Association of dietary intake and cervical cancer: a prevention strategy

Abstract

Introduction

Cervical cancer is one of lethal cancers in women. As a global concern, identifying important factors of cancer is a useful strategy for prevention. Due to the role of diet/nutrition factors for cancer, the purpose of our study was to determine the impact of 150 nutrition/vitamin factors and 50 non-nutritional factor in cervical cancer and phase.

Methods

Population samples of 2088 healthy subjects and patients with cervical cancer were investigated. 200 factors such as vitamin E, B1, B6, fruits, HPV, and age were gathered. Deep learning, Decision tree, and correlation matrix were used for modeling and identifying important factors. SPSS 26, R4.0.3, and Rapid miner were utilized for implementation.

Results

Our findings indicated that zinc, Iron, Niacin, Potassium, Phosphorous, and Cooper have a beneficial impact in reducing the risk of cervical cancer and progression of phase in Iranian women, as well as Salt, snacks and milk Were identified as high-risk food factors (P value < 0.05 and coefficient correlation > 0.6). Also, alcohol, and sex patient with two groups, HPV positive have an impact on cervical cancer incidence. Phosphorus and selenium in the Micronutrients category (R² = 0.85, AUC = 0.993) and polyunsaturated fatty acid and salt in the Macronutrients category and other categories of nutrients were identified as the most effective factors in cervical cancer using deep learning (R² = 0.93, AUC = 0.999).

Conclusions

A diet and rich nutrition can be helpful for the prevention of cervix cancer and may reduce the risk of disease. Additional research is necessary for different countries.

Introduction

Cervical cancer is the third most common gynecological cancer in young women and the leading cause of mortality [1,2,3] every 2 min, one woman dies from cervix cancer [4]. 85% of new cases are in developing countries [5]. Cervical cancer is one of the disastrous menaces to women’s lives. Most of them have not been diagnosed [6]. In recent years cervical cancer has been a concern to primary physicians [7]. This is cancer, a largely preventable disease [8]. Therefore, the Development of national control and prevention program for cervical cancer should be considered to decrease cervical cancer incidence, morbidity and mortality, improve the quality of care and cost reduction [9]. Identifying the factors that affect cancer is an important prevention strategy. Many studies have pointed out the role of human papillomavirus (HPV) as a necessary cause of cervical cancer [10, 11]. Marriage age, marital status, age of first pregnancy, smoking, family history, multiparty, Education level and Economical status, also are mentioned in other studies [12]. In recent Research, dietary antioxidants, such as vitamins A, C, D, E and nutrition hold a rather great share in cervical cancer prevention [13,14,15,16]. There is growing evidence related to the effect of nutrients on cancer prevention. On the other hand, because there are geographical differences in the incidence, risk factors and mortality of cervical cancer, studies in different countries are necessary [12, 16]. Considering the lack of a comprehensive study related to many aspects of dietary/nutritional factors on Cervical cancer, the current study is mainly aimed to investigate the importance of 91 nutritious and vitamin factors also 31 demographic, sexual factor and medical examination factors on the development of cervical cancer and then formulate a strategy for prevention.

We considered three objectives for meeting the main purpose of the study:

Aim1 Regarding to the importance of diet/nutrition in cervical cancer prevention, the correlation between all dietary/nutritional variables on cervical cancer and phases were studied and the preventive and reductive effects of the nutritional intake on cervical cancer, HPV and phases were reported.

Aim2 Due to the importance of HPV and sexually related factors on cervical cancer incidence, the binary correlation of these variables on cervical cancer and phases were examined using a correlation matrix.

Aim3 The effect coefficient of each category of variables such as macronutrients, micronutrients and junk foods on cervical cancer were calculated using deep learning and decision tree models.

Material and methods

Participants and data source

This was a population-based study involving 2088 Iranian cases (Mashhad). All participants gave informed, written consent to contribute to the survey, reviewed and approved by the ethics committee of Mashhad University of Medical Sciences (MUMS). A semi-quantitative Food Frequency Questionnaire (FFQ) was used in the clinic to assess dietary habits. The FFQ took roughly 40 min to complete and collected data on 65 different food products. To reduce estimating errors in portion and consumption frequency, the FFQ was administered by specialized nutritionists via face-to-face interview. For the previous year, the frequency of consumption of different food products was recorded on a daily, weekly, monthly, rarely, and never basis. The reference serving size was applied to determine portion sizes. Food items were categorized into 17 food groups, including fast food, fruits, and vegetables. The total energy intake was calculated by adding all of the food energy intakes together. 200 factors were collected from samples. 91 nutritional factors and 31 non-nutritional factors, were identified as suitable for modeling based on expert opinion in two rounds of the Delphi method, the descriptions of variables can be seen in Table 1.

