In this case control study, we tried to use machine learning methods to develop a risk prediction model for bladder cancer according the lifestyle and occupational risk factors. Also, we present a dynamic web-based nomogram to calculate the probability of bladder cancer after entering the lifestyle and occupational information for a person. This user-friendly nomogram is available in the following link ( https://nbshiny.shinyapps.io/BladderCancer/ ).
In this study, the univariable results showed that half of the 12 important factors for UBC were related to the past medical history including: chronic renal failure, bladder stone, neurogenic bladder, spinal paralysis, recurrent infections, diabetes, and a family history of previous UBC. And finally, three of factors related to UBC were from the lifestyle items: smoking, opium and analgesics use. The last two ones were fruits and vegetables, and pickles consumption per day, which belong to the Dietary factors category. In addition to the importance of acquiring a complete patient history, lifestyle and dietary risk factors must be considered. The main reason is that these are factors that can be affected simply by educating the patients and raising their awareness of these significant factors. Having a better understanding of these risk factors will prove to be a great asset to the prevention and management of bladder cancer in the future.
In 2018, a systematic study by Cumberbatch et al. divided the risk factors into six major groups: smoking, occupational exposure, dietary factors (such as alcohol/vitamins and antioxidants/fluid consumption/fruits and vegetables/meat/diversity), environmental carcinogens (e.g., arsenic, nitrates, selenium, cadmium, nuclear power plants and hair dye) gender, race, socioeconomic status and lastly the interaction between genes and the environment [17]. A more recent study in 2020 listed 9 groups of risk factors: gender, age, hereditary factors, smoking, environmental and occupational exposure, alcohol, red meat, obesity, and pathogens [18]. In our study we found out that previous diseases of a patient such as recurrent infections and smoking and opium use negatively affect the risk of UBC. On the other hand, the daily use of vegetables, fruits and pickles could have a beneficial effect on preventing UBC. This is in coherence with the aforementioned study; moreover, our study sheds light on the importance of past medical record of patients and emphasizes the roles of smoking and dietary factors on UBC.
Despite their importance, original research by Westhoff et al. showed that most UBC survivors were not aware of any risk factors contributing to their disease. Only 20 percent of the survivors assumed some possible cause. Others were completely oblivious to the role of risk factors in the development of UBC [19]. It can be deduced that general knowledge about bladder cancer risk factors is still scarce and that we need more and more educational programs informing the audience about risk factors and how to prevent cancers.
Genetical, anatomical, external, and internal factors have been suggested by different studies to play major roles in this incident. Frequent use of hair dye products is an example of the external factors. Internal factors mainly revolve around the sex hormones and the hormone replacement therapy in women and but the results have been somewhat contradictory [20]. In contrast, the use of hair color products was not significantly associated with UBC in our results.
Smoking and the use of tobacco-related products have been the center of attention in bladder cancer. Many studies have considered smoking to be the most important and well-known risk factor for UBC [17, 18, 21]. A systematic review discovered that smoking not only has a carcinogenic effect on the transitional cells of the bladder but it also negatively affects the treatment process and may exacerbate the progression and recurrence of the disease [6]. Another study found out that smokers who had quit for a long time are still at a much higher risk of developing cancer compared to never smokers [7]. Also, according to our results, smoking was one of the significant risk factors for UBC with a p-value of less than 0.001 in Table 1 and an importance of second position in Fig. 2.
Dietary factors contain numerous variables that may lead to or prevent UBC depending on their quantity and quality; nonetheless, studies that investigated the association between dietary factors and UBC have yielded inconsistent and controversial results. According to Piyathilake et al. the role of fruits, vegetables and micronutrients is still being debated. Despite all these inconsistencies, we can’t avert our eyes from the fact that a diet rich in fruits and vegetables and low in processed meat might have beneficial and protective qualities against UBC [22]. In our study, the daily consumption of fruits and vegetables, and pickles were also associated with a significant reduction in the risk of bladder cancer with an importance of fifth position in Fig. 2. In 2017 a meta-analysis surrounding total fluid consumption was conducted. Based on 21 case – control and 5 cohort studies, the authors found out that when all data were pooled together, there was no association between the risk of UBC and the total amount of fluids consumed [23]. Two dose-response meta-analysis studies regarding tea consumption and alcohol consumption were performed in 2017 and 2021 [24, 25]. In both studies no significant association or relationship was observed between tea or alcohol consumption and risk of urinary bladder cancer in the total population.
