Clustering 46 Asian countries according to the trend of breast cancer incidence rate from 1990- 2016: An application of growth mixture model

Breast cancer among Asian countries is the second cause of cancer death and remains as a challenging issue in woman health as 39% of all new cases of breast cancer have been diagnosed in Asia. The current study was devoted to recognize different patterns of breast cancer incidence rate among Asian countries. Information about the incidence rates of female breast cancer within 1990-2016 years was extracted from Gapminder web site and growth mixture model was developed to describe the growth patterns and provide a set of tools to investigate the individual differences in change. Our findings suggest an overall increasing trend throughout the continent, but individual trajectories show different behaviors among countries. Bayesian information creation showed that 3-cluster model was the best. Cluster one countries including: Bangladesh, Israel, Kyrgyz republic, Maldives, Nepal, North Korea, Tajikistan and Timor-Leste, the slope of -0.13 suggests a slight negative trend for the incidence rate of breast cancer. 17 countries including: Armenia, Bahrain, Brunei, Cyprus, Iraq, Japan, Jordan, Kazakhstan, Kuwait, Lebanon, Malaysia, Pakistan, Philippines, Qatar, Singapore, South Korea, United Arab Emirate which belongs to cluster 2, had not only a high intercept that means higher amounts of incidence rate in 1990 year, but also a slope of 0.96, indicating a sharp increase trajectory. Also slope of 0.38 showed a slow increase in the incidence rate of breast cancer over time for another 21 countries. In conclusion, increase of breast cancer incidence among cluster 2 countries is tremendous therefore effective strategies for prevention are urgently needed.


Background
Breast cancer (BF) is known as the most common cancer among women in several countries worldwide [1] [2]. Globally, it includes 23% of all cases of cancers and 14% of the cause of cancer death in women [3]. According to the 2012 GLOBOCAN, the incidence of age standardized BF is 73.4 per 100000 in developed countries and 31.3 per 100000 in developing countries [4].
44% of deaths and 39% of all new cases of breast cancer have been diagnosed in Asia.
Approximately, 25% of women's cancer in India is breast cancer [5]. Breast cancer among Asian countries is the second cause of cancer death and remains as a challenging issue in woman health [6]. In Arab countries like Qatar, Saudi Arabia, Iraq and Bahrain, breast cancer is a crucial issue [7]. In 3 Saudi Arabia, the International Research Agency announced that the incidence rate of breast cancer in 2009 was 22.4 per 100000 in women and mortality rate was 10.4 per 100000 [8]. BC is also the most common cancer in China and the fifth cause of women's death due to cancer [9] as in 2015, 268,600 new cases and 69,500 deaths because of breast cancer were reported in China [10].
Although the trend of breast cancer incidence across most of the Asian countries is increasing, heterogeneity among countries is particularly apparent due to different health policies, regional behaviors, and genetic and lifestyle variety [6]. While the incidence rate of breast cancer has doubled or more in Korea, Singapore, Hong Kong and Japan, besides socioeconomic advancement [10] in the last decades [6, 11], a substantial increase was found in China (2.96% per year) [12], Thailand [13] and Taiwan [3]; also, a gradual increase was reported in Sri Lanka [14], Pakistan [15], India [16], and Iraq [17].
Growth mixture modeling as one of the well-known and practical statistical methods is capable of dividing a heterogeneous population into multiple homogeneous clusters based on their temporal trends (detecting, growth mixture). To the best of our knowledge, there are a few studies about diversity of temporal trends of breast cancer in Asia, and previous studies are limited to a country or regional location. Therefore, the current study was devoted to cluster most of the Asian countries according to the trend of breast cancer incidence rate for the 27-year period (1990-2016) via growth mixture models to show longitudinal heterogeneity among the countries.

Methods
In the current study, information about the incidence rates of female breast cancer (new cases per 100,000 women) within 1990-2016 years was extracted from Gapminder web site for 46 Asian countries. Gapminder is a non-profit venture registered in Stockholm, Sweden and the current dataset provided by the World Health Organization in Gapminder.
Growth mixture model (GMM) was developed to describe the growth patterns and provide a set of tools to investigate the individual differences in change. GMM tries to explore latent profiles of temporal trends by incorporating the mixed models to consider the intra-correlation of individuals in repeated measures. This method also identifies and discriminates heterogeneous unobservable 4 clusters called latent groups using the latent models (Growth mixture model, GMM paper). In GMM, estimating the optimal number of latent classes is a critical issue. In the current manuscript, Bayesian information creation (BIC) was used to estimate the optimum number of latent clusters; all the statistical analyses were performed in Mplus software, version 7.4.

Results
Visualized trend of breast cancer incidence rate among 46 Asian countries is shown in Figure 1 and related descriptive statistics in 5-year intervals from 1990 to 2016 are reported in Table 1. Reported results suggest an overall increasing trend throughout the continent, but individual trajectories show different behaviors among countries. To consider heterogeneity in growth trajectories and model these complex patterns, we implemented the growth mixture model.
Bayesian information creation (BIC) showed that 3-cluster model was the best model with minimum error (results not shown). Therefore, growth trajectories for 3-cluster model were fitted and for each cluster, intercept and slope of the incidence rate were obtained. As shown in Table 2, eight countries belonged to cluster 1, seventeen countries to cluster 2, and the other 21 countries belonged to cluster 3.
The entropy was another goodness of fit index which was in an acceptable level (entropy=0.96). which belongs to cluster 2, had not only a high intercept that means higher amounts of incidence 5 rate, but also a slope of 0.96, indicating a sharp increase trajectory (Table 2 and Figure 2).