Table 1 Variables used in the study

Statistical analysis

Descriptive analysis, normality test (Kolmogorov–Smirnov tests) and Spearman correlation were performed by SPSS 26. A significance level of 0.05 was considered for analysis.

Machine learning methods

Deep learning and decision tree models were used to identify the effective factors in a category of variables including macronutrients and micronutrients. The significant variables gained from the feature selection method (Weight by Correlation) were the final parameters in creating the model. The two machine learning techniques used in the study, the decision tree and deep learning, are described following. Also the present study used the correlation matrix to investigate the dependence between variables. A correlation matrix depicts the coefficient of correlation between variables. The correlation coefficient is measured from − 1 to 1. A positive correlation points that the variables are in the same direction, while a negative correlation shows the variables in opposite directions. The lack of correlation is displayed by 0.

Decision tree

The Decision tree is a very popular class of predictive models due to their interpretability and best performance special on categorical data. It is a tree-based technique in which a data separating sequence characterizes any path from the root node to the leaf until a Boolean outcome is obtained. Decision trees are an effective tool that may be utilized in various domains, including machine learning, image processing, and pattern recognition [17].

Deep learning

Deep learning is part of machine learning methods based on artificial neural networks. Deep learning allows computational models with several layers to learn multiple degrees of abstraction for data representations and can automatically learn feature selection from many varying data, Deep learning uses the backpropagation algorithm to show how a machine should adjust its internal parameters that are used to compute the representation in each layer from the representation in the previous layer, revealing intricate structure in massive data sets [18].

Computational workflow

R 4.0.3 and Rapid miner version 9.10 were utilized for modeling. For Decision tree modeling, the max depth = 10, minimal gain = 0.01, minimal leaf size = 2, minimal size for split = 4, number of pre-pruning alternatives = 3, confidence = 0.1 and were some of the tuning hyper parameters which were considered. In deep learning, parameters of epochs = 20, activation function = Rectifier and learning rate = 0.01 were set.

The standard workflow was utilized to create, evaluate, and optimize methods explained as follows.

Splitting data into training and test sets

To provide some independent evaluation levels, it is common practice to split the source data set into two parts: training and test data. The model is then optimized using the training data and independently evaluated using the test data.

Performance measures optimization and generalized predictive ability

In the current study, 70/30 train/test ratios were determined for machine learning models. For each workflow, a model with the fixed optimal hyper parameter values is retrained on data and randomly sampled from the complete data set, and then evaluated on the unused data.

Model evaluation using a test set

Machine-learning methods assessment was performed by 5 indicators, including the accuracy, R², MSE, and AUC.

$${\text{Accuracy}} = \left( {{\text{TP}} + {\text{TN}}} \right)/\left( {{\text{TP}} + {\text{TN}} + {\text{FP}} + {\text{FN}}} \right)$$

where TP—true positive; FP—false positive; TN—true negative; FN—false negative.

$${\text{MSE}}\;({\text{Mean Squared Error}}) = \left( {1/{\text{n}}} \right)*\Sigma \left( {{\text{actual}} - {\text{forecast}}} \right)2$$

where Σ—a symbol that means “sum”; n—sample size; actual—the actual data value; forecast—the predicted data value.

R² (R-Squared) = 1 − unexplained variation/total variation. It is the coefficient of determination and tells you the percentage variation in y explained by x-variables.

AUC (area under the curve): It represents the degree of separability. It illustrates how much the capability of the model in distinguishing between classes.

Results

Data description

Table 2 shows the mean and standard deviation of the quantitative variables. The Frequency and percentage of cancer patients and healthy people participating in the study are also mentioned.

Table 2 Characteristic of population

Data analytics

The results of data analysis can be seen in the following text:

Response to the aim1: the correlation between the dietary/nutritional intake and the risk of cervical cancers and phase progression

Tables 3 and 4 show the correlation between dietary/nutritional intake and the risk of cervical cancers and phase progression.

