3.1. The pooled prevalence estimate of cervical cancer among women in Ethiopia
Seven studies conducted at health facility level with a total simple size of 3951 were included to compute the pooled prevalence estimate of cervical cancer among women in Ethiopia. The highest prevalence was observed in a study done in gynecology referral Clinics in 2018 (39.5%) (45) whereas the lowest point estimate was observed at a study conducted in North Ethiopia in 2017(6.7%)(25). A Statistically significant heterogeneity between studies was also detected (I2=94.35% (p<0.0001) (Table 4).
This study based meta-analysis forest plot with random effects model discovered that more weight was given for a study conducted in Yirgalem General Hospital in 2017(19) (i.e. the bigger the box, the more participants in the study, the smaller 95% confidence intervals and the higher the precision to predict the pooled estimate).As a result, studies given the higher weights were had the more influence on the pooled effect estimate. Therefore; the combined point estimate of cervical cancer was found to be 15.7%, 95%CI (10.78%, 21.31%) (Figure 3)
In figure 4, publication bias had been indicated by funnel plot symmetry or asymmetry. This visual inspection of the funnel plot confirmed that the plot was somewhat symmetrical. This indicates that there was no possible publication bias up on discovering the prevalence of cervical cancer among women in Ethiopia.
3.2. Factors associated with cervical cancer among women in Ethiopia
(Table 5 &: Table 6)
At a meta-analysis level, each study was not given equal weighting in the comparison. Those studies with wider confidence intervals were given a lower weighting than those studies having narrow confidence intervals or those studies with a larger square were showing a higher weighting and vice versa. As a rule of thumb studies with a greater number of participants have a narrower confidence interval and a smaller horizontal line. That means the bigger the study, the smaller the horizontal line and the bigger the square of the point estimate or the smaller the study, the wider the horizontal line and smaller the square of the point estimate
A study conducted in gynecology clinics of Addis Ababa in 2019 (22) had more effect to pull the overall effect estimate to the left (favours to the bad event /cervical cancer). The 95% CI of this study did not overlap 1(the no effect). So, there was statistical significance association with cervical cancer at the study level like a study conducted in Addis Ababa (46) and in Jimma (47). Both had the lowest effect on the overall effect estimate (OR=1.34, 95% CI (0.56, 3.12) respectively. More weight (19.2%) was given for studies conducted in Addis Ababa (46) and in the Gurage zone (48). This tells us the weight was given based on the size of the sample they used for their studies.
Therefore, the overall effect estimates of women’s age (OR=1.43, 95 %CI (0.65, 3.12) was slightly associated with cervical cancer but it is not statistically supported (Figure 5).
From this sub group meta-analysis, the study done at Yirgalem General Hospital in 2017 had more effect size and weight on the combined effect estimate (19). Its 95% confidence interval did not overlap 1 (the no effect). The 95% confidence intervals of all other the studies overlap 1. But the overall effect estimate 95% CI did not the no effect line That means marital status had statistically significant association with cervical cancer (Figure 6).
Except one study(19), all other studies did not show a statistically significant association with cervical cancer at a study level (22, 46, 49). As a result, the overall effect estimate touches the line of no effect (Figure 7).
The study conducted at Yirgalem General Hospital in 2017 had more effect, high precision and statistically significant (19). But a study conducted in Addis Ababa in 2018 had low precision and statistically significant at a study level (22). Due to the higher weight (35.8%) and more effect was given for a study done at Yirgalem General Hospital, it pulls the overall effect estimate towards to the better event (favours to Negative)(19). Moreover, studies were had higher heterogeneity (I2=96%) at (p<0001). The 95%CI of the overall effect estimate did overlap the no effect line. That means there was no statistical significance (Figure 8).
In this analysis more effect was observed at a study conducted in Jimma University Specialized Hospital in 2015(47) where as more weight (52.7%) was given for a study conducted at Yirgalem General Hospital in,2017 due to its larger simple size (19). Higher heterogeneity was also observed (I2=92%) between these studies. However, the combined effect estimate, OR=1.44, 95%CI (0.37, 5.58) did not show statistical significance (Figure 9).
Income was another important variable to predict cervical cancer distribution among women in Ethiopia. The 95% confidence intervals of all the studies did not overlap 1 and the 95% confidence intervals of the overall effect estimate did not overlap 1. This indicates that there was statistical significance at the study level (Figure 10).
As the forest plot indicates that the combined point estimate did not touch the null value. Therefore, there was statistically significant effect difference on cervical cancer between women of modern contraceptive users and non-users (Figure11).
All studies except a study done at Family Guidance Association of Ethiopia in Jimma in 2015(50), the 95% confidence intervals did not touch the line of no effect(46, 47, 49). As a result, their effect pulled the overall effect estimate towards the left (bad event/cervical cancer episode). Thus, there was a statistically significant effect difference on cervical cancer prevalence between women having >1 life time sexual partners and women had 1 life time sexual partner (OR=0.40, 95%CI (021, 0.75) (Figure 12).
Like other variables history of STI predicted the prevalence of cervical cancer among women in Ethiopia. The overall effect estimate, OR=1.99, 95%CI (1.02, 3.87) showed that there was statistical significance effect difference between women having a history of STI and women did not have a history of STI (Figure 13).
This forest plot shows that the highest effect but the least weight (7.4%) was given for a study done in North Ethiopia in 2016 even though it did not show statistical significance(25). The 95% confidence intervals of studies conducted at Debre Markos referral hospital, Addis Ababa and Yirgalem General Hospital did not overlap 1(19, 46, 51). But there was statistical significance effect difference at the meta-analysis level (Figure 14).
All studies(22, 46, 47, 50) except studies done in the Gurage zone(48) and Adama town(49), the 95% CIs cross the no effect line (1). Similarly, the combined effect estimate (the diamond) also overlaps on the no effect line (Figure 15).
This forest plot shows that the 95% confidence intervals of both studies(46, 47) overlap 1 as a result the 95% confidence intervals of the overall effect estimate also overlaps 1 (Figure 16).