Discussion
In our study on 46 Asian countries, three types of behavior were noticed. The first cluster consists of the countries with a slight negative trend for incidence rate of breast cancer incidence rate: Although the rates are different in various countries, Korea and Thailand had the highest (44.9% and 36.0%) and Philippines had the lowest (5.2%) incidence rates. Korea had the significantly increased incidence rates during the 10 year period. Four patterns in the incidence rates and age specific incidence in the study of Shin et al. were suggested. Low incidence with slow increase until age 50 years and then a gradual decrease were seen in rural China, Korea and Thailand. Intermediate incidence with increase until age 50 years and then slow decrease in the elderly is the second pattern in Japan and Taiwan. Hong Kong and Shanghai showed an intermediate incidence with rapid increase until age 50 years and then a platue. The last pattern is high incidence with rapid increase peaking after age 65 in Singapore and Philippines. This study claim that early age at menarche and late in menopause, delay in marriage, fewer children, changes in infant feeding, and lifestyle are the risk factors of breast cancer [18]. Based on a study conducted in Turkish Republic of North Cyprus that was done from 2007 to 2012, similar to our results, it was reported that the trend of BC incidence rate in Asian was rising. Hormonal therapy, fewer children, late menopause, early menarche, late reproductive and obesity are considered to be the risk factor of breast cancer [19].  [20]. The difference between the obtained results may be due to different periods of time. Also, previous studies about breast cancer incidence during 1993 to 2012 years showed that North Korea had a slight decrease and South Korea had a sharp increasing trend that was consistent with obtained results in the current study [18].
A study in Pakistan claimed that total projected breast cancer incidence will increase by approximately 23.1% in 2020 to 60.7% in 2025 and cases of breast cancer diagnosed in younger women aged 30 to 34 years will increase from 70.7 to 130.4% in 2020 and 2025 as compared to 2015. From 2004 to 2015, breast cancer rates were lower and stable in the age of 15 to 29 years, while the incidence rates for women aged 25 years and older were increasing overall with a slight variation. In 2004 to 2015, the incidence rates were the highest among those aged 60 to 64 years.
However, from 2016 to 2025 increases in breast cancer rates among women aged 55 to 64 years and older than 75 years were expected. Lifestyle factors like smoking, physical activity, westernized diet, illiteracy, and cultural and economic status are the influencing factors [21].
In line with our study, previous studies in Lebanon confirmed a high increase in breast cancer incidence. Lakkis et al. showed that age standardized rate of BC in Lebanon was 46.7 per 100000 in 1998 which increased to 52.5 in 2002. This increase is higher than several Asian countries like Iran, Malaysia and Japan. The reasons of increase may be related to improvement of NCR data collection and the rise in the use of mammography. Reproductive factors such as mean age at marriage of the women increased from 23.2 in 1970 to 27.5 in 1996, and also the fertility rate decrease from 4.4 to 2.5 during the same period. The age of menarche and menopause became younger and later, respectively. Hormone replacement therapy and obesity, smoking and nargileh use among Lebanese women were the other possible risk factors [22].
In a study done in Singapore, the overall incidence rate of breast cancer was totally increased from 7 Malaysians and 1.7% per year for Indians who were living in Singapore. Singapore's transition from an industrialized country to a developed country has brought about changes in lifestyle in the population: late marriage, delay in child bearing, smaller family sizes, and genetic interaction are the reason of the change in the breast cancer incidence in Singapore. Other risk factors like reproductive issues, height, dietary habits, obesity, duration of breast feeding, childhood growth, age at menarche and menopause make variations in different ethnic groups and could also influence breast cancer risks [23].  [26].
A study in India showed that the age standardized incidence rates of breast cancer increased from 24.8 in 1988 to 41.0 in 2012 in Delhi. Also, they claimed that a statistically significant increase in breast cancer was also seen in Delhi over 25 years. The same pattern was also reported among Indian urban registries. This trend has occurred due to adherence to more westernized lifestyle like adverse changes in diet, physical activity, and fertility and reproductive pattern [16].
Our findings showed a slow increase in the incidence rate of breast cancer in China that was approved in previous studies. Srilanka showed the increasing age standardized incidence of breast cancer from 9.2 in 2001 to 12.9 per 100000 in 2010. However, a slight drop in the incidence was observed in 2007. These trends are related to westernization of lifestyle such as the use of unhealthy food, inactive life that causes obesity, delay in childbearing, and duration of breastfeeding [14].
Changes in demographic and the risk profile of population are related to the increasing trend of BC in Thailand. In two surveys in 2004 and 2009, the percentage of overweight, diabetes or hypertension was found to be increasing. The rate of parity and total fertility was 6 in 1970, 3 in 1985, 2 in 1998,and 1.6 in 2020 [13].

Conclusion
In conclusion, increase of breast cancer incidence among cluster 2 countries is tremendous therefore 9 effective strategies for prevention are urgently needed.

Declarations
Ethics approval and consent to participate: not applicable Consent for publication: not applicable Availability of data and materials: data are available at Gapminder web site.
Competing interests: The authors declare that they have no competing interests Funding: this research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

12.
Huang Z, Wen W, Zheng Y, Gao Y, Wu C, Bao P, Wang C, Gu K, Peng P, Gong Y: Breast cancer incidence and mortality: trends over 40 years among women in  Figure 1 The trend of breast cancer incidence rate among 46 Asian countries 15 Figure 2 trend of marginal mean and estimated linear trend in different clusters