Differences in Cancer Incidence among Predominantly Muslim and Buddhist Subpopulations in Songkhla

Songkhla is a unique province in Southern Thailand, where approximately 25% of its population is Muslim (National Statistical Office, 2012). The proportion of Muslims in its three southeast districts, namely Chana, Thepha, and Saba Yoi, is around 70%, while it is approximately 14% in the remaining 13 districts of Songkhla (Figure 1). In countries and areas inhabited by a mixture of Muslims and people of other religious beliefs, Muslim people usually live in clusters composed of members of their own community and have their own life style, i.e., their dietary and cultural practices are different from those of nearby religious communities. Such differences in practice, e.g., the kind and amount of spices in food originating in the Middle East and abstention from alcoholic beverage, may be related to differences in health status, especially cancer incidence, in their population. Such phenomenon is also true in Thailand; however, the


Introduction
Songkhla is a unique province in Southern Thailand, where approximately 25% of its population is Muslim (National Statistical Office, 2012). The proportion of Muslims in its three southeast districts, namely Chana, Thepha, and Saba Yoi, is around 70%, while it is approximately 14% in the remaining 13 districts of Songkhla ( Figure 1).
In countries and areas inhabited by a mixture of Muslims and people of other religious beliefs, Muslim people usually live in clusters composed of members of their own community and have their own life style, i.e., their dietary and cultural practices are different from those of nearby religious communities. Such differences in practice, e.g., the kind and amount of spices in food originating in the Middle East and abstention from alcoholic beverage, may be related to differences in health status, especially cancer incidence, in their population. Such phenomenon is also true in Thailand; however, the Muslim communities there are not isolated. Close contact among Buddhist and Muslim communities is a fact in the Southern provinces of Thailand. A few studies on the effect of religion on disease prevalence and healthcare system conducted in Southern Thailand have focused on differences in some diseases and health conditions, but, to date, there has been no study on cancer (Golomb, 1985;Baelum et al., 2003;Sengupta et al., 2006;Liabsuetrakul et al., 2009).
To avoid conflict between different ethnic groups in Thailand, the Thai government promoted the idea of a unified Thai nationality and devalued the sense of race and ethnic differences (Chuah, 2010;Laungaramsri 2014). This policy has caused both national unification and conflict at the same time. Even though the majority of Thai Muslims in this area have been able to cope well with the cultural differences between them and the non-Muslim populations, some have failed to. Their resistance has caused difficulties in healthcare services offered to this category of people. This phenomenon is more pronounced in the three southernmost provinces of Thailand: Pattani, Yala, and Narathiwat, but less in the three districts of Songkhla, which is close to these three provinces. Thus, access to the healthcare system in Songkhla is good, while it is less so in the three southernmost provinces.
Since no cancer registry exists in the three southernmost provinces of Thailand on the eastern coast of Peninsular Thailand, where the percentage of Muslim population is around 85%, and also Satun, a province on the West (Andaman) Coast with a proportion of Muslims of around 68%, the incidence of cancers in the three districts in Songkhla may provide some clues regarding cancer burden to justify a cancer-control plan for the Muslim-predominated provinces in Southern Thailand. The objectives of this study were to 1) describe the incidence of various cancers in predominantly Muslim subpopulation (PMSP), which comprises the residents of Chana, Thepha, and Saba Yoi districts, and 2) compare their incidence between PMSP and the predominantly Buddhist subpopulation (BPSP), who reside primarily in the remaining districts of Songkhla Province.

