This study shows extremely few cases of ovarian cancer in East Africa compared to Alberta population, using data from the Alberta cancer registry. There were limitations, and likely but unknown bias influencing the results. The age standardized incidence rate could not be calculated for the East African population, because of the lack of census data. On the one hand, a large proportion of the female population in East Africa may not yet be at an age of risk for ovarian cancer. On the other hand, the proportion of older females which carry the highest risk is likely still lower in East Africa compared to Alberta (12). The WHO 2014 cancer country profiles show that the life expectancy of Canadian women is 20 years higher compared to East African women. The larger population per cancer center in East Africa and other impediments such as access to health care facilities will most likely result in under-diagnosis of cases. For example, older women in rural areas may die of undiagnosed disease, unaware of the significance of the symptoms, and lacking the resources to seek medical attention(9). While our data provide a first benchmark of ovarian cancer incidences in East Africa, these limitations lead us to assume that cases from selected areas in East Africa may not be representative of the whole population. There is a need to establish cancer registries and census data in Eastern Africa similar to those in developed countries.
Ovarian cancer shows a wide geographical variation with higher incidence in North America compared to Asia and Africa (13). In our study, the Alberta population had more cases of epithelia cancer (89%) compared to East Africa (60%). In contrast, germ cell tumors and sex cord stromal tumors were more common in East Africa compared to the Alberta population. A large population-based study (CONCORD-2) reported a high proportion of germ cell tumors in Asia (4.2%), compared to Europe (1.3%), and North America (2.0%). We report a high proportion of germ cell tumors in East Africa (12.5%) similar to the one reported in Russia (11.4%)(14). Similarly, the proportion of sex cord stromal tumors was 14.3% for East Africa, and 0.92% for the Alberta population. Because germ cell and sex cord stromal tumors occur at a younger age, this could reflect the age distribution of the underlying population and a selection bias. There is a possibility that younger women with germ cell tumor or sex cord stromal tumor might seek medical attention and receive treatment, while older women with advanced epithelial cancer may not.
Morphology-based diagnosis of cancer remains a mainstay in developing countries, and ancillary immunohistochemical tests are rarely available. The diagnostic accuracy of ovarian cancer in East Africa showed a substantial agreement with the diagnosis of major categories of ovarian cancer, but was only fair for specific histotypes. This is a remarkable fact, given the poor laboratory resources in these settings, compared to North America, exemplified by the tremendous differences in the quality of the basic H&E stains. The fact that pathologists in East Africa are able to make highly accurate diagnosis with virtually zero contrast H&E stains is an example of the remarkable adaption and ingenuity of the human brain. There is no reason to believe that they could not achieve a similar diagnostic accuracy regarding more specific histotyping with simple standardization of the H&E staining protocol.
In the past 15 years, there has been a significant change in the understanding and diagnosis of ovarian cancer, such that histotype has emerged as an important prognostic and predictive marker (15, 16). Historically, reproducibility of diagnosis of ovarian carcinoma based on the cell types (17–19) was low, but nowadays, the inter-observer reproducibility is very good, and, if not excellent, can be improved by the use of ancillary immunohistochemistry tests (20). In this study, the morphological diagnosis of ovarian carcinoma was only fair, likely a result of a failure to use the current criteria for tumor cell types in a routine practice,(21) and the majority of the cases of carcinoma were unspecified. Significant numbers of unspecified carcinomas (~ 43%) in this study were reclassified as HGSC in a revised diagnosis, and these are the diagnoses likely made before 2014, where serous carcinoma were separated into HGSC and LGSC. For instance, a study performed in Canada necessitated a review of diagnoses; the majority (78%) of unspecified carcinoma based on 2003 WHO classification were reclassified to HGSC (22). The problematic areas in ovarian carcinoma histotyping remain the differentiation of HGSC versus high grade EC and vice versa (6, 20, 23). In the current study, 29% of the EC were reclassified to HGSC, contrary to 17% seen in a study carried out in Canada (22).
