Introduction

Regular physical activity (PA) has been shown to reduce the risk of gastric cancer in observational studies. Previous meta-analyses suggest that people with a highly active lifestyle would benefit from a 10%-20% reduction in stomach cancer risk as compared to their least active counterparts [1–4]. Variations in risk estimates have been reported by sex (largest effect in women) [1–3], body mass index (weaker association at higher BMI) [2], smoking (weaker protection in smokers) [2], or tumour location [1–3,5]. However, these results are based on a limited number of observational studies, and the evidence to claim a preventive effect for PA against stomach cancer is still judged as insufficient [5–8].

Overall gastric cancer incidence has been decreasing for decades [9–11]. However, tumours of the cardia region are on a steady rise [11], and stomach cancer still represents the fifth most common cancer worldwide and the third leading cause of cancer-related mortality [12]. Furthermore, survival of gastric cancer patients is poor [13,14]. It is thus important to improve our knowledge of factors, such as PA, that may be effective in the primary prevention of the disease. Guidelines for cancer prevention [6] recommend doing a minimum of 150 minutes/week of moderate-intensity PA (or 75 minutes/week of vigorous-intensity PA) for a physically active lifestyle. However, it is yet unclear whether such recommended levels of PA could be effective in the prevention of gastric cancer, and the importance of domain-specific PA effects on health at different combinations of frequency and intensity should be further investigated [15].

The aim of this study was to gain insight into the association of domain-specific physical activity and gastric cancer risk by sex, tumour site, and histological type, in the MCC-Spain case-control study.

Subjects and methods

Results

Median age at diagnosis was higher in cases than controls (68.8 vs. 65.8) (Table 1). Cases were predominantly men, had lower socio-economic status, were more likely overweight, and more frequently reported a family history of gastric cancer, and lower use of non-steroideal anti-inflammatory medication. Since men were over-represented among the cases, daily intake of energy, alcohol, and red meat were accordingly higher in this group (p < 0.001). For physical activity variables, significant differences by case status were found when comparing the proportion of participants regularly engaging in household (58.9% of cases vs. 76% of controls) or recreational activities (47.9% of cases vs. 60.1% of controls). However, median MET·min/week values did not differ between cases and controls who reported any amount of activity in these domains. Finally, no differences were found in sitting time between the two groups with a median of 5 hours/day of sitting time in both cases and controls.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive characteristics of gastric adenocarcinoma cases and controls.

MCC-Spain study.

https://doi.org/10.1371/journal.pone.0179731.t001

Multivariate logistic regression models of total GAC by physical activity domains showed highly significant inverse associations for household and recreational PA categories (Table 2). For the household domain, higher levels of PA were associated with lower OR of GAC by up to one half (OR = 0.50, 95% CI: 0.38–0.66; model 3), as compared to the lowest activity group ('none' for men, '<1000 MET·min/week' for women). Furthermore, participants who adhered to the recommended 500–1000 MET·min/week showed 35% lower risk of GAC (OR = 0.65, 95% CI: 0.43–0.98; model 3), as compared to those not engaging in RPA. On the other hand, associations were null for occupational activity and sedentary time. The significant results found were independent of all potential confounders considered (model 3), and each PA domain was further independent of all other domains evaluated (model 4). The strength of the association with RPA was slightly attenuated after accounting for BMI and dietary variables but remained, nonetheless, highly significant.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Odds ratio (OR) and 95% confidence interval (CI) of gastric adenocarcinoma by levels of physical activity (PA) variables.

MCC-Spain study.

https://doi.org/10.1371/journal.pone.0179731.t002

By site and histology, household PA (HPA) was found to be inversely associated with GAC irrespective of tumour location, but mainly of intestinal type (Table 3). On the other hand, RPA was more strongly associated with non-cardia GAC. The relationship of different variables for the household and recreational domains according to pre-specified levels of PA is presented in Table 4. Significant inverse linear trends were found for HPA or recreational walking (at any pace). RPA of moderate intensity was associated with lower GAC risk below the recommended 500 MET·min/week, whereas high-intensity RPA showed no relationship with to GAC risk overall. Finally, we tested for the combined effect of a sedentary behaviour (≥ 6 h/d sitting during leisure time) and adherence to PA recommendations (Table 5). Results revealed that recommended amounts of RPA were effectively associated with lower GAC risk, but only among non-sedentary participants. There was no evidence of effect modification by socio-demographic factors, smoking, BMI categories or H. pylori seropositivity. Although the protective association was mainly observed among men (Table 6), interactions with sex were not significant for any physical activity domain (p > 0.10).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Odds ratio (OR) and 95% confidence interval (CI) of gastric adenocarcinoma, by site and type, according to levles of physical activity (PA) variables.