Recruitment of cancer cases
Cancer cases diagnosed between 1990 and 2010 were drawn from the database of Songkhla Cancer Registry. The population-based cancer registry of Songkhla is kept at the Faculty of Medicine, Prince of Songkla University. It involved both passive and active case finding from 23 sources of data consisting of 3 hospitals with over 400 beds, community hospitals, private hospitals, special hospitals, the Provincial Health Office, and the Population Registration Office of the province. The completeness of death registrations was acceptable. However, the number of undiagnosed and unregistered cancer cases was difficult to estimate, but was one among registries with expected completeness greater than 95% by capture-recapture method (Suwanrungruang, et al., 2011). A few Muslims from the remote areas of Southeastern Songkhla visited large hospitals in the two nearby provinces, and some of them, especially the elderly, used traditional herbal medicines and never accessed the governmental healthcare system. The situation has been continuously improving since the healthcare reform started in 2002 under the universal coverage scheme of the National Health Security Office, which has led to a much better organized patient registration system and ensured the coverage of essential cancer treatment. The cancer statistics of the Songkhla Cancer Registry meet international standards and have been regularly published in the series 'Cancer in Thailand' by the Thai National Cancer Institute (Vatanasapt et al., 1993;Deerassamee, et al., 1999;Sriplung et al., 2003;Khuhaprema et al., 2007;2012;2013), and 'Cancer Incidence in Five Continents' published by the International Agency for Research on Cancer (Parkin et al., 1997;2002;Curado et al., 2007).

Population denominators
The population denominators used in the calculation of incidence rates were estimated from the 3 population censuses conducted by the Thai National Statistical Office in 1990, 2000(National Statistical Office, 19922002;2012). The population denominators by both sexes for all districts were readily present in the censuses. Intercensus populations were estimated using the log-linear function between two consecutive censuses.

Statistical analysis
Cases and populations were stratified into PMSP and PBSP by district of residence as mentioned above, and the age-standardized incidence rates (ASR) per 10 5 population (Boyle et al., 1991) of various cancers were calculated for the two groups. The population denominators in each calendar year from 1990 to 2010 were estimated from population censuses of 1990, 2000, and 2010 using the log-linear equation for 5-year age group strata. Annual ASRs were analyzed by subpopulation, sex, and year of diagnosis by means of simple linear and Poisson regression models using the R statistical software version 3.0.2 (R Core Team, 2013). The annual ASRs for each particular major cancer in PMSP and PBSP were shown as graphs of cancer trends. ASRs of leading cancers in the last 5-year period, from 2006 to 2010, were calculated and shown via horizontal bar charts for a clear comparison of the incidence in the two subpopulations. The standardized incidence ratio (SIR) and its variance were calculated according to the formula documented by Boyle and Parkin (Boyle et al., 1991).
The Poisson regression models revealed a significantly lower incidence of cancers of the cervix and breast in PMSP compared to PBSP with odds ratios of 0.54 (95% CI: 0.45-0.64) and 0.51 (95% CI: 0.43-0.60) respectively, while the incidences of some other cancers were also significantly different as shown in Figure 3. Among males, the incidences of cancers of the lung, liver, colonrectum, and other cancers, as shown in Figure 3A, were significantly different between the two populations. Independent of sex and year of diagnosis, the incidence of lung, liver, NHL, and colorectal cancers was lower in PMSP compared to BPSP, with odds ratios of 0.75 (95% CI: 0.65-0.85), 0.74 (95% CI: 0.62-0.88), 0.74 (95% CI: 0.60-0.91), and 0.67 (95% CI: 0.56-0.78), respectively. However, the rate of cancer incidence increase, e.g., that of the liver, in PMSP was higher than that in BPSP and it reached the same rate in the last 5-year period.
In the comparison of rates using the SIR method, we used the incidence in Songkhla (2004Songkhla ( -2007 reported in the 'Cancer Incidence in Five Continents Vol. X' as the standard (reference) incidence. The expected rates for PMSP and PBSP in Songkhla, the whole Thailand, Penang Malay, and Singapore Malay populations were plotted for those cancers that showed some difference among the five populations, i.e., cancers of the liver and bile duct (female), lung, prostate, oral (male), breast, cervix and thyroid (female) [ Figure 4]. Other cancers without significant difference among the populations of interest were not plotted.

