2-PG TB estimates

Estimates of TB burden

Philippe Glaziou Harare, 2010

Can we measure TB incidence?

Sources of data on incidence

• Incidence surveys

• PPD surveys

• Notifications

• Capture-recapture

• Indirect estimation

http://www.flickr.com/photos/giacomofrigerio

duration

prevalenceincidence =

www.flickr.com/photos/bogenfreund/

ratefatalitycase

mortalityincidence =

ratedetectioncase

onsnotificatiincidence =

http://www.flickr.com/photos/teenytinyturkey/313457260/

Onion model: size of the non-

notified TB population

Inventory study

Survey

Survey

DHS

Source

Capture-

recapture

Diagnosed but not

reported

Seek care but not

diagnosed

Have access but do not

seek care

Do not have access to

health care

Sources of data

19Tuberculin surveys

3Inventory / capture re-capture

6Prevalence surveys of disease

89Mortality (vital registration)

212Case notifications

Number of

countries

Mortality

• Measurements from VR in 88 countries– Coverage > 80%

– Ill-defined causes < 20%

• Prior distributions of Case Fatality Rates (CFR)– Commissioned literature review (Masja

Straetemans)

– WHO database

• Posterior distributions of CFRs from Bayesian models

Interpreting mortality time trends

in South Africa

0%

4%

8%

12%

1988 1992 1996 2000 2004

TB

& H

IV

0%

5%

10%

15%

20%

25%

Ill-

de

fin

ed

VR: HIV/AIDS

VR: ill-defined

VR: TB

Vital registration:

More reported deaths (VR) per

notified TB case in East Europe

Trends in TB incidence and

mortality in Latvia

From Vital

Registrations

Trends in TB incidence and

mortality in Brazil

CFR = probability of death at 4 years of follow-up among linked records

probability of linkage of TB death records

From Vital

Registrations

How about prevalence?

• Measurable through cross-sectional

surveys, but …

• Indirect estimates very uncertain

P = I * d

Kenya profile

TB incidence all forms

(including HIV+)

TB incidence all forms

(including HIV+)

TB notifications (black)

TB prevalence all

forms (including HIV+)

Tb prevalence per 100,000 population

Disaggregation of estimates

• Smear status: estimates discontinued

• Age group and sex

– Based on distribution of TB deaths by age

and sex in countries with VR

• HIV-status

• MDR-TB

HIV in TB indirectly estimated in 13

out of 46 countries in AFR

logit

(HIV

in T

B)

logit (HIV in the general population)

Insufficient coverage of drug

resistance surveillance in Africa

Percent MDR in

previously untreated

Insufficient capacity

to diagnose MDR-TB

Insufficient lab capacity

to diagnose TB

In conclusion

1. Incomplete data leading to uncertainty

2. Use of vital registration records

3. Incidence trajectories based on notifications and time changes in case finding effort

4. Need to strengthen surveillance, including VR, drug resistance

5. Targets and planning should be based on measured indicators (as opposed to indicators estimated with uncertainty)

Acknowledgements

• Katherine Floyd

• Babis Sismanidis

• Ana Bierrenbach

• the TB estimates sub-group of the Task Force on TB Impact Measurement