MCC-Spain study.

https://doi.org/10.1371/journal.pone.0179731.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Odds ratio (OR) and 95% confidence interval (CI) of gastric adenocarcinoma, by intensity of household and recreational physical activity variables.

MCC-Spain study.

https://doi.org/10.1371/journal.pone.0179731.t004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Odds ratio (OR) and 95% confidence interval (CI) of gastric adenocarcinoma by combined levels of sedentary behaviour and recreational physical activity (RPA) recommendations.

MCC-Spain study.

https://doi.org/10.1371/journal.pone.0179731.t005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Odds ratio (OR) and 95% confidence interval (CI) of gastric adenocarcinoma by sex, according to levels of physical activity (PA) variables.

MCC-Spain study.

https://doi.org/10.1371/journal.pone.0179731.t006

Discussion

Household and recreational physical activities were strongly associated with a lower risk of GAC in the MCC-Spain study. The associations found were independent of a wide set of potential confounders and other physical activity domains. In stratified analysis, associations were highly consistent for non-cardia tumours, and mainly among men, although interactions with sex were not significant. On the other hand, sedentary time was not found to be a risk factor for gastric cancer in this study.

The promotion of physical activity is acknowledged as one of the hallmarks of chronic disease prevention, and an important contributor to wellbeing and healthy ageing [6,7,21]. In order to achieve such goals, guidelines recommend spending at least 150 min/week in activities of moderate intensity (or equivalently 75 min of high-intensity PA). Expressed as volume, this would represent 500–1000 MET·minutes/week of moderate-intensity PA (or an equivalent volume of high-intensity activities), which is advised to adults and older people with no limiting conditions [20,21]. While guidelines encourage people to remain active in all domains of everyday life, many times practical indications or specific recommendations to the population are formulated in terms of recreational outdoor activities. However, in the epidemiological literature there is also support for a protective effect of non-recreational PA as determinant of chronic disease risk. Our results revealed a strong association for household PA, with the lowest GAC risks among all PA variables observed for household PA (50% lower risk for the 'highest' versus 'lowest' group comparison). The strong inverse association of HPA with chronic disease adds to the accruing epidemiological evidence showing decreased risks of cancer and overall mortality at increasing levels of HPA[22–24]. A previous meta-analysis [23] found that overall risk of mortality was reduced by 36% (RR = 0.64, 95% CI: 0.55–0.75; p = 0.039) when comparing the highest and lowest groups of activities of daily living (household plus walking and cycling). In a more recent dose-response meta-analysis of HPA and cancer risk, Shi et al. [22] described an inverse linear relationship, with an estimated RR = 0.84 (95% CI: 0.76, 0.93) for the highest vs. lowest comparison, and 2% risk reduction per additional 10 MET·h/week. Shi et al.'s meta-analysis provides suggestive evidence for a protective effect of HPA in cancer risk. However, such association was mainly dominated by studies on PA and breast cancer. Unfortunately, only two studies on gastric cancer were available for meta-analysis (1 cohort [25] and 1 case-control [26]), both reporting null results. However, the first one did not estimate separated effects for HPA, whereas the assessment of HPA in Wen et al.'s paper was not quantitative (furthermore, PA was not the main exposure in the study and multivariate models lacked important confounders such as education, H. pylori infection, or red and processed meat intake). Thus, this is the first study to provide specific quantitative estimates of the association of HPA of different intensities and gastric cancer risk, with extensive adjustment for relevant confounders (demographics, diet, lifestyles, family history, medication, or H. pylori serostatus). Further studies are needed to corroborate our results.

