704The preventable future burden of cancer in Australia

Abstract

Background Estimates of the future burden of invasive cancer attributable to current modifiable causal exposures can guide cancer prevention. Methods We linked pooled data from seven Australian cohort studies (N = 367,058) to national cancer and death registries, and estimated exposure-cancer and exposure-death associations using adjusted proportional hazards models. We estimated exposure prevalence from contemporary national health surveys and calculated population attributable fractions (PAFs) and 95% confidence intervals, using advanced methods accounting for competing risk of death. Results Current levels of past and current smoking explain 36.1% (95%CI 33.2%-38.9%), body fatness 13.6% (10.9%-16.2%) and alcohol consumption exceeding two drinks/day 2.3% (1.0%-3.6%) of cancers causally related to these exposures, corresponding to 210,000, 81,300 and 14,800 cancers in Australia in the next 10 years, respectively. Ever smoking is the leading modifiable cause of lung (82.1%), bladder (49.8%), oesophageal (42.8%), liver (39.8%), head and neck (35.6%), and pancreatic (21.3%) cancer burden. Body fatness is the leading modifiable cause of corpus uteri (42.5%), gastric cardia (33.6%), renal cell (29.1%), thyroid (20.1%), colorectal (12.6%) and postmenopausal breast (12.6%) cancer burden. The absolute numbers of cancers in the next 10 years attributable to smoking are highest for lung cancer (114,000). The numbers of cancers attributable to body fatness and alcohol are highest for colorectal cancer (23,000 and 9,900, respectively). Conclusions More reliable advanced methods demonstrate large proportions and numbers of cancers are preventable by modifying behaviours. Key messages Ever smoking and body fatness are the leading causes of preventable future burden of causally related cancers in Australia.

The future excess fraction of cancer due to lifestyle factors in Australia

“…Recent studies in Australia have also estimated the future burden at a number of specific cancer sites over 10 years (2017-2026). Using a modified PAF method, Laaksonen and colleagues estimated that 19% of colorectal cancers [55], 31-63% of lung cancers [15], 19% of premenopausal breast cancers, 24% of postmenopausal breast cancers [56], 33% of kidney cancers, and 24% of bladder cancers were attributable to modifiable factors [16]. Our estimates accord closely to many of these findings (36% lung, 25% postmenopausal breast, 28% kidney, 22% bladder); however, we estimated a much higher attributable burden for colorectal cancer (50%).…”

“…In addition, there may be differences in cancer incidence rates, levels of exposure, and risk factors examined. For example, while Laaksonen and colleagues estimated the future burden of lung cancer due to fruit consumption and physical inactivity [15], we excluded these risk factors as the updated WCRF evidence shows only a suggestive relationship [60]. We also excluded risk factor/cancer combinations where a protective effect has been found and the risk factor could not be conceptualised in terms of insufficient exposure (for example, oestrogen-only MHT and cancer of the colon and rectum [29]; OCP use and ovarian and endometrial cancers) [61].…”

“…In Australia, variations on the PAF method have been used to predict future cancers attributable to lifestyle exposures. For example, Laaksonen and colleagues used pooled cohort data to estimate the PAF of various cancers due to modifiable factors over 10 years (from 2017-2026) [14][15][16]. They found that smoking, overweight and obesity, and alcohol consumption contributed the highest number of cancers during that period [17].…”

“…For example, Laaksonen and colleagues used pooled cohort data to estimate the PAF of various cancers due to modifiable factors over 10 years (from 2017-2026) [14][15][16]. They found that smoking, overweight and obesity, and alcohol consumption contributed the highest number of cancers during that period [17]. In an alternative approach, Wilson and colleagues used a dynamic simulation model incorporating latency estimates to predict future cancers attributable to lifestyle factors over a 25 year period (2013-2037) [18,19].…”