Bull F, McKerracher S, Platell C, Fritschi L. Timing and intensity of recreational physical activity and the risk of subsite-specific colorectal cancer. Cancer Causes Control. 2011;22(12):1647–58.View ArticlePubMedGoogle Scholar
  37. Cong Y, Gan Y, Sun H, Deng J, Cao S, Xu X, et al. Association of sedentary behaviour with colon and rectal cancer: a meta-analysis of observational studies. Br J Cancer. 2014;110(3):817–26.View ArticlePubMedGoogle Scholar
  38. Crump C, Sundquist J, Winkleby MA, Sundquist K. Interactive effects of physical fitness and body mass index on the risk of hypertension. JAMA Intern Med. 2016;176(2):210–6.View ArticlePubMedPubMed CentralGoogle Scholar
  39. Nunez C, Bauman A, Egger S, Sitas F, Nair-Shalliker V. Obesity, physical activity and cancer risks: results from the cancer, lifestyle and evaluation of risk study (CLEAR). Cancer Epidemiol. 2017;47:56–63.View ArticlePubMedGoogle Scholar
  40. VanderWeele TJ, Knol MJA. Tutorial on interaction. Epidemiologic Methods. 2014;3(1):33–72.Google Scholar
  41. Gebel K, Ding D, Chey T, Stamatakis E, Brown WJ, Bauman AE. Effect of moderate to vigorous physical activity on all-cause mortality in middle-aged and older Australians. JAMA Intern Med. 2015;175(6):970–7.View ArticlePubMedGoogle Scholar
  42. Pearce N, Greenland S. Confounding and interaction. Handbook of epidemiology: Springer; New York 2014. p. 659–84. https://doi.org/10.1007/978-0-387-09834-0

Copyright

© The Author(s). 2018

Advertisement