Table 3 The preventive and reductive effects of the dietary/nutritional intake on cervix cancer

Table 4 The preventive and reductive effects of the dietary/nutritional intake on phase progression

Note that the correlation coefficient less than 0.3 is considered weak, the coefficient between 0.3 and 0.6 is moderate and the coefficient greater than 0.6 is considered strong. Significant values with medium and high coefficients were listed in Table3. Positive numbers indicate high-risk diets/nutrients and negative numbers indicate Preventive and Reductive diets/nutrients effect.

The findings of Tables 3 and 4 revealed that zinc, iron, niacin, potassium, phosphorous, cooper and folate have an impact in reducing the risk of cervical cancer and progression of phase (see Fig. 1), as well as salt, snack and milk were identified as high risk factors. Dietary fiber, starch and Tot.N2.g also has a beneficial impact on cervical cancer and phase. In Fig. 1 is displayed important macronutrients and micronutrients affecting on cervical cancer and its phases. Seasonal and tree fruits also have a good effect on cancer and phase, Meat and vegetables as well have a reducing effect on phase progression.

Fig. 1

Important macronutrients and micronutrients affecting cervix cancer and phase

Response to aim2: correlation of HPV and sexually related factors with cervix cancer and phases

Using the correlation matrix, we examined the dual correlation between cervical cancer/phase with crucial medical examination variables. The results are shown in Figs. 2 and 3.

Fig. 2

The correlation between crucial medical examination and cervical cancer

Fig. 3

The correlation between phase and crucial medical examination

The correlation between crucial medical examination and cervical cancer

Figure 2 showed that Data is not normally distributed and a high coefficient was detected between cervix and exocervix, hpv_positive, smear, wart and hpv_sign_cat in positive direction.

The correlation between crucial medical examination and phase

Figure 3 showed that Data is not normally distributed and the high coefficient was detected between phase and HPV_sign_cat and smear in negative direct.

Correlation between sexual factors and demographic on cervical cancer can be seen following (Tables 5 and 6).

Table 5 Correlation between sexually related variables and cervix cancer

Table 6 Correlation between demographic variables and cervix cancer

As shown in Table 5, Conttaception_method, Menstrul disorder, Number of sex in a month, the age of first sex has a positive correlation with cancer and successful pregnancies have negative correlation on cervix cancer.

As can be deduced from Table 6, Residient_place, Financial_status, Marriage_status, Education_status and Alcohol are associated with cancer.

Response to aim3: identifying important coefficients on cervical cancer in each category

According to the results obtained from Table 7, the most important coefficients in combination with other variables in each category from high to low can be seen in each category, for example, in the macronutrients category, phosphorus, selenium and zinc have the most effect on cancer, respectively.

Table 7 Important coefficients in each category with machine learning methods

Discussion

Cervical Cancer has a high mortality rate in women and endangers their lives. Identifying the most important factors of cancer is a critical challenge in prevention strategy and can even be helpful in early diagnosis. Among the factors that not be overlooked is the influence of diet/nutrition. Therefore, in the present study, the factors affecting cervical cancer, especially dietary factors and vitamins, were studied. Also, non-nutritional factors affecting cervical cancer and phase were identified.

Our findings indicated that, Phosphorous, selenium, Iron and zinc have an impact in reducing the risk of cervical cancer and progression of phase, as well as salt, snacks and milk were determined as high risk food factors. dietary fiber, starch and Tot.N2.g also has a beneficial impact on cervical cancer and phase. Meat and vegetables as well have a reducing effect on phase progression. Seasonal and tree fruits also have a good effect on cancer and phase.

Similar and contradictory results have been reported in various studies, which are mentioned following. In Meta-analysis by Myung et al. reported that carotene was associated inversely with cervical cancer risk and vitamin A had no effect on cervical cancer risk [19]. In a meta-analysis by Cao et al. Vitamin C was significantly associated with cervical cancer reduction risk [20]. In study by Hosono et al. Vitamin D intake, in Guo et al. study α-carotene, β-carotene and vitamins E and C, in research of Manju et al. vitamin C, E had a preventive role in cancer [14, 21, 22]. Beneficial effects of fruits and vegetables on cancer prevention have been reported in some studies [23,24,25].

