Risk Factors for Mortality among Participants of Investigation into Cancer and Nutrition - Statistics Project Example

An Investigation of the Risk Factors for Mortality among Participants of the European Prospective Investigation into Cancer and Nutrition (EPIC) Institutional Affiliation Abstract The aim of this report was to investigate what factors contribute to mortality among a host of risk and non-risk (safety) variables. The sample comprised 5872 participants of the European Prospective Investigation into Cancer and Nutrition (EPIC). The risk factors included high Body Mass Indexes (BMIs), old age, smoking status (exploring options ranging from whether the individual ever/ never smoked to whether they currently smoke), to whether the individual consumed too much meat. The safety factors included high consumption of vegetables, legumes and fruits. Equally, younger ages and lower BMI measures were considered as favourable factors for health, and unlikely to contribute to mortality. Using descriptive statistics, the average age of participants was found to be 49.54 years (min. = 34.17, max. = 67.34). Males comprised 62.3% of the entire sample, while only 36.5% had not completed high school education. 39.2% of participants had never smoked in their lives, while 58.3% were either obese or overweight. 8.3% were above the age of 60. The results of the binary logistic model indicated that none of the four dietary components (vegetables, legumes, fruits and meat) were significantly involved in alleviating the incidence of death. However, the smoking status was a significant predictor of the life/ death status of participants. Equally, some age groups were also found to be significant moderators of the response (life/ death status). Introduction The aim of this report is to establish the risk factors for mortality among a group of 5872 participants of the European Prospective Investigation into Cancer and Nutrition (EPIC). EPIC is touted as one of the largest cohort studies globally, with over 500 thousand participants followed over a fifteen-year duration. Its primary role is to investigate the relationship between diet, environmental factors, lifestyle, nutrition, and the prevalence of cancers and other chronic diseases (International Agency for Research on Cancer, 2014). By using data pertaining to various risk and safety factors for the participants, the report presents a scenario that exposes the individual roles played by these factors in either aggravating or alleviating the risk of cancer, and the consequent deaths the condition may cause. Based on the contrasting contributions of the risk and protective factors of mortality due to cancer acquired through smoking or body weight complications, the factors were examined concurrently in this report. The discussion draws from a brief review of literature that reveals the expected roles of each of the variables used. Besides the abstract and this introduction, the report covers other sections on methods, analysis, and discussion. Smoking is one of the risk factors for cancer globally. The United Kingdom (which has the highest rate of female smokers in Europe) also has the leading rates of cancer deaths in the continent (Ferlay et al, 2013). Investigating the prevalence of lung cancer in the UK, Kiri, Sorino, Visick and Fabbri (2010) observed that the average prevalence of lung cancer among people living in the UK between 1991 and 2004 was higher for those with a prior history of chronic obstructive pulmonary disease. Apparently, a high percentage of lung cancer is attributable to cigarette smoke – there is a strong correlation between rates of smoking and lung cancer incidence (Proctor, 2001). One of the factors that have led to increase in global rates of smoking is the push to acquire slim figures that are touted as being more attractive. Apparently, smoking is associated with slimmer figures and loss of weight. Consequently, some women end up engaging in the risk-laden act of smoking, which inevitably exposes them to lung cancer, among other forms of cancer. High uptake of junk foods is likely to cause weight-associated problems, thereby increasing the possibility of deaths from weight-related complications. As a result, individuals are more encouraged to take healthier foods, such as legumes, vegetables and fruits. The contrasting contributions of meat and the listed healthier foods have been documented in various studies; with notable correlations being observed between high consumption of each class of food and the body mass. For instance, Hu (2003) noted that plant-based foods (including fruits, legumes and vegetables) were better placed to help combat excessive weight gain; which could be the result of high uptake of meaty foods. These plant based foods are, therefore, able to prevent accumulation of unneeded weight (Tuso, Ismail, Ha & Bartolotto, 2013). This underlines why their consumption is highly encouraged by health and medical practitioners. Ferlay et al. (2013) recorded intriguing statistics showing that in Europe, lung cancer ranks fourth in prevalence, but is the leading cause of