Introduction

Breast cancer is one of the most common cancers with leading causes of cancer mortality among women worldwide [1]. It is also the most common cancer affecting women in the Sub-Saharan region including Ethiopia. In Ethiopia, breast cancer commonly affects the younger women as compared to older women in low-income countries. Moreover, this cancer is also the leading cancer mortality among women in developing countries as well as in Ethiopia [2, 3].

There are multiple factors contributing to a higher breast cancer mortality rate (e.g., staging at diagnosis, type of breast cancer, therapeutic advance and compliance to breast cancer treatment) [4, 5]. Stage of breast cancer at diagnosis is the main predictor of breast cancer mortality [2]. There is stark gap in breast cancer survival rates between low- and high-income countries. Women in low-income countries suffer from disproportionately higher rates of breast cancer mortality in comparison to high-income countries. Breast cancer survival rates are increasing in developed countries because of early-stage diagnosis with the help of community awareness, wide spread screening, and advances in treatment options [4, 5]. According to the WHO, at least 60% of breast cancer should be diagnosed at early-stage to achieve a reduction of 2.5% in breast cancer mortality every year by 2040 [6].

In Ethiopia, 71% of breast cancer patients are diagnosed late, stages III and IV, leading to an increase in the likelihood of mortality [3, 7, 8]. Adequate management of all stages breast cancer requires a hospital facility that includes a multidisciplinary team (surgery, medical oncology, radiation oncology) to reduce mortality; which is lacking in most Ethiopian hospitals [3]. Therefore, in low-income countries like Ethiopia, early detection of breast cancer is vital to reduce breast cancer mortality. There is no country level summary data in Ethiopia on what proportion of breast cancer is detected in the early-stages in Ethiopia [9,10,11,12]. Therefore, the aim of this study is to conduct a systematic review and meta-analysis on the proportion of early-stage breast cancer at diagnosis in Ethiopia.

Methods

Search strategies

This systematic review and meta-analysis was conducted to estimate the pooled proportion of early breast cancer at diagnosis in Ethiopia. The protocol for this review was registered in the International Prospective Register of Systematic Reviews (PROSPERO) the University of York Centre for Reviews and Dissemination (registration number 339368),) on June 3rd, 2022. To adhere to the scientific standard, the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines were used [13, 14].

International databases: MEDLINE through Pub-Med, Google Scholar, Science Direct, HINARI and medley were systematically searched. Also, references of identified articles were searched to increase the chance of detecting missed articles in grey literature. To search relevant articles for this study, we used the following Mesh terms: “Breast” OR “Lactation” OR “Mammoplasty” OR “Mastectomy” OR “Mammography” OR “Mammary Glands, Animal” AND “Neoplasms” OR “Cancer” OR “Malignancy” OR “Tumors” OR “Tumor” OR “Malignant Neoplasms” OR “Malignant Neoplasm” OR “Neoplasm” OR “Neoplasia” AND “Ethiopia”. The key terms were used separately and/or in combination using Boolean operators like: “OR” or “AND”. The literature search from those databases was done from May 22nd to June 21st, 2022. All papers published until June 21st, 2022 were included in this systematic review and meta-analysis. Following every search, all identified citations were collated and uploaded into EndNote version 4/2020 and duplicates removed. Following a pilot test, titles and abstracts were then screened by two independent authors (KB and GB) for assessment against the inclusion and exclusion criteria for the review (Table 1).

Table 1 Inclusion and exclusion criteria on study conducted early-stage breast cancer in Ethiopia
Full size table

Potentially relevant studies were retrieved in full document. The full texts of selected citations were assessed in detail against the inclusion criteria by KB and GB. Reasons for excluding of papers at full text that did not meet the inclusion criteria were recorded and reported in the systematic review. Any disagreements that arose between the authors at stage of search, title and abstract screening, as well as full text screening of the selection process was resolved through discussion. If no resolution was found, we used an additional third author (FN). The results of the search and the study inclusion process were reported in the final systematic review and presented in our PRISMA flow diagram (Fig. 1).

Fig. 1
figure 1

PRISMA flow diagram of included studies which shows process of selection possible

Full size image

Eligibility criteria

Participants

We included all studies, which were conducted in Ethiopia and describe the stage of breast cancer on initial diagnosis

Study designs

For this review, we included all observational studies reporting stages of breast cancer conducted in Ethiopia.

Language

We included all articles reported in English language only.

Setting

The setting was restricted to Ethiopia.