Household PA comprises a plethora of indoor and outdoor activities of light, moderate, or intense effort that actually represents the largest PA domain for different population groups. In the present study, HPA accounted for 83% of total (non-occupational) activity volume, on average, among women, and for more than half total non-occupational activity in men below 60 years old. Such heterogeneity in HPA allowed for more powerful comparisions across exposure categories which could partly explain the larger effect estimates found for HPA as compared to recreational activities. As expected, men and women had very different distributions of HPA (only 3% of women reported no HPA at all vs. 43% of men). Thus, sex-specific tertiles were defined to preclude a severe imbalance of the sex ratio among HPA groups. However, the 'low HPA' group in women covered a PA range (< 1000 MET·minutes/week) for which effects in men were significant and strong (Table 6). The lack of an appropriate reference group of women not engaging in HPA, and the reduced statistical power due to the fewer number of cases, would mostly explain the null associations found in this group, as very few women had HPA levels low enough as to allow for powered statistical comparisons. Nevertheless, since there was no statistical evidence of effect modification by sex (p for sex interaction = 0.306), no specific conclusions should be drawn from sex-specific analyses. Rather, the accumulation of data supporting a protective effect of HPA on cancer and chronic disease risk in cross-sectional and longitudinal studies, advocates not to neglect the household dimension of physical activity either when conducting epidemiological research or when translating study results into recommendations or interventions for the population.

Associations of recreational PA with GAC were highly significant within the recommended 500–1000 MET·min/week range, with an estimated OR (95% CI) of 0.65 (0.43, 0.98). It is noteworthy that moderate RPA (but not walking or high-intensity activities) was associated with lower GAC risk even at levels below such recommendation (OR = 0.54, 95% CI: 0.34, 0.85). Previous meta-analyses of case-control studies found slightly lower OR for sufficient vs. insufficient (OR = 0.78, 95% CI: 0.66, 0.91) [2] or high vs. low levels of RPA (OR from 0.75 to 0.84) [1,3,4], while pooled estimates from cohort studies were of lower magnitude (OR from 0.81 to 0.87) [1–4]. Our results suggest that prevention of GAC through RPA should mainly rely on the promotion of moderate-intensity activities (such as walking, bicycling, swimming, or home exercise) rather than on RPA of higher intensity, for which we found null results. Of note, Behrens et al. [1] described a curvilinear association for RPA and gastroesophageal cancer with the lowest risk at 5 times/week of moderate-to-high RPA, and non-significant risk reductions at increased frequency. Thus, while our data support moderating the intensity of RPA to get the maximal benefit, others have suggested moderating the frequency of PA for optimal prevention of gastric cancer.

The question as to whether PA effects would differ by tumour site or histological type of GAC remains elusive. Our data support previous results from large-scale prospective studies [25,27] showing larger effects of RPA on distal (non-cardia) tumours. However, a recent report by Moore et al. [5] using pooled prospective data from 1.44 million participants showed no significant association overall for non-cardia gastric cancer (HR = 0.92, 95% CI: 0.81, 1.06), whereas the significant association reported for cardia tumours was shown to be largely mediated by BMI. It is important to note that despite the large number of subjects, standardised analytical methodology, and prospective nature of data included in Moore's dataset, the number of studies used for estimation of GC risk was low (n = 7), and thus further research is warranted to be able to draw more definitive conclusions on this topic. Even fewer investigations have considered a potential heterogeneity by histology of GAC. We have found the effect of HPA to be only significantly associated with tumours of intestinal type, while for recreational PA, similarly borderline associations were found for intestinal or diffuse types among participants in the upper vs. lower PA categories (Table 3). The paper by Moore et al. did not report on the effect of RPA on histological subtypes of gastric cancer since papers addressing the study of such association are very scarce. In a previous paper with prospective data from the EPIC study [25], no significant associations in GAC risk by histological sub-types were reported, although the number of cases was also low for sub-group analyses, limiting the statistical power. To our knowledge, no other epidemiological study has addressed this association. Due to the scarcity of data, further studies with a sufficient number of cases are needed that analyse the specific effect of PA by histological sub-types of GAC.

Sitting time was not associated with GAC overall or in sub-group analyses by site or histology. Associations were null also when considering working days and weekends separately (data not shown). There is only one previous paper, from the NIH-AARP Diet and Health Study [27], that looked at sedentary time in relation to gastric cancer risk, which reported similar null findings either for television watching or daily sitting time. A sedentary behaviour has been estimated to account for 10% of colon cancers and 10% of breast cancers globally [28], but the evidence so far does not support an independent effect on stomach cancer risk. Nevertheless, cross-classification of subjects according to sedentary time and level of RPA in the present study revealed that RPA was only associated with GAC among non-sedentary participants (Table 5). This finding suggests that physical inactivity may be able to counterbalance the benefits of physical activity on chronic disease risk (as supported by previous data on HPA and premature mortality [29]), and reinforces the message that active lifestyles cannot be defined based solely on the amount of PA, but also on limiting sedentary behaviours.

