Precancerous cervical lesion in Ethiopia: systematic review and meta-analysis

The Preferred Reporting Items of Systematic Reviews and Meta-Analysis (PRISMA) checklist guideline was used to report the result of this systematic review and meta-analyses. In addition, the PRISMA flow chart was utilized to show the selection process of studies for a systematic review and meta-analysis [24].

This systematic review and meta-analysis was conducted to estimate the pooled prevalence of precancerous cervical lesion in Ethiopia. To conduct this study, all potentially relevant articles and gray literatures were meticulously searched. PubMed, Web of Science, SCOPUS, Science Direct, Google Scholar, African Index Medicus (AIM) and African Journals Online databases, and Addis Ababa and Bahir Dar Universities research repositories were searched using the following search terms: “Epidemiology,” “Burden,” “Magnitude,” “Prevalence,” “Cervical cancer,” “Cervix neoplasm,” “Cervix lesion,” “Pap Smear Positive,” “VIA Positive,” “Cervix precancerous lesion,” and “Ethiopia.” Search strings were developed using “AND” and “OR” Boolean operators. In addition to this, gray literature studies were searched from a research repository online library and a secondary search technique known as “footnote chasing” was utilized to identify additional articles from the included articles.

The search was carried out between August 28 and October 10, 2019, and all articles published until October 10, 2019, were included in the review.

This systematic review and meta-analysis encompassed studies conducted only in Ethiopia using the English language. Research articles published in scientific journals and gray literature that reported the prevalence of precancerous cervical lesion were included in the review. Excluded studies included any which focused on the assessment of knowledge, attitude, and practice towards CC without the outcome of interest of this study, program evaluation studies, studies with only abstracts, case studies, and qualitative studies.

Three authors (ASK, TD, and YB) extracted all necessary data by a standardized data extraction format using Microsoft Excel. The extracted parameters were primary author, publication year, region where the study was conducted, method of assessment, study design, mean age of the study participants, response rate, sample size, and prevalence of precancerous cervical lesion. Then, three authors (GD, AE, and YW) checked the data extraction process. Finally, five authors (BZ, EA, SK, GM, and WA) participated in approving the extraction process.

To appraise the quality of the included articles, we used the Joanna Briggs Institute (JBI) quality appraisal tool adapted for studies reporting prevalence data [25]. The following items were used to appraise cross-sectional studies: (1) inclusion criteria, (2) description of study subject and setting, (3) valid and reliable measurement of exposure, (4) objective and standard criteria used, (5) identification of confounder, (6) strategies to handle confounder, (7) outcome measurement, and (8) appropriate statistical analysis. Studies were considered low risk whenever fitted to 50% and/or above in the quality assessment checklist criteria [25, 26]. Using the aforementioned quality appraisal tools, no study was excluded (Table 1).

The data were analyzed using STATA version 15 Statistical Software. Heterogeneity across studies was checked using the inverse variance (I²) and Cochran Q statistics. The cut-offs of 25%, 50%, and 75% were used to declare the heterogeneity as low, moderate, and severe, respectively [44, 45].

As the preliminary output of the test statistics revealed a significant heterogeneity among studies (I² > 70%, P< 0.05), a random effects model was used to estimate the pooled prevalence of precancerous cervical lesion with a 95% confidence interval (CI). Subgroup analysis was also performed based on study area (regions) and human immunodeficiency virus (HIV) status in relation to the outcome variable. Funnel plot, Egger, and Begg tests at 5% significant level were employed to assess publication bias [46, 47].

Up to October 10, 2019, we identified a total of 1139 articles using different databases, Addis Ababa and Bahir Dar University research repositories. Nine hundred twenty-four articles were excluded because of duplication and relevance-related issues. Among the remaining 215 articles, 183 articles were excluded because the outcome was not clearly measured. Fifteen articles were excluded because they were program evaluation and knowledge, attitude, and practice-related reports. Finally, 17 studies fulfilled the eligibility criteria and were included in the final systematic review and meta-analysis (Fig. 1).

Among the total 17 studies included in this current review, five of them were from Addis Ababa City Administration [29, 35‐37, 40]. Five studies were included from the Southern Nations Nationalities and People Region (SNNPR) [30, 32, 38, 39, 41]. In addition, five studies were included from the Amhara region [27, 28, 31, 33, 42]. The remaining two articles were from Oromia [43] and Tigray [34]. Most of the studies used VIA as the measure of assessment for the outcome variable. The total participants in the included studies were 26,112 with samples ranging from 189 [38] to 16,632 [36]. The response rate of all the included studies was greater than 90%, and the quality score of the studies ranged from 62.5 to 100% (Table 2).

To assess publication bias, both the funnel plot and the Egger’s test were conducted in the meta-analysis. The visual examination of the funnel plot exhibited a symmetric distribution of studies (Fig. 2). In addition to the funnel plot, Egger’s regression test was (β = −0.0063, SE=0.08, P=0.88) showing no statistical evidence of publication bias for the included studies.

The lowest prevalence of precancerous cervical lesion was 1.56% which is reported from SNNPR [37] whereas the highest prevalence was 28.4% [29] reported from a study done in Addis Ababa. Overall, the pooled prevalence of precancerous cervical lesion in Ethiopia was 15.16 (95 CI 10.16–19.70) (Fig. 3).

I ² statistic test for heterogeneity indicated that the studies differed significantly (I²= 95.40%, p < 0.001). Hence, we used a DerSimonian and Laird random effects model [48, 49] to estimate the pooled prevalence of precancerous cervical lesion. Studies that showed the largest weight were Plezer et al. [37] that showed 6.35, Kebede et al. [34] with 6.11, Temesgen et al. [42] with 6.10, and Belayneh et al. [28] that showed 6.80 weight, whereas Birara et al. showed a relatively smaller weight which was 5.62 (Fig. 3).

Despite having one of the lowest single study prevalence rates, the highest pooled subgroup analysis of precancerous cervical lesion prevalence was found in SNNPR, 19.65% (95% CI 15.51–23.80), followed by Addis Ababa City Administration which the pooled prevalence was 15.10 (95% CI 4.77–25.44).

An additional subgroup analysis was done using the HIV status of study participants. Accordingly, the highest pooled estimate of precancerous cervical lesion was found in those study participants whose sero-status is unknown, 17.27% (95% CI 5.12–29.41). However, the pooled prevalence rate of precancerous cervical lesion in those study participants whose status was confirmed to be HIV-positive was 15.27% (95% CI 9.06–21.48) (Table 3).

The meta-regression analysis was conducted considering publication years, HIV status, sample size, and study area. However, the results showed that none of these variables was a statistically significant source of heterogeneity.

Trends of the prevalence of precancerous cervical lesion were done to observe the patterns of the disease in Ethiopia using the reported prevalence and year of study. A significant upward trend in the prevalence of precancerous cervical lesion was observed from 1992 to 2019 (B =0.56, P=0.014) (Fig. 4).

In the current systematic review and meta-analysis, we noticed the following limitations. The pooled prevalence which was determined using the eligible 17 studies only represented four different administrative regions in Ethiopia. Hence, it does not represent the whole country as studies in more than half of the regions were not eligible for inclusion in the current review. Utilization of various data collection tool in the included studies might have its own influence in generalizing the finding. In addition, potentially unpublished studies were not considered and latent sources of heterogeneity may also have an impact in the overall finding of the result.