Discussion
It was found that, in the past, the incidence rates of some cancers in both males and females among PMSP were significantly lower than those of PBSP, but the rate of incidence increase for some cancers was higher in PMSP and caused the ASRs of both populations in the last 5-year period to be similar (Figure 3). The underlying reasons need to be identified. One among the possible explanations may be under-registration among MPSP in the first decade of the registration period. After the 2002 healthcare reform in Thailand, the improvement of healthcare facilities, settlement of the patient referral system, and establishment of the country-wide patient database have led to a better registration of cancer patients in PMSP. Due to the possibility of the improved patient catchment as a result of the registry, the feature of cancers in the last 5-year period, as shown in Figure 2, better explains the true incidence of cancer among PMSP.
Though the time period of cancer incidence among the Penang Malay population is almost 5 years earlier than that of this report, it is the latest official publication available in CI5 vol. X. The higher incidence of cancers of the colonrectum, stomach, breast, ovary and corpus uteri among the Penang Malay population than that of both MPSP and BPSP points to the trends towards the urbanization of Penang, which is a highly developed province in Malaysia. The lower incidence of tobacco-related cancers in the Penang Malay population such as that of the oral cavity, pharynx, esophagus, and larynx is rather interesting, since a recent report from Malaysia indicated a smoking prevalence of 55.9% among Malay ethnics aged 18 or above (Lim, et al., 2013). However, WHO states that the lowest smoking prevalence in Malaysia is among urban Penang and Sarawak populations (WHO-WEPRO, 2014). A correlation between those cancers and smoking prevalence in the past decades is needed to confirm such an association. The incidence of lung cancer among Singapore Malays was slightly higher than that among PMSP and PBSP in Songkhla (Figure 4), but without statistical significance. The smoking prevalence among Southern Thai males aged 15 years or above in 2008 was reported at 48.2% (National Health Examination Survey, 2009). Unfortunately, data on smoking prevalence in Songkhla in the previous few decades are not available.
Liver cancer was much lower in incidence among the Penang Malay population than our study population. These findings are worth comparing and studying in detail in order to find the underlying reasons for the difference. The reasons behind the lower incidence of female liver cancer among PMSP than that of PBSP are not clear. It should be noted that cholangiocarcinoma (CCA), not hepatocellular carcinoma (HCC), was the major histologic type of female liver cancer in Songkhla. At present, the incidence of CCA outnumbers that of HCC in both sexes and both subpopulations in Songkhla, and the decrease in HCC incidence among young people is evident (unpublished analysis).
The Poisson regression modelling ( Figure 3) and SIR (Figure 4) methods resulted in the same conclusion regarding the difference in rates for cancers of the cervix and female breast among both subpopulations, but the results were different in other cancers. One of the reasons for these differences in sensitivity is the use of long-term data in the regression models, where the SIR method used only a single 5-year period point to compare. Another reason is the convergence of trends in those cancers in the last 5-year period among the two populations, as can be obviously seen in Figure 3A. Thus, the evidence seems to demonstrate that only cervix and female breast cancers were statistically different among PMSP and PBSP, while the incidences of other cancers was different in the far past but converging in the recent years.
It should be noted that oral cancer incidence was high in both subpopulations compared to the Penang and Singapore Malay populations (Figure 4). It is not easy to explain this phenomenon. Malay ethnics and Islam do not seem to be the good explanations. The risk factor(s) for male oral cancer in Songkhla should affect both Buddhists and Muslims at the same degree. The lower incidence of prostate cancer among PMSP than PBSP seemed to be related to the distance to the urban centers of Songkhla, namely Hat Yai Municipality and Songkhla City. Among females, PMSP had lower cancer incidences than PBSP. The phenomenon might be related to the lower degree of awareness and accessibility to screening facilities, which are more prevalent in the urban areas of Hat Yai and Songkhla cities. However, the incidence of cervical cancer was as low as 10 per 10 5 population among Muslim women, i.e, lower than among Buddhist women. Nevertheless, its rate is continuously decreasing due to better awareness and utilization of cervical cancer screening facilities. Thus, the incidence of cervical cancer in both populations will most likely be at the same level in the near future.
In conclusion, in Songkhla, the incidence of some cancers is generally lower among Muslims than Buddhists. The difference in cancer incidence as well as in religionrelated culture between the two subpopulations suffices consideration of a specific surveillance and monitoring system and two sets of cancer control plans and goals, especially for cervical cancer, in order for them to fit the unique population context of the deep Southern Thailand.