Incidence and determinants of extrapulmonary tuberculosis in Egypt: a retrospective cohort study

Categories: Disease & Virus

April 20, 2025

Demographics and characteristics of TB patients

A total of 8,223 TB patient records were reviewed, with 978 records excluded from the analysis due to incomplete records or diagnoses made outside of MoHP hospitals and clinics. Among the analyzed patients 7245, 33.5% were aged > 15–30 years, 66.8% were male, and 74.3% were residents in urban areas. The vast majority (98.7%) had no documented comorbidities, and 42.5% presented with EPTB. Among the patients tested for HIV or HCV coinfection, 2.0% and 0.74% tested positive, respectively. Furthermore, when categorized by HDI, the majority (84.8%) of individuals resided in high-HDI areas, followed by 14.3% in medium-HDI regions, and only about 0.9% in very high-HDI locations, as shown in Fig. 1 and Table 1.

Table 1 Characteristics of TB patients and children living in 25 Egyptian governorates.

Figure 2 illustrates the percentage of EPTB across Egyptian governorates, highlighting a high prevalence in Suez and Luxor, where over 60% of patients had EPTB. This is followed by Minia and Beni Suef, with approximately 50% of patients affected. In contrast, the lowest percentages were observed in North Sinai and Alexandria, where less than 20% of patients had EPTB.

Figure 3a illustrates the distribution of EPTB cases across various anatomical sites, with the lymph nodes (27.1%) and pleural cavity/effusion (24.6%) being the most frequently affected sites. Other commonly impacted sites included the vertebrae (9.4%), bones and joints (7.1%), and the gastrointestinal and peritoneal cavity (6.9%). Less commonly affected sites included breast, genitourinary system, skin, brain or meninges, neck and salivary glands, and eye, while cases involving the ear, nose, throat, and pericardium/heart were rare.

Figure 3b further categorizes these EPTB cases based on HIV and HCV positivity, illustrating the distribution of infections across different anatomical sites. HCV-positive cases were more frequently observed in the pleural cavity and lymph nodes, with up to 6 cases in some locations, whereas HIV-positive cases were distributed across multiple sites, including the pericardium/heart (2 cases), bones & joints (2 cases), and skin & underarm (2 cases). Notably, certain sites, such as the brain/meninges (3 cases), genitourinary system (2 cases), and eye (2 cases), showed a higher prevalence of HCV-positive cases compared to HIV-positive cases.

Clinical and epidemiological associations with EPTB and PTB

Table 2 presents the distribution of EPTB and PTB cases across various demographic and clinical variables. HIV status showed a notable relationship, with HIV-positive individuals having a higher percentage of PTB (78.6%) compared to EPTB (21.4%), indicating a significant association (p < 0.001). Gender analysis revealed that females had a significantly higher percentage of EPTB (62.2%) than males (37.8%) with significant association (p < 0.001). Age was a significant factor, with younger individuals showing higher EPTB percentage, decreasing with age (p < 0.001). Urban residents had a higher percentage of PTB (59.7%) than in EPTB (40.3%), with a significant association (p < 0.001). Comorbidities also affected TB clinical presentation, with a higher percentage of PTB observed in individuals with one or more comorbidities (69.8%) compared to those with EPTB (29%, p = 0.017). Additionally, the HDI was significantly associated with TB type, with lower rates of EPTB in higher HDI categories (p < 0.001). In contrast, HCV status did not show a significant difference in TB percentage between HCV-positive and HCV-negative individuals (p = 0.601).

Table 2 TB diagnosis by patient characteristics.

Multilevel logistic model results

In the null model with only an intercept, a likelihood ratio test comparing models with and without a random component revealed a significant difference (χ² = 369.66, df = 1, p = 0.0001). The ICC was 0.010, and the design effect was 12.9, indicating the necessity of a multilevel model with a random intercept, reflecting variability among governorates relative to the total variance. Table 3 shows the odds ratio (OR) for EPTB compared to PTB as 0.93, with a fixed intercept of −0.07, corresponding to a 48% probability of EPTB. The within-group variance in the logit of the odds is 3.29 (σ²), and the random intercept variance is 0.36 (σ00). The data from 25 Egyptian Governorates, totaling 4374 observations, demonstrated considerable variability in random effects, with a governorate-level variance (σ00) of 0.44 and an ICC of 0.13, indicating that 13% of the variance in TB type is explained by differences between governorates.

Adding random slopes for significant variables did not produce substantial results. HIV-positive individuals were less likely to have EPTB (OR = 0.46, 95% CI: 0.30–0.71, p < 0.001). Males had a 69% lower likelihood of EPTB compared with females (OR = 0.31, 95% CI: 0.27–0.35, p < 0.001). Younger age significantly increased the odds of EPTB compared to the reference category (30–44 years), particularly in children under five years (OR = 4.75, 95% CI: 2.29–9.84, p < 0.001) and those aged 5–15 years (OR = 2.99, 95% CI: 2.07–4.33, p < 0.001). Comorbidities were associated with a 41% reduction in the odds of EPTB (OR = 0.59, 95% CI: 0.35–0.98, p < 0.05).

Contextual variables, including place of residence (OR = 1.05, 95% CI: 0.87–1.27, p > 0.05) and HDI categories; high HDI (OR = 1.69, 95% CI: 0.86–3.3, p > 0.05), and very high HDI (OR = 1.29, 95% CI:0.64–2.64, p > 0.05) did not significantly increase the odds of having EPTB. The marginal R² indicated that fixed effects explained 11% of the variability, while the conditional R², incorporating random effects, explained 22%.

Figure 4 highlights the predicted probabilities of EPTB across various governorates. The highest probabilities were observed in Suez, Sohag, Minia, Luxor, and Assiut, with predicted probabilities of 0.69, 0.69, 0.68, 0.67, and 0.67, respectively. Other governorates, such as Red Sea, Qena, and Gharbia, also showed relatively high probabilities (0.63, 0.57, and 0.53). In contrast, lower probabilities were observed in governorates like Alexandria (0.25), Buhira (0.31), and North Sinai (0.14). This variability suggests that geographical factors may influence the likelihood of EPTB.

Table 3 The multilevel model with random intercept (EPTB patients).