In an experiment to find out more about occupational risk factors connected to UBC, it was discovered that being an underground hard coal miner, a varnisher and a car mechanic was associated with a higher risk for bladder cancer [26].
Finally, the past medical conditions of patients must be considered when estimating the risk of UBC. Metabolic syndrome, Diabetes, and infections such as UTIs and parasites especially schistomiasis are all linked with an elevated risk to urinary bladder cancer [27, 28]. According to the results we discovered, recurrent infections, neurogenic bladder and bladder stones were all significantly associated with UBC in Table 1 and had an importance position of first, third and fourth in Fig. 2 respectively.
To the best knowledge of authors, there have not been any nomograms which have delved into the risk factors of UBC and predicting their role in carcinogenesis. One study has attempted to develop a prognostic one for patients with breast cancer [29]; however, other nomograms for UBC investigated the survival of patients with UBC [30]; and metastasis to lymph nodes [31].
In recent years, with the development of technology, studies have tried to use deep learning (DL) and machine learning (ML) models in order to come up with programs that can help in the diagnosis of cancers; predict their prognosis and survival outcomes; and select the best route of treatment for patients [32]. One study in 2019 used personal health data from patients and combined various machine learning methods to achieve better results in terms of breast cancer prediction [33]. Considering DL and ML studies on UBC, Tsai et al. tried to predict the neoplasm by utilizing laboratory data and machine learning methods. Their light GBM model differentiated bladder cancer from cystitis and other cancers with an accuracy of 84.8–86.9%, a sensitivity of 84–87.8%, a specificity of 82.9–86.7%, and an area under the curve (AUC) of 0.88 to 0.92 [34].In another study in 2021, machine learning algorithms were used on H&E-images of UBC. The results showed a higher efficiency for diagnosing (with an AUC of 96.3%, 89.2%, and 94.1% in the training cohort, test cohort, and external validation cohort), distinguishing UBC from cystitis (AUC of 93.4% in the general cohort) and predicting the survival rates (with AUC values of 77.7%, 83.8%, and 81.3% for one-, three-, and five-year survival) of patients with UBC [35]. In a review study, Suarez‑Ibarrola et al. concluded that machine and deep learning methods for UBC have been employed to predict treatment response, recurrence of tumors, and survival rates for the patients [36].
Up until now there have not been any machine or deep learning algorithms predicting the risk of UBC based on risk factors in the lifestyle, dietary, environmental, and occupational groups, and that’s the novelty of our study.
This study has some strength and limitations. We used a relatively large sample size of patients randomly chosen from the Iranian Cancer Registry system and developed risk prediction models. Also, our study led to production of a dynamic web-based nomogram which can be used to calculates the probability of UBC for everyone.
One limitation in our study was that we were not able to witness the effect of other risk factors belonging to various groups mainly including: dietary, environmental, and occupational groups. There are many more risk factors some of which have been investigated in previous studies such as the high intake of red meat [37], high levels of nitrate [38], and/or arsenic [27] in drinking water, nitrosamines in foods and drinks [27], exposure to pesticides [39] and many more are still to be addressed. Further and extensive research is needed to include these risk factors in a more comprehensive epidemiology research in the future and to include these risk factors in future nomograms to achieve results with increased sensitivity and specificity. Another limitation is that dietary and life style factors are greatly influenced by the culture, ethnicity, and geographical and historical factors and our study was conducted on the Iranian population so the results might differ in various contexts we suggest further studies to be conducted using larger sample sizes from all over the world in order to make the prediction model even more accurate.