The present study has some limitations. Case-control studies are prone to recall bias and potential reverse causation. In the present study, cases were recruited as close to the date of diagnosis as possible, with over 76% of GAC cases interviewed within three months of cancer diagnosis. Since pre-diagnostic conditions might affect exposure assessment, data were collected excluding recent exposure whenever possible. Thus, recreational physical activity was assessed using an open-ended questionnaire that recorded age at starting (and quitting) of every activity reported, which allowed to exclude exposures within the last year previous to study entry. Diet, smoking status and estimated BMI were also defined for the previous year, and alcohol intake was referred to consumption in the 30–40 year-old period (over 97% of cases were diagnosed after the age of 40 years). However, reverse causation cannot be totally discarded for HPA or sitting time variables, meant to reflect the habitual engagement in such activities on a mid-term basis, but that could have been partly affected by pre-diagnostic conditions among the cases. Nevertheless, it is unlikely that reverse causation is a major explanation for our findings, since there is increasing evidence from prospective studies with long follow-up (thus, free of reverse causation bias) supporting a protective role for household activities against risk of chronic disease and mortality. Another concern is the scarcity of cases among certain strata of tumour location (e.g. cardia) or histology (e.g. diffuse type) that may have limited the statistical power to detect potential associations. Confounding (either residual or unmeasured) is always an issue in the analysis of observational data. Because of the multicase-control design of the study, controls were matched to the whole pool of cancer cases, and the separate analysis of individual cancer locations (e.g. stomach) limits the efficiency of the matching procedure. Thus, all models included matching factors (age, sex, and region) as covariates. Furthermore, the final models included the most relevant potential confounders, plus other important factors, such as H. pylori serostatus, not available to adjust for in most previous epidemiological studies. Unfortunately, the sensitive approach used to define H. pylori seropositivity did not allow us to stratify results on H. pylori serostatus, and stratified results are only presented for positive cases (Table 6). Finally, selection bias was addressed by aiming to recruit all cancer cases diagnosed within the study period in the selected areas, and the use of population controls selected from general practitioners lists. However, the non-universal sampling frame of controls and the low response rate (57% for gastric cancer patients) [16] may account for some degree of selection bias, which would limit the generalizability of our results.

Strengths of the study include the availability of data on gastric cancer site and histology to allow for sub-group analyses, the large set of confounders available to adjust for, including H. pylori infection, and the exhaustive, open-ended, PA questionnaire that included specific questions on the main PA domains (work, household, and recreational) and sedentary time.

Conclusion

Our results support a protective role for PA on gastric cancer risk, particularly non-cardia tumours, with the largest effects found for household PA. Recommended levels of recreational activity were significantly associated with lower GAC risk, but only among non-sedentary participants. Our data advice promoting PA in all domains of everyday life, not neglecting household activities, and concomitantly limiting sitting time. Such recommendations, as formulated in the 4th European Code Against Cancer [6], should be the framework for effective public health action.

Supporting information

Acknowledgments

The study was partially funded by the ‘Acción Transversal del Cancer’, approved by the Spanish Ministry Council on 11 October 2007, by the Instituto de Salud Carlos III-FEDER (PI08/1770, PI08/1359, PI09/00773, PI09/01286, PI09/01903, PI09/02078, PI09/01662, PI11/01403, PI11/01889, PI12/00265), the Fundación Marqués de Valdecilla (API 10/09), the Junta de Castilla y León (LE22A10–2), the Consejería de Salud of the Junta de Andalucía (2009-S0143), the Conselleria de Sanitat of the Generalitat Valenciana (AP_061/10), the Recercaixa (2010ACUP 00310), the Regional Government of the Basque Country, the Consejería de Sanidad de la Región de Murcia, the European Commission grants FOOD-CT-2006–036224-HIWATE, the Spanish Association Against Cancer (AECC) Scientific Foundation, the Catalan Government DURSI grant 2014SGR647, the Fundación Caja de Ahorros de Asturias, and the University of Oviedo.