Mucinous carcinoma showed a poor concordance between the original diagnosis and the revised diagnosis, and significant number (26%) were reclassified to EC, whereas 17% were metastatic gastrointestinal neoplasms. Differentiating primary ovarian from metastatic carcinoma has been an area of challenge. In previous years, the diagnosis of MC was made frequently, constituting up to 14% of ovarian carcinoma (24). Currently, with an improved understanding of ovarian carcinoma histotypes, and the use of ancillary tests, MC has become a rare subset of ovarian carcinoma, making it less than 5% of the cases (25–27). In this study, we used SATB2, a recently identified marker, for differentiating primary ovarian from secondary colorectal/appendiceal tumors (28, 29). The tumors which showed mucinous, or endometrioid like morphology suspicious for metastasis with expression of SATB2 were considered as metastatic.
In this study, the cases of MC which were reclassified as EC were probably related to failure of recognition of confirmatory endometrioid features or overreliance on features like mucinous differentiation which do occur in EC, (30, 31) and mucinous carcinoma with unapparent intracytoplasmic mucin can also mimic EC (27). Also, 2 cases initially diagnosed as CCC were actually endometrioid carcinoma with clear cell changes(30).
For non-epithelia tumors, the reproducibility in assigning tumor types was substantial, with the highest concordance in lymphomas (92%), followed by sex cord stromal tumors (83%), and in the germ cell tumors 81%. We did not go into specific histological types in these smaller categories because of lower case numbers, but noted a high number of “primary” ovarian lymphomas.
This study identified one case of ovarian small cell carcinoma hypercalcemic type (SCCOHT) which was originally diagnosed as a neuroblastoma. This is a lethal cancer which affects young women, characterized by aggressive clinical course, poor prognosis and a recently discovered pathognomonic mutation in SMARCA4 (32–34). The reasons for such diagnosis could be the morphology of a round dark cell tumor which can easily attract a differential diagnosis of neuroblastoma in a setting where IHC tests are not routinely performed. Yet, obarian neuroblastoma at the age of 30 years is quite uncommon, although it has been reported as primary ovarian or associated with mature teratoma in a few cases (35, 36). Lack of awareness for the existence of this diagnosis in the current classification was a likely contributing factor, as well as the rarity of this tumor, noting that most of the pathologists would see at most one such case during their practice(21). Foremost, this case illustrates the inability to render correct diagnosis without access to confirmatory molecular testing for molecularly defined rare cancers.
This study applied various IHC markers used in ovarian cancer diagnosis and histotyping. The expression patterns followed the expected trend with some differences. TP53 mutations are ubiquitously present in HGSC (37, 38) and optimized p53 immunohistochemistry is used as a surrogate marker of TP53 mutation (38, 39). A notable difference was seen in the frequency of mutant p53 staining; only 63% of HGSC had mutant p53, whereas, in EC, it was seen in 18% of the cases as expected.(22) The low rate of abnormal p53 seen in HGSC could be due to various reasons. Some cases lost the internal control, which was even commonly observed with ARID1A. In this study, we used a p53 optimized antibody, and a standardized staining platform, as well as known positive and negative controls. Therefore, poor staining and loss of internal control cannot be explained by the analytical process; the issue probably occurred in pre-analytical phases specifically related to tissue fixation. In Tanzania, there is no uniform, standardized protocol for tissue fixation. The samples are sent from various hospitals to the pathology centres, and the quality of formalin and time for fixation vary considerably. Furthemore, FFPE tissue blocks are not stored in a controlled environment.
Hormonal receptor expression in EC was also lower compared to the previous studies, (40, 41) and this could probably be attributed to weak staining or a biological behavior of the EC which cannot be justified. Low hormonal receptor expression is common in high grade EC, (42) but, in this study, tumor grades were equally distributed across all EC cases. Interestingly, we observed only one case (3%) (41 year old patient with endometrioid carcinoma) that showed abnormal mismatch repair, contrary to 13% as reported previously (43). MSH6 was lost indicating probably Lynch syndrome. This indicates that Lynch syndrome likely exists in East Africa, but larger studies on endometrioid carcinomas are needed to assess its prevalence in the East African population.