Diet and dietary change
The results above show that tomatoes and canned meat consumption seem to decrease the odds of developing breast cancer. On the contrary, beans, low sodium salt and cabbage consumption seem to increase the odds of developing cancer.
There are a number of confounding issues when identifying food items as risk or protective factors of breast cancer.
Other food items are being consumed with the item analysed, which are contributing to the risk or protective factors of breast cancer; however, these items are not included in the study.
Dietary changes followed by subjects after being diagnosed with breast cancer.
The list of item being analysed may be too generic—for instance, the study makes reference to fava beans; however, the subject may mix it up with other types of beans that may have a different effect on breast cancer.
The onset of a health condition is very often followed by a change in the food diet. For instance, concerns over high blood pressure levels may lead subjects to change regular salt intake with low sodium salts. Similarly, patients diagnosed with breast cancer may change their food diet to a healthier one. Beans and cabbages are often considered to be part of a healthy diet and normally are not associated with adverse effect on breast cancer. Our study shows that cabbages, beans and low salt intake are more likely to increase the risk of breast cancer, while canned meat and American coffee are more likely to protect against cancer. These anomalous results may be attributed to the fact that subjects diagnosed with breast cancer tend to change their food diet to a healthier one. Hence, there are larger proportions in the case group eating cabbages, beans and low salt and larger proportions in the control group taking canned meat and American coffee. At least one study has found that in premenopausal women, regular coffee drinking appears to be linked with a reduced risk of the disease .
In the case of tomato consumption, our result complements other findings in literature, which identify tomato consumption as a protective factor against breast cancer. Other studies [44, 45] found that lycopene compounds have antioxidant effects that protect against free-radical damage which is associated with an increased risk of developing cancer. In fact, tomatoes are known to be particularly rich in lycopene. Tomatoes are frequently consumed in Malta and other Mediterranean countries, and subject recall bias is less likely when they quantify tomato consumption per week.
Exposure to sunlight seems to be associated with a decrease in the risk of breast cancer. This corroborates findings found in other studies [24, 46]. It is highly possible that in both controls and cases, subjective recall bias influences the accuracy of estimated exposure to sunlight. Most people find it difficult to recall the exact period of time they spent directly exposed to sunlight, and some subjects might have different ideas as to what qualifies as “direct exposure” even if the interviewer made it as clear as possible. The time of day of sunlight exposure is another confounding factor since sunlight exposure is more intense during certain hours of the day, which may actually counteract any benefit.
Our study shows that exposure to sunlight has a positive effect on the prevention of breast cancer. This finding is also corroborated in other scientific analyses of this nature [24, 46, 47]. Though prolonged sunlight exposure damages the skin, it has been found that it reduces the risk of breast cancer—possibly due to vitamin D production and its role in breast cell growth. Therefore, one cannot discount these findings as mere artefacts or results of recall bias.
Oral contraceptive pill use
As reported in the results section, the odds of having breast cancer is 54.6 % lower in subjects not taking oral contraceptive pills compared to those who take oral contraceptive pills. This is quite a large effect, and corroborates established scientific evidence that oral contraceptive pill use increases breast cancer .
The three physiological factors emerging as highly significant were height, a history of myocardial infarction (heart attack) and menopausal status.
Height came as a much unexpected factor in this analysis, appearing as significant in virtually every model run in this study. The model selected showed a 4.8 % increase in the odds of having breast cancer for every centimetre increase in height (p value <0.001; CI = 1.021 cm, 1.077 cm). Large-scale studies carried out elsewhere which have examined height as a continuous variable also found such an association . One possible explanation could be that taller individuals have a greater number of cells, and this would statistically lead to a greater risk of a cell going awry and turning cancerous; however, few studies have been carried out linking height to cancer, and further research is necessary prior to making any conclusions. There seems to be no association between breast size and height .
The height effect on the prevalence of breast cancer is doubtful due to a number of confounding factors. Breast tissue density was not taken into account, and neither was the type of breast cancer. Both these parameters are known to be far more associated with breast cancer risk than height. Although these two effects could not be included in this study, the height effect as risk factor of breast cancer cannot be entirely disregarded.
