Clustering time trends of breast cancer incidence in Africa: a 27-year longitudinal study in 53 countries

Background Breast cancer is the most common, frequently diagnosed cancer with the highest incidence among female worldwide. Although the incidence is decreasing in developed countries, it is on increase in most of the African countries. Objective This study aimed to identify different time trends of breast cancer incidence among African countries using latent mixture approach. Methods The information includes newly diagnosed breast cancer patients per 100,000 women for 53 African countries in a period of 1990–2016. Latent mixture modeling was performed in Mplus 7.4 software. Results The overall trend of breast cancer in Africa was increasing. Latent mixture model with 5 clusters was estimated as the best using fit indices and linear growth trajectories were specified for each cluster. Nigeria was the only country which belongs to a cluster with negative slope indicating a slow decrease in the breast cancer incidence; also, Seychelles was the only country that showed a sharp increase over time. 31 countries belonged to a cluster with a slope of 0.08, indicating that the incidence of breast cancer is almost constant over time. Cluster 3 including Algeria, Angola, Botswana, Central African Republic, Cote d'lvoire, Equatorial Guinea, Lesotho, Libya, Namibia, Somalia, Sudan, Swaziland, Uganda and Zimbabwe and cluster 2 including Gabon, Mauritius, Morocco, South Africa, Tunisia and Congo showed a slow and moderate increase in the incidence of breast cancer, respectively. Conclusion Providing health education programs is essential in African countries with rising trend of breast cancer during the last decades.


Annually 1.3 million cancer patients are identified
worldwide. At present, cancer is known as one of the most important issues in public health because in developed countries more than 50% of the burden of diseases is related to cancer and in developing countries the incidence of cancer is increasing 1 . Breast cancer is the most common and frequently diagnosed cancer with the highest incidence and eminent cause of cancer death among females worldwide [2][3][4][5][6][7] .
In year 2012, 1.7 million cases and more than 500,000 deaths due to breast cancer have been reported 5,6,8 . Although the incidence is decreasing in developed countries, breast cancer is increasing in most of the African countries 8,9 . The International agency for cancer research (IARC) announced that the incidence rate of female's breast cancer is 39 per 100 000 in southern Africa and 27 per 100 000 in central Afria in 2012 2 . The burden of breast cancer will be double in Africa by 2030 because of the increase in the population growth and unhealthy lifestyle 2 . Recently, modeling time trends in health sciences has been more attentive. Analyzing time trends requires extensive datasets including repeated measures in a longitudinal design and also a practical model to describe complex growth patterns 10 . Latent mixture approach as the most appropriate model in this setting tries to capture unobserved heterogeneity, using categorical latent variables 10,11 . In the current manuscript, we call this categorical latent variable as cluster. Despite the numerous studies about the incidence and mortality of breast cancer in African countries, lack of studies about heterogeneity of breast cancer temporal trends among African countries is obvious. Therefore, this study aimed to identify different time trends of breast cancer incidence among African countries using latent mixture model approach.

Materials and methods
The information including newly diagnosed breast cancer patients per 100,000 women for 53 African countries which was reported in Gapminder web site (Stockholm, Sweden. Freely available at www.gapminder.org/ data) in a period of 1990-2016 was collected. Estimation of number of clusters is the most important issue in latent mixture modeling. To address this problem, we used likelihood ratio test and due to small sample size, P-value <0.1 was considered as statistically significant. Latent mixture modeling was performed in Mplus 7.4 software which uses maximum likelihood estimator based on EM algorithm.

Results
Descriptive statistics of breast cancer incidence rate for African countries from 1990 to 2016 (in 5-year intervals) are summarized in Table 1 and growth trajectory for each country is displayed in Figure 1. Although these rates indicate an overall increasing trend, trajectories show different trends among African countries so latent mixture modeling was utilized to divide the countries into different clusters based on their incidence trend over time.   Table 2 and 5-cluster model was selected as the best based on the solutions of likelihood ratio test (P-value=0.08). Based on 5-cluster model, linear growth trajectories with differ-ent intercepts and trends were specified for all of the clusters, as reported in Table 3. Besides, the entropy of 0.89 revealed good quality for the membership of latent classes. Overall mean and estimated linear trend of each trajectory are displayed in Figure 2. Swaziland, Uganda and Zimbabwe) and can be defined as having a moderate and slow increase in the incidence rate over time, respectively. Seychelles was the only country which belonged to cluster 4 with slope 1.15 indicating a sharp increase over time and other 31 countries belonged to cluster 5 with a slope of 0.08, indicating that incidence of breast cancer is almost constant over time (Table 3).  13 . The difference between the aforementioned study and our findings might be due to assessment of an overall long term trend in the current study. There were also some related articles which have reported different patterns in different population or periods of time. Elima et al. analyzed data from 2 population Ibadan and Abuja in two years (2009 to 2010) and showed that the incidence of BC as increased but cervical cancer had stable incidence 14 . Besides, the results of a research on black African population in Nigeria showed that the incidence rate was 63.7 to 130.9 in two periods (1971)(1972)(1973)(1974)(1975)(1976)(1977)(1978)(1979)(1980)(1981)(1982)(1983)(1984)(1985)(1986)(1987)(1988)(1989)(1990). The incidence rose in 1978 because of the low level of the people's awareness and negligence of the consultant. Also, in 1982 in collaboration with TV and radio for increasing awareness, we see raising incidence and in 1986 because of development of economic conditions BC incidence decreased. In 1987, because of the use of radiotherapy in the hospital Ibadan, the incidence rate increased 15 .
Although studies about the incidence of BC in Seychelles is rare, the observed sharp increasing trend may be due to greater proportion of early stage diagnosis as a results of development and better screening compared to other African countries 16 .
Our findings about moderate increase of incidence in Gabon, Mauritius, Morocco, South Africa, Tunisia and Congo are consistent with most of the related articles in these countries. Khalis 21 .
Formal statistics about the incidence of BC in Algeria from cancer registry of Setif confirmed the increasing trend from 10.4 in 1987-1989 to 24.8 in 2005-2007 which was slightly higher than the reported rate for cluster of Algeria in this study 22 . The observed increase may be due to the prevalence of risk factors like obesity, level of sex hormones, late child bring, having fewer child and unhealthy diet 23  Using an advanced statistical model and investigation of longitudinal trajectories in a large scale are two key features of the current study. As a limitation, there was no related information about a few number of African countries. Investigation of the incidence of breast cancer in different age categories or mortality patterns of breast cancer in Africa is highly suggested for future studies.

Conclusion
The incidence of breast cancer was increased in most of the African countries. Therefore, providing health education programs is essential in countries with rising trend of breast cancer during the last decades.