Exclusion criteria

Articles that used any other staging system different from TNM staging or did not mention the staging of breast cancer were excluded. Additionally, studies that did not include stage IV breast cancer were excluded from this review.

Outcome of interest

Primarily, this study aimed to estimate the pooled data on what proportion of breast cancer was diagnosed at the early-stage in Ethiopia. The proportion of early breast cancer patients was calculated by dividing the number of early breast cancer diagnoses (stage 1 and stage 2) to the total amount of breast cancer patients for whom TNM staging system was done [15].

Operational definition

Breast cancer staging was usually done by TNM staging system. We used the term “early-stage” of breast cancer for stage I and stage II tumors and “advanced stage” of breast cancer for stage III and stage IV tumors based on the TNM staging system [6, 16].

Data extraction

KB and GB independently extracted all the data required using a standardized data extraction format on excel. The data extraction format included the primary author, year of publication, study period, site of studies, regional state, study design, sample size, response rate, proportion of early breast cancer and total breast cancer for whom staging was done.

Quality assessment

Newcastle–Ottawa Scale (NOS) quality assessment tool for observational studies was adapted to assess the quality of each independent study. The tool has three main sections. The section of the tool graded out of 5 stars and mainly focuses on the methodological quality of each original study. The second component of the tool graded out of 2 stars and mainly focused on the comparability of each study. The last section of the tool graded out of 3 stars and was used to assess the outcomes and statistical analysis of each original study used for data analysis [17]. KB and GB independently assessed the quality of included research articles using the stated tool. Differences in the scoring of articles between the two reviewers were addressed by discussion. For the differences that could not be resolved by agreement between the two authors, we used a third author (FN) as a result. After reviewing various literatures, we declared that articles scoring ≥ 7 points out of 10 stars were considered to be good quality and therefore included in this study.

Data management and analysis

Microsoft Excel spreadsheets were used for data extraction while STATA Version 16 and compressive meta-analysis (CMA) software was used for data analysis. The descriptive data was presented using a table to describe the characteristics of all included studies. For analysis we used the random effects model. The magnitude of each original study as well as the overall magnitude of pooled proportion is described using a forest plot graph. The horizontal line of the forest plot shows the 95% CI, and the blue box represents the weight of each study. Presence of heterogeneity was confirmed by Cochrane Q-statistics and the degree of heterogeneity was explored by using predictive interval (PI) using 95%. In identifying the degree of heterogeneity, a predictive interval was used to evaluate the dispersion or variation of effect size across included studies. I2 statistics is a proportion not an absolute index for variability but can be used to describe what proportion of the observed variance reflects variation in true effects rather than due to sampling error or random error. It cannot be used to make a conclusion on the heterogeneity of the disease burden but, can provide estimates (Borenstein M, Higgins J, Rothstein HR, Hedges LV: I 2 is not an absolute measure of heterogeneity in a meta-analysis, Unpublished). To identify the source of heterogeneity we did a subgroup analysis based on study design, site of study, years of publication and sample size. Additionally, meta-regression was conducted using sample size and year of publication as study level covariate.

Publication bias was explored by using a funnel plot to assess symmetry by observation and statistically, we used Egger’s test to objectively assess the presence or absence of publication bias or small study effect.

Results

Characteristics of the included studies

A systematic search of the electronic databases yielded a total of 288 articles. After the removal of duplicates (n = 92), articles were screened by title and abstract. A total of 138 articles were excluded based on non-compliance with their titles and abstracts. As described in the Fig. 1, a total of 57 full articles were accessed and screened based on compliance with the inclusion and exclusion criteria. Finally, 43 research articles fulfilled the inclusion criteria from which 2 articles were excluded because of repetition of same sample for a different study objective and 41 of them were checked and fulfilled the optimal quality standard evaluated by the NOS tool. In this systematic review and meta-analysis, 41 studies published between 2006 and 2022 were used to estimate the pooled proportion of early breast cancer at diagnosis conducted in Ethiopia. A total of 10,123 breast cancer patients were included in the study. All original studies estimated the proportion of early-stage of breast cancer using an observational study design. The sample size of each study ranged from 13 to 787 (Table 2).

Table 2 Descriptive summary of original included studies in systematic review and meta-analysis of the proportion of early breast cancer detection at diagnosis in Ethiopia
Full size table

Meta -analysis

This meta-analysis found that the pooled proportion of early-stage breast cancer among all breast cancer patients for whom TNM staging was carried out at diagnosis in Ethiopian hospital was found to be 36% (95% CI: 31%—41%) and 95% PI  was ranging from 28 to 45% (Fig. 2).