For women with no history of heart attacks, the odds of having breast cancer was 66.9 % lower than their counterparts with a family history. This difference cannot be attributed to age differences between cases and controls, as the study cases and controls were age-matched. Literature shows that living a healthy lifestyle not only reduces the risk of cardiovascular ailments but also lowers the risk of cancers, including breast cancer.
Menopausal status (categorical)
Postmenopausal women are more likely to develop breast cancer than their younger counterparts. Logistic regression analysis shows that for premenopausal women and perimenopausal women, the odds of having breast cancer rather than not having breast cancer are, respectively, 93.3 and 38.3 % lower compared to those for postmenopausal women. Table 4 outlines the menopausal status according to age category and sorted as case-control.
It is well known that age is a strong risk factor of breast cancer and postmenopausal women tend to be older than premenopausal and perimenopausal women. Another explanation, however, can be attributed to the fact that women who start menopause at a later age tend to have had greater exposure to oestrogens in their lifetime, and therefore are at an increased risk of developing breast cancer.
Table 4 shows a smaller proportion of women in the control group (4.9 %) who were in a postmenopausal state between the age of 40 and 49 years, compared to women in the case group (56.7 %). On the other hand, there is a larger proportion of women in the control group (13.5 %) who were in a premenopausal state between the age of 50 and 59 years, compared to women in the case group (0.0 %). These figures show that women who commence their postmenopausal state at a young age are more at risk of breast cancer than women who have a delayed menopause.
Limitations—bias and confounding
Retrospective case-control studies are prone to recall bias and cannot be used to establish causality. The main problem with case-control studies and dietary questionnaires is that recall bias tends to be unavoidable. In this study for example, people were not recruited upon diagnosis, and therefore, one could not assess diet prior to diagnosis. It is known the dietary changes do occur more often in those who have suffered a life-threatening pathology such as breast cancer. This was reflected in our findings in the study, as shall be discussed below.
Recall could not be avoided in this questionnaire, as people in general do not really know the amount of an item they typically consume in a week. Additionally, case subjects in the study might be more cautious as to which items they eat and consume in light of their pathology, and therefore, the reporting from cases is probably more accurate in this regard. In the same way, however, the cases might not be so willing to report consumption of items they deem as “harmful to health” so as to avoid embarrassment. One can therefore never truly know how often an item is consumed in a case-control study.
Recall bias also becomes a problem in the older strata of the population where due to a decrease in cognitive function accurate recall might be a problem. This study included a large number of women defined as “elderly” that might or might not have been suffering from memory impairment due to ageing.
In an attempt to address the problem of recall bias, questions asked were limited to “how much of item X do you consume per week” and then added up to a monthly figure. Additionally, images of containers were shown in an attempt to introduce standardisation of quantity consumed across subjects. These attempts probably went some way at addressing the issue of quantity standardisation and gave a picture as to a “typical” diet. However, recall bias remains a factor in this study.
Confounding factors are also a significant problem in retrospective case-control studies such as the one carried out here, as in most situations there are various confounding factors that cannot be catered for. One example is work exposure. While questions were asked as to possible exposures, the subjects who provided an answer were too few for statistical significance to be reached. This could also be true for various food factors, where a complete dietary analysis was impossible to carry out given timeframes and available funds.
Another major limitation in the study is that breast cancer was not categorised by individual subtype. Different cancers might have different aetiologies and different risk factors contributing to their possible emergence. However, this data was not readily available and while it would have been obtainable, it was not deemed central for this study, which was conducted as an exploratory exercise. Additionally, the relatively small cohort of people involved in the study would have made any statistically significant results hard to achieve if one were to look at sub-categories.
However, despite the obvious limitations, these results can form the basis of further studies focusing on the factors that were indicated as showing a significant effect on the occurrence of breast cancer in general. The odds ratios discussed in the results section should therefore be viewed with caution, given the magnitude of recall and confounding bias. One cannot view the results in a quantitative matter but rather as possible indicators for future studies in this field. On the other hand, a number of factors found to be associated with breast cancer are actually substantiating other published work.