death among all types of cancer. Apparently, the more age-advanced individuals are more prone to cancer and other chronic diseases (including lifestyle diseases such as diabetes and cardiovascular diseases) than their younger compatriots. Equally, the lesser active, older individuals are more likely to accumulate excessive weights, and for aging smokers who have smoked for substantial parts of their lives, it is more likely that they experience these conditions at a higher rate than the younger, less exposed individuals (Tuso et al, 2013). This implies that age, by extension, also serves as either a risk or protective factor, based on one’s current tally. The ratio of men engaging in smoking is larger than that of women. Consequently, a larger percentage of men are likely to suffer from lung cancer than do women. While there is no anatomical inter-gender bias in the causation of cancers (especially lung cancer), it is important to credit the greater participation of men in smoking to the higher rates of cancers resulting from this practice. As such, gender is likely to play a role in determination of incidence of deaths in the report. However, this is restricted to the scenario explained above. Methods The sample comprised some 5872 participants in the European Prospective Investigation into Cancer and Nutrition (EPIC). Participants were drawn randomly from across the participating nations, with no set limits for the number of participants from any country. Details about their country of residence were not included in this report. For purposes of ethics and concealment of participants’ identities, each participant was assigned an eight digit identification number. This way, their personal attributes were concealed as required by research ethics. The variables were measured at three levels: scale, ordinal and nominal. The scale variables include the ages, BMI, and mean daily consumption of vegetables, legumes, fruits, and meat (measured in grams). The nominal variable was basically the highest levels of education attained. The ordinal variables included the smoking status (whether participants smoked or not) and the genders of participants. The initial nominal variable is described below. School: 1 = none/ primary school, 2 = technical/ professional school, 3 = secondary school, 4 = university. That way, the variable resembles an increasing hierarchy, with the increasing levels of educational attainment corresponding to higher ranks. For the purposes of identifying how certain classes within the variables BMI and age could impact the overall logistic models developed, both variables were transformed into nominal before they were introduced into the binary logistic model. The transformation of BMI saw every participant’s measure classified as normal (represented by digit 1), overweight (2), or obese (3). The logically increasing hierarchy of increasing ranks for weights takes a similar trend to that of “schooling”. Equally transformed were the ages of participants. Based on minimum and maximum ages reported in the initial analysis, the ages were classified into classes of 10, spanning between 34.01 and 70. The age groups were developed as 34.01 - 43.00, 43.01 – 52.00, 52.01 – 61.00, and 61.01 – 67.30. Using the Statistical Package for the Social Sciences (SPSS), the data was analysed for interesting characteristics using the descriptive and frequency approaches. The data was also analysed for suitable regression models to determine which variables were influential in determining the mortality incidence of participants. For this purpose, binary logistic regression was used because the response (mortality) was split into a binary variable with the possibilities of being alive or dead at the moment of compiling the data. Results The results were divided into descriptive and inferential. Preliminary statistics were obtained to identify important characteristics of the data, including distribution. Descriptive Statistics Table 1. Descriptive statistics. N Range Minimum Maximum Mean Std. Dev. Variance Skewness Kurtosis Statistic Std. Error Statistic Std. Error Years 5872 33.17 34.17 67.34 49.55 7.96 63.38 -0.006 0.032 -1.024 0.064 Body Mass Index 5872 31.83 15.92 47.75 26.08 3.64 13.26 0.845 0.032 1.686 0.064 Mean daily consumption of vegetables (grams) 5872 1106.61 0.00 1106.61 182.77 100.05 10009.39 1.40 0.032 3.930 0.064 Mean daily consumption of legumes (grams) 5872 92.57 0.00 92.57 6.47 7.65 58.54 3.055 0.032 17.06 0.064 Mean daily consumption of fruits (grams) 5872 2635.44 0.00 2635.44 337.90 195.51 38226.03 1.877 0.032 9.444 0.064 Mean daily consumption of meat (grams) 5872 406.86 0.00 406.86 111.63 54.76 2999.03 0.825 0.032 1.154 0.064 Valid N (listwise) 5872 Based on frequency analysis, 3657 of the participants were male, making up 62.3% of the entire sample, while females comprised 37.7% of the sample. 1093 (18.6%) of participants had only acquired the basic primary school education at most, 1053 (17.9%) had technical/ professional qualifications, and the largest proportion (3118 = 53.1%) had a high school education as their highest qualification. 