We included TB notification data from 45 countries with a high TB, TB/HIV or MDR-TB burden: 21 countries in the African Region, two in the Region of the Americas, two in the Eastern Mediterranean Region, nine in the European Region, seven in the South-East Asia Region and four in the Western Pacific Region. There were insufficient years of appropriately disaggregated data to include Angola, Mozambique, Papua New Guinea and Uganda. A total of 2,334,656 (6.9% of all notifications disaggregated by age) children (aged < 15 years), 27,448,386 (81.7%) adults (aged 15–64 years), and 3,828,418 (11.4%) elderly (aged ≥ 65 years) people with TB were notified from 2013–2019 across all 45 countries, including 19,961,383 (63.8% of all notifications disaggregated by sex) men (aged ≥ 15 years) and 11,346,972 (36.2%) women (aged ≥ 15 years). A total of 325,964 (6.5%) children, 4,079,324 (81.6%) adults, 595,840 (11.9%) elderly, 2,917,005 (62.4%) men and 1,758,159 (37.6%) women were notified in 2020 (see Table 1).

Country-specific Poisson models (see Additional file 1) suggest that compared to these observed notifications, an estimated 195,449 (95% confidence interval, CI: 189,673–201,562) of an expected 517,168 children (37.8%), 1,126,133 (CI: 1,107,146–1,145,704) of an expected 5,170,592 adults (21.8%) and 235,402 (CI: 228,108–243,202) of an expected 826,563 elderly (28.5%) were missed or delayed in their TB diagnosis in 2020 as a result of the pandemic. This included 511,546 (CI: 499,623–523,869) of an expected 2,250,097 women (22.7%) and 863,916 (CI: 847,591–880,515) of an expected 3,763,363 men (23.0%) (Fig. 1).

Observed and expected tuberculosis notifications in 2020 for 45 high TB, TB/HIV and MDR-TB burden countries for a children aged < 15 years, b adults aged 15–64 years, c elderly aged ≥ 65 years, d men aged ≥ 15 years and e women aged ≥ 15 years. Colours indicate the World Health Organization region for each country and labels indicate the iso3 code. 95% confidence intervals have been omitted as these are not visible at this scale

In 24 of 42 countries (57.1%) with fewer than predicted notifications for either children or adults, there was evidence (strong evidence in 21 countries, evidence in a further three) that missed or delayed diagnoses due to COVID-19 were associated with being a child (i.e. child-to-adult RR > 1). In ten countries (23.8%), there was evidence (strong evidence in eight countries, evidence in a further two) that missed or delayed diagnoses due to COVID-19 were associated with being an adult (i.e. child-to-adult RR < 1). There was no evidence of any association between risk and being either a child or an adult in the remaining eight countries (19.0%) (Fig. 2a).

Risk ratios for disruption to tuberculosis notifications due to the pandemic for 45 high TB, TB/HIV and MDR-TB burden countries by WHO region for a children aged < 15 years compared to adults aged 15–64 years, b elderly aged ≥ 65 years compared to adults aged 15–64 years and c women aged ≥ 15 years compared to men aged ≥ 15 years. Risk ratios > 1 imply that the first population (children, the elderly or women) have had a larger proportion of diagnoses missed or delayed in 2020 as a result of the pandemic. Risk ratios < 1 imply that the second population (adults or men) have had a larger proportion of diagnoses missed or delayed in 2020 as a result of the pandemic. Countries where there were more notifications in both comparator and reference group were excluded from the meta-analysis. Colours indicate strength of evidence; no evidence for a risk ratio different to 1 (grey), strong evidence for a risk ratio > 1 (dark blue), evidence for a risk ratio > 1 (light blue), weak evidence for a risk ratio > 1 (green), strong evidence for a risk ratio < 1 (purple), evidence for a risk ratio < 1 (dark pink) and weak evidence for a risk ratio < 1 (light pink)

In 29 of 41 countries (70.1%) with fewer than predicted notifications for either the elderly or adults, there was evidence (strong evidence in 24 countries, evidence in a further five) that missed or delayed diagnoses due to COVID-19 were associated with being elderly (i.e. elderly-to-adult RR > 1). In five countries (12.2%), there was evidence (strong evidence in four countries, evidence in a further one) that missed or delayed diagnoses due to COVID-19 were associated with being an adult (i.e. elderly-to-adult RR < 1). There was no evidence of any association between risk and being either elderly or an adult in the remaining seven countries (17.1%) (Fig. 2b).

