Beyond the Meat of the Matter: A Systematic Review and Meta-Analysis of the Hepatitis E Seroprevalence and Food-Borne Transmission Potential in the Balkans

Figures

• 

  Figure A1. Overall pooled HEV seroprevalence using the Freeman–Tukey double arcsine transformation (PFT + Inverse variance model). Forest plot showing the pooled anti-HEV seroprevalence across 28 studies using the Freeman–Tukey (PFT) transformation with inverse-variance pooling. The pooled prevalence was 7.49% (95% CI 4.76–10.74), with high between-study heterogeneity (I2 = 97.9%, τ2 = 0.0204, p < 0.001). The result was consistent in direction and magnitude with the main PLOGIT–GLMM model (5.68%; Figure 2), confirming the robustness of pooled estimates to transformation choice.

• 

  Figure A2. Overall pooled HEV seroprevalence using logit transformation with inverse-variance pooling (PLOGIT + Inverse model). Forest plot of the pooled anti-HEV seroprevalence under a logit transformation with inverse-variance weighting. The estimated pooled prevalence was 6.44% (95% CI 4.1–9.96), with substantial heterogeneity (I2 = 95.2%, τ2 = 1.44, p < 0.001). These findings are nearly identical to those obtained with the PLOGIT–GLMM approach (Figure 2), demonstrating stability across alternative pooling methods.

• 

  Figure A3. Forest plot of pooled HEV seroprevalence stratified by national pork consumption category (very low/low, moderate, high). Pooled estimates were calculated using a random-effects generalized linear mixed model (GLMM) with logit transformation. Squares represent individual study estimates with 95% confidence intervals (CIs); diamond shapes indicate pooled subgroup estimates. For Anițǎ et al., 2014, only data from the 2012 cohort were included [73]. For Voiculescu et al., 2010, only data from the healthcare professionals cohort were included [74].

• 

  Figure A4. Forest plot of pooled HEV seroprevalence stratified by serological assay group (Abbott ELISA, Dia.Pro ELISA, Euroimmun ELISA, Mikrogen ELISA, Genelabs ELISA, other commercial ELISA, and in-house/unclear assays). Pooled estimates were calculated using a random-effects GLMM with logit transformation. Squares represent individual study estimates with 95% CIs; diamonds indicate pooled subgroup estimates.

• 

  Figure A5. Forest plot of pooled HEV seroprevalence stratified by population type (general population vs. blood donors). Pooled estimates were calculated using a random-effects GLMM with logit transformation. Squares represent individual study estimates with 95% CIs; diamonds indicate pooled subgroup estimates.

• 

  Figure A6. Forest plot of pooled HEV seroprevalence stratified by study period (<2000, 2000–2010, >2010). Pooled estimates were calculated using a random-effects GLMM with logit transformation. Squares represent individual study estimates with 95% CIs; diamonds indicate pooled subgroup estimates.

• 

  Figure A7. Forest plot of pooled HEV seroprevalence stratified by sex (male vs. female). Pooled estimates were calculated using a random-effects GLMM with logit transformation. Squares represent individual study estimates with 95% CIs; diamonds indicate pooled subgroup estimates.

• 

  Figure A8. Forest plot of pooled HEV seroprevalence stratified by country. Pooled estimates were calculated using a random-effects GLMM with logit transformation. Squares represent individual study estimates with 95% CIs; diamonds indicate pooled subgroup estimates.

• 

  Figure A9. Doi plot for assessment of small-study effects in the included studies. The Luis Furuya–Kanamori (LFK) index was −4.27, indicating major asymmetry.

• 

  Figure 1. PRISMA flow diagram of study screening and selection.

• 

  Figure 2. Random-effects forest plot showing anti-HEV seroprevalence in the general population and blood donors across 28 studies conducted in the Balkans. Individual study estimates are presented as proportions (events/total) with corresponding 95% confidence intervals (CIs). Squares represent study-specific prevalence estimates, and horizontal lines indicate 95% CIs. In this framework, study weights are not displayed because pooling is based on binomial likelihood rather than inverse-variance weighting. The final row presents the overall pooled estimate across all included studies, with the pooled prevalence (diamond) estimated using a generalized linear mixed model with logit transformation (PLOGIT–GLMM). The width of the diamond corresponds to the 95% CI of the pooled estimate. Between-study heterogeneity was assessed using Cochran’s Q test and quantified with Tau2 (between-study variance) and I2, which reflects the proportion of total variability attributable to true between-study heterogeneity rather than sampling error.

• 

  Figure 3. The prevalence of anti-HEV total immunoglobulins across the Balkans throughout the decades.