Fig. 2
figure 2

Forest plot showing pooled proportion of early breast cancer detection from 41 studies on diagnosis among Ethiopian hospital

Full size image

Degree of heterogeneity (dispersion of effect size) is best measured by predictive interval. By using estimate of between-study variance Tau^2 = 0.02 and upper confidence level 0 0.41 (upper confidence level of pooled proportion), we calculated 95%, PI by using CMA software and it was found to be in range of 0.28 to 0.45 (28%—45%). This shows that if studies were carried out on proportion of early-stage breast cancer detection in Ethiopian hospital in 95% of the case proportion of those studies falls between 28% and 45. This 95% PI range (28%—45%) is low for early breast cancer proportion at diagnosis. Hence, there is only a mild degree of heterogeneity (Fig. 3).

Fig. 3
figure 3

The pooled proportion of early breast cancer at diagnosis is 0.36 with a 95% confidence interval of 0.31 to 0.41

Full size image

Subgroup analysis and meta regression

In this systematic review and meta-analysis, both subgroup analysis and meta regression were carried out. Accordingly, subgroup analysis described in Table 3 and meta regression described in Table 4.

Table 3 Subgroup analysis based on four study level moderator using predictive interval to assess source of heterogeneity
Full size table
Table 4 Meta-regression of moderators related to the heterogeneity of included studies in estimating the pooled proportion of early breast cancer at diagnosis among Ethiopia hospital
Full size table

Publication bias

Publication bias was checked graphically and statistically. On visual inspection of standard funnel plot, it looks there was publication bias (Fig. 4). As the result publication bias was statically evaluated by using the Egger test and it showed that there was no publication bias (p = 0.1486).

Fig. 4
figure 4

Funnel plot testing publication bias (random, N = 41)

Full size image

Discussion

This systematic review and meta-analysis was conducted to estimate the proportion of early-stage breast cancer at diagnosis among Ethiopian hospitals. Early-stage at diagnosis is one of strongest determinants of survival after breast cancer treatment. When breast cancer is diagnosed at early-stages, there is a higher chance of good outcomes and a higher survival rates [5]. This study showed the pooled prevalence of early-stage breast cancer at diagnosis in Ethiopia was 36.0% which is lower than the target set by GBCI of 60% of breast cancers diagnosed at stages I and II [6]. This pooled proportion of early-stage breast cancer at diagnosis in our study is slightly higher than what was reported in a systematic review and meta-analysis in 17 Sub-Saharan African countries in 2014 (33%). The difference might be because of time and population difference [5]. This low proportion of early-stage breast cancer at diagnosis in Ethiopia contribute to high mortality rates from breast cancer and hence makes it difficult to attain the breast cancer annual mortality reduction target of 2.5% by 2040 envisioned by WHO [6].

Subgroup analysis by site of study showed that there is a slightly higher proportion of early-stage breast cancer detection (39.9%) in Addis Ababa, the capital city, as compared to outside Addis Ababa (29%). This variation between study sites could relate to a difference in the community breast cancer awareness and the availability of diagnostic facilities in Addis Ababa as compared areas outside of the capital city.

There are well-established strategies to improve the early detection of breast cancer. Among these, community health education and awareness about early signs and symptoms of breast cancer, health professionals education at primary healthcare centers on these early signs and breast cancer screening are important [6].

Since breast cancer advanced care is limited to few hospitals in Ethiopia, the Government should adopt and work practically to implement some of said strategies at the primary healthcare level to facilitate the early detection of breast cancer cases. The Ministry of Health must focus on addressing the delays in diagnosis of symptomatic breast cancers to reduce the current high mortality rates breast cancer in Ethiopia. Additionally, there is also strong need for expansion of mammography for the screening of breast cancer to detect it at the early-stage in order to lower the number of preventable deaths of women with a breast cancer diagnosis. We recommend that the Ministry of Health give special focus on methods of early diagnoses of breast cancer to reduce the current high mortality rates of women with breast cancer in Ethiopia. Additionally, in settings with diagnostic capacity and access to effective cancer treatment, screening mammography could be considered.

Conclusion

Early breast cancer detection rates in Ethiopia are very low. To reduce breast cancer mortality rates to meet GBCI’s targets, the Ethiopian government should work strongly on the early detection of breast cancer. The Ethiopian Ministry of Health, different non-governmental organizations, Ethiopian Women’s Affair and Health Workers, and health care professionals should work on incorporating evidence-based methods of early breast cancer detection in Ethiopia.