608 (10.4%) had attained a university education. The majority (2303 = 39.2%) had never smoked, while 1886 (32.2%) of the participants had already quit smoking at the time of the survey. 1552 (26.4%) were smoking at the time of the survey while 131 participants did not indicate their smoking statuses. Only 92 (1.6%) of participants had died at the end of the four years follow up, implying that 5780 (98.4%) of them were still alive after the four years. Based on their BMIs, 2444 (41.6%) had normal weights, 2659 (45.3%) were overweight, and 765 (13.0%) were obese. Four respondents did not provide valid responses to the question. After grouping participants based on their ages, 1455 (24.8%) were found to be between 34 and 43 years, 2027 (34.5%) were between 43.1 and 52 years, 1899 (32.3%) ranged between 52.1 and 61 years, and 488 (8.3%) were aged between 61.1 and 67.3 years. Table 1 above has the descriptive statistics of the variables pertaining to different features of the study participants. The average age of the participants was 49.5 (std. dev. = 7.96) years. The youngest recruit into the program was 34.1 years while the oldest was 67.3. Therefore, the difference between the ages of the youngest and oldest participants was 33.17 years. The average BMI for the participants was 26.08 (std. dev. = 3.64); the lowest BMI recorded was 15.92 while the highest was 47.75. The average mean daily consumption of vegetables for the participants was 182.78 (std. dev. = 100.05) grams; the lowest level of consumption for an individual was 0.00 and the highest was 1106.61 grams. The daily average consumption of legumes, fruits and meat equally had the lowest consumption levels recorded at 0. Legumes (mean = 6.47, std. dev. = 7.65) had a highest consumption level of 92.57 grams daily for the leading consumer, with fruits (mean = 337.9, std. dev. = 195.51, max. consumption = 2635.44 grams) recording the highest average consumption levels, and meat (mean = 111.63, std. dev. = 54.76, max. = 406.86 grams) recording the third highest consumption levels. Inferential Statistics Table 2. Variables in the equation. B S.E. Wald df Sig. Exp(B) 95% C.I.for EXP(B) Lower Upper Step 1a BMIRecoded 1.240 3 .743 BMIRecoded(1) -17.168 19505.756 .000 1 .999 .000 .000 . BMIRecoded(2) -.319 .311 1.056 1 .304 .727 .395 1.336 BMIRecoded(3) -.298 .294 1.034 1 .309 .742 .417 1.319 AgeRecoded 38.439 4 .000 AgeRecoded(1) -18.330 22927.963 .000 1 .999 .000 .000 . AgeRecoded(2) -2.435 .476 26.179 1 .000 .088 .034 .223 AgeRecoded(3) -1.506 .323 21.679 1 .000 .222 .118 .418 AgeRecoded(4) -.592 .276 4.579 1 .032 .553 .322 .951 Smokes .315 .127 6.162 1 .013 1.370 1.069 1.757 Vegetables -.001 .001 1.140 1 .286 .999 .996 1.001 Legumes -.011 .016 .510 1 .475 .989 .958 1.020 Fruit -.001 .001 .705 1 .401 .999 .998 1.001 Meat .000 .002 .001 1 .972 1.000 .996 1.004 Sex -.345 .251 1.878 1 .171 .709 .433 1.160 Constant -2.546 .657 15.020 1 .000 .078 a. Variable(s) entered on step 1: BMIRecoded, AgeRecoded, Smokes, Vegetables, Legumes, Fruit, Meat, Sex. Table 2 above shows the results of the binary logistic model for predicting whether the factors listed in the survey would ultimately contribute to the death or life status of a participant. From the table, it is important to note that many variables (including dummies) did not have significant predictors based on the 0.05 level of significance criterion. As such, their application in the model was discontinued. These variables include all dummies of the variable BMI, the age dummy, age-recoded 1 (the age between 34.1 and 43 years), and participants’ consumption of vegetables, legumes, meat and fruits. As such, the new model for predicting the death/ life status is as shown below. Life/ Death Status = 0.315*Smoking Status – 2.435*(Age between 43.1 and 52 years) – 1.506*(Age between 52.1 and 61 years) – 0.592*(Age between 61.1 and 67.3 years) – 2.546. These results indicate that consumption of any of the four diet components does not alleviate the incidence of death. However, the smoking status tends to increase the incidence of death (the variable has a positive predictor). Since age is generally applied as a control variable, we can freely drop it from the model. The new model becomes: Life/ Death Status = 0.315*Smoking Status – 2.546. We can now calculate the likelihood that smoking leads to higher incidence of death. The odds of dying after engaging in smoking are expressed in the statistic 0.107. From this figure, we can calculate the probability of dying as: Odds/ (1 + Odds) = 0.107/ (1 + 0.107) = 0.097%. The odds that an individual who does not smoke died within the four-year study period is 0.078. The probability of dying in this case is 0.073%. Using the Odds ratio, we can calculate the number of times that an individual who smoked was more likely to die compared to one who did not as follows: Odds (smoked and died)/ Odds (did not smoke but died) = 0.107/ 0.078 = 1.37. This means that an individual who engaged in smoking within the study period was 1.37 times more likely to die compared to one who did not. Discussion The results of this analysis indicate