The effect of nutrients on HPV has been studied in several studies. A study in Sao Paulo reported that the consumption of papaya plays a preventive role against HPV infection [26]. In the Chih et al. study, the consumption of fruits, vegetables, yogurt, fish, tofu and meat was considered to be effective to decrease the risk of HPV [27]. The Result of Barchitta et al. study shows that a high intake of red and processed meats, dipping sauces, chips, and snacks with a low intake of olive oil in the Western diet, was related to a higher risk of HPV. In contrast, Mediterranean diet (MD), Consisting vegetables, legumes, fruits and nuts, cereals, fish, and a high ratio of unsaturated to saturated lipids had a lower risk of HPV [28].

Sedjo et al. reported high consumption of vegetables and carotenoid be beneficial in reducing HPV risk [29]. In the review of Koshiyama M publisher in 2019, multi-vitamin, vitamin A, vitamin C, vitamin D, vitamin E, papaya, Mediterranean diet, carotenoids, fruits, vegetables, legume, lycopene, green tea, folate, sulforaphane, polyphenol Flavonoids, polyunsaturated fatty acid, calcium (±) reported generally as main preventive and reductive factors against CC risk and introduced cigarette, Western diet and oleic acid as high-risk diets/nutrients. Also Mediterranean diet, papaya Vitamin-C, vegetables, carotenoids and fruits had a reductive effect on HPV infection [16].

Piyathilake et al. showed that folate has a significant inverse association with HPV infection [30]. A study by Giuliano et al. found a relationship between persistent HPV infection and low intake of vitamin C [26]. Sedjo et al. showed that vegetables and fruits and juices were associated with a reduction in the risk of HPV persistence [29]. In studies, the effect of a diet containing vegetables on cancer prevention and HPV was positive [24, 27, 31,32,33]. We did not find any correlation between nutrient intake and HPV in the present study.

Some research has studied the association of socioeconomic characteristics with cancer. In some studies, Age at first marriage, Number of deliveries and Contraceptive methods have been reported as important factors on cervical cancer [34]. In study of Nojomi et al. low marriage age, high prevalence of pregnancy, family history, contraceptive pills and Low age at first pregnancy associated with cervix cancer [35]. In the study by Vaisy et al. marriage at age below 16, marital status, married more than once, consumption of Protective factors were reported as influence factors and Contraceptive pill [36]. In contrast, in the study by Mohaghegh et al. multiple marriages and multiple sexual partners were significant, but smoking, diet, and being widowed or divorced, have no significant correlation with cancer [37]. Tadesse showed poverty, early marriage, and high parity influence factors were associated with cervical cancer [38]. In Ansari study, socio-economic factors were reviewed as key factors in cervical cancer, such as increasing age, education, knowledge, marital status, multiple sexual partners, financial status, using inappropriate clothes or having bad sanitary during menstruation, sexually activeness, HPV, post-menopausal bleeding, offensive vaginal discharge, having many pregnancies, pills and injections [39].

In our research, constipation method, menstrual disorder, number of sex in a month, the age of first sex and Sexmate_patient_2group has a positive correlation with cancer and successful pregnancies has a negative correlation with cervix cancer. Among demographic factors, Residient_place, Financial_status, Marriage_status, Education_status and Alcohol were correlated with cancer. Smear, hpv_sign_cat, wart and hpv positive also had important examination factors for cervix cancer diagnosis.

Conclusions

With Developing a comprehensive strategy, cervical cancer can be effectively controlled. Discovery factor affecting cervical cancer facilitates prevention, diagnosis and treatment. As in the present study, all factors affecting the incidence of cervical cancer were investigated to develop an appropriate prevention protocol.

In research during recent years, antioxidant vitamins have attracted much attention in cancer prevention. Because they protect cells from oxidative DNA damage and enhance the immune system.

Based on the results of the present study, healthcare workers should educate women to consider vitamins and useful nutrition in their dietary regime to prevent the development of cervical cancer. Education about cervical cancer should offer to women, families and communities.

The influence of diet and nutrition mechanisms on cervical cancer is unknown. Further research is needed to clarify these mechanisms. Foods consumed by humans are widespread and the effects of all nutrition cannot be measured, it is recommended that other foods be considered in research. Most articles were epidemiological and clinical trials. There were few experimental studies in this regard. It is recommended more experimental studies will do. Otherwise, different nutrition/diets may have differing impacts on cancer in the different geographic places. Therefore, it is necessary that similar studies would be done in different countries in the future. In general, we must continuously avoid the consumption of large amounts of high-risk diets and nutrients, and at the same time continuously consume preventive and reductive diets and nutrients. Therefore, Lifestyle modifications, including attention to diet, social habits, sexual behavior and vaccination can greatly prevent cancer.