In nine of 40 countries (22.5%) with fewer than predicted notifications for either men or women, there was evidence (strong evidence in five countries, evidence in three and weak evidence in a further one) that missed or delayed diagnoses due to COVID-19 were associated with being a woman (i.e. woman-to-man RR > 1). In nine countries (22.5%), there was evidence (strong evidence in seven countries, weak evidence in a further two) that missed or delayed diagnoses due to COVID-19 were associated with being a man (i.e. woman-to-man RR < 1). There was no evidence of any association between risk and being either a woman or a man in the remaining 22 countries (55.0%) (Fig. 2c).

Evaluating the strength of evidence as defined in the methods above, there was strong evidence in the WHO Eastern Mediterranean (RR = 1.77 [95% CI 1.26–2.48, I2 = 84.9%]) and evidence in the European (RR = 1.32 [95% CI 1.04–1.68, I2 = 85.7%]) regions that notifications for children had been disproportionately affected compared to adults (see Fig. 2a). However, globally (RR = 1.09 [95% CI 0.41–2.91, I2 = 99.9%]) and in remaining WHO regions, there was no evidence that notifications for either children or adults have been disproportionately affected relative to one another.

There was strong evidence the WHO European (RR = 1.33 [95% CI 1.13–1.57, I2 = 84.8%]) and Western Pacific (RR = 1.18 [95% CI 1.06–1.32, I2 = 96.7%]) regions that notifications for the elderly have been disproportionately affected compared to adults (Fig. 2b). However, globally (RR = 1.40 [95% CI 0.62–3.16, I2 = 100.0%]) and in remaining WHO regions, there was no evidence that notifications for either the elderly or adults have been disproportionately affected relative to one another.

There was strong evidence that notifications for men have been disproportionately affected compared to women in the WHO Region of the Americas (RR = 0.78 [95% CI 0.70–0.87, I2 = 70.1%). Globally, and in the remaining WHO regions, there was no evidence that notifications for either men or women have been disproportionately affected (RR = 0.59 [95% CI 0.25–1.42, I2 = 99.9%) (Fig. 2c).

Further region- and country-specific results can be found in Additional file 1.

Although there are some differences in regional RR when considering high TB, high TB/HIV and high MDR-TB burden countries separately, our conclusions are broadly qualitatively similar, with no evidence globally for any difference in risk by age or sex.

Whilst overall and in most age and sex groups there was a reduction in TB cases notified in 2020 compared to expected numbers based on 2013–2019 trends, in 18 countries, there were more notifications than expected in at least one age or sex group; Cameroon, Central African Republic, China, Congo, Democratic People’s Republic of Korea, Eswatini, Ethiopia, Guinea-Bissau, Kazakhstan, Malawi, Mongolia, Nepal, Nigeria, Somalia, South Africa, United Republic of Tanzania, Viet Nam and Zambia. Removing these countries from our analysis, we found that there was strong evidence globally (RR = 1.57 [95% CI 1.22–2.03, I2 = 99.6%]) that notifications for children have been disproportionately affected compared to adults. There was also strong evidence globally (RR = 1.36 [95% CI 1.25–1.48, I2 = 98.7%]) that notifications for the elderly have been disproportionately affected compared to adults. However, there remained no evidence globally (RR = 1.02 [95% CI 0.96–1.09, I2 = 97.5%]) that notifications for either men or women have been disproportionately affected.

We considered there to be a poor model fit or limited data for children compared to adults in 26 countries, for the elderly compared to adults in 24 countries and for women compared to men in 22 countries. Removing these countries from our analysis, we found that there was no evidence globally that notifications for either children or adults (RR = 0.65 [95% CI 0.32–2.03, I2 = 99.9%]) were disproportionately affected. However, there was strong evidence globally (RR = 1.24 [95% CI 1.08–1.42, I2 = 99.0%]) that notifications for the elderly were disproportionately affected compared to adults and weak evidence (RR = 0.39 [95% CI 0.13–1.15, I2 = 100%]) that men were disproportionately affected compared to women.

See additional file 1 for further details.