that smoking is a significant determinant of the incidence of deaths among participants. Consumption of vegetables, fruits, legumes and meat did not appear to significantly reduce the incidence of death. This shows that these foods do not help in the alleviation of incidence of death – perhaps, some other factors are more influential in this respect. The results are partly aligned with previous findings, including that by Proctor (2001) that found that smoking is significantly correlated with deaths from lung cancer. However, they also differ with the common notion that has been developed through several research studies. For instance, Hu (2003) observed that certain classes of foods significantly reduced the incidence of deaths among relatively prone sections of the population; some of the classes of foods including fruits, vegetables and legumes. As such, the results contradict the common notion that these classes of foods partly deter deaths. The basis of the current argument is that certain classes of foods and cigarette smoking statuses work in contrasting modes to contribute to varying incidences of deaths among respective users. Generally, the research banked on the assumption that greater consumption levels for the various classes of foods would contribute to lowered incidence of deaths. This is the notion that Hu (2003) advanced through his research. Equally, the research banks on the assumption that smoking effectively increases the incidence of deaths, a fact that this study has positively corroborated. Some of the negative implications of smoking on the body is that the behaviour often leads to lung cancer, a condition that was at the centre of the European Prospective Investigation into Cancer and Nutrition (EPIC). The researchers intended to investigate whether the mentioned classes of foods are, indeed, significant contributors of reduced incidences of deaths. This way, they would act as alleviators of the incidence of deaths from smoking. The combined effect is that though either type of death-incidence control pulls in a specific direction, the mechanism there would be a partial “cancelling” effect that consuming vegetables, legumes, meat and fruits would have on the overall levels of death/ life incidence. Effectively, the present results indicate that though there could be such a counter effect, the contribution of the foods is insignificant in cancelling that of smoking. Conclusion, Limitations and Recommendations The results effectively annulled the notion that fruits, vegetables, meat and legumes are not significant alleviators of the effects of smoking on the incidence of deaths, particularly those from smoking-motivated cancer. However, smoking significantly motivated higher incidences of death from cancer. The results further established that smokers were 1.37 times more likely to die within the study period compared to non-smokers. The results are, however, valid to the extent that the data effectively conforms to the assumptions of normality. In this case, referring back to the measures of kurtosis and skewness, it is evident that the variables that mainly violated these assumptions (mainly the average daily consumption levels of meat, vegetables, fruits and legumes) eventually ended up yielding results that did not conform to those indicated in relevant studies. For that reason, it is important that the data is considered for further exploratory analysis, or the use of alternative tests (particularly the non-parametric tests). Equally, the data could be transformed in such a way that the more obvious effects of violations to the requirement for normality are reduced. Such could include logarithmic transformations, which serve as a powerful tool for converting data that disobeys the assumptions of normality into obeying the same. Notably, the fact that the data was positively skewed indicates that there could have been more exaggerated recordings from the relevant variables. To control for such faulty recording of data, the analyst could consider such remedy as the removal of outliers spotted using the box plots. References Ferlay, J., Steliarova-Foucher, E., Lortet-Tieulent, J., Rosso, S., Coebergh, J. W. W., Comber, H…and Bray, F. 2013. Cancer incidence and mortality patterns in Europe: Estimates for 40 countries in 2012. European Journal of Cancer. 49: 1374-1403. Hu, F. B. 2003. Plant-based foods and prevention of cardiovascular disease: An overview. American Journal of Clinical Nutrition. 78(3): 5445-5515. International Agency for Research on Cancer 2014. EPIC study. World Health Organization. Retrieved from epic.iarc.fr. (Accessed 23rd December 2014). Kiri, V. A., Soriano, J. B., Visick, G. and Fabbri, L. M. 2010. Recent trends in lung cancer and its association with COPD: An analysis using the UK GP Research Database. Primary Care Respiratory Journal. 19(1): 57-61. Proctor, R. N. 2001. Tobacco and the global lung cancer epidemic. Perspectives. 1: 82-87. Tuso, P. J., Ismail, M. H., Ha, B. P. & Bartolotto, C. 2013. Nutritional update for physicians: Plant-based diets. The Permanente Journal. 17(2): 61-66. Read More