Measuring the burden

1
Measuring the burden of TB directly Surveys of
TB infection and disease - Brian Williams
(STB/TME) - Geneva Workshop June 2005
2

• We want to reduce
• Morbidity and mortality
• New infections
• Recurrence of old infections
• Prevalent cases

3
Routine programme data

• Notifications
• Cure rates
• Drug resistance
• Mortality
• Process indicators

4
But...

• Biases
• Incompleteness
• Recording
• Reporting
• Analysing
• AIDS

5

• Four things we can measure
• Prevalence of infection (tuberculin surveys)
• Incidence of infection (ARTI)
• Incidence of disease
• Prevalence of disease
• Prevalence of HIV in TB patients

6
Measuring the prevalence of infection Tuberculin
surveys
7
Sampling Districts Probability proportional to
size Schools in districts Random Children in
schools Total sample within schools Sample
size Small to minimize costs and logistical
problems Large to get the most accurate
estimate How many districts? How many schools in
districts? How many children in schools?
8
Design effect Multiply your sample size by D. D
1 (m - 1)? where m is the number of people
in each cluster and the intra-cluster correlation
coefficient, ?, is ? VB/VW with VB the
between-cluster variance and VW the total
variance. No difference between clusters VB
0, ? 0, D 1. Members of a cluster are
identical VB VW, ? 1, D m
9
(No Transcript)
10
Calculating ARTI (R) from prevalence (P) and age
(A) R 1 (1 P)1/A Probability of being
infected per year R Probability of not being
infected per year 1 R Probability of not
being infected after A years (1 R)A 1 P
(1 R)A 1 R (1 P)1/A
11
Assumptions The annual risk of infection is
constant over time and independent of
age. Surveys in school children, aged about 5 to
10 years estimate of the ARTI is averaged over
about five to ten years.
12
South Korea
Halving in ln(2)/0.07 ? 10 years
10 to 20 years
5 to 10 years
13
Measuring the incidence of disease
14
Calculating incidence from the ARTI Styblo
showed that for every 1 increase in the annual
risk of infection, the incidence of SS disease
increases by (50?10)/100,000/yr. If we multiply
the ARTI by 50 this gives us an estimate of the
incidence of SS TB.
15
Estimating sample sizes quickly If you count N
positives then ? ??N 95 confidence limits ?
?2?N Fractional error ? ?2?N/N ?2/?N Incidence
(South Korea, 1990) ? 40 per 100,000 95
confidence limits ? ? 2?N ? 13 (? ?32) Sample
of 800,000 gives 320 positive ? ?11
16
South Korea Incidence Survey 1998
Baseline 920,000 had chest radiography 840,000
did not have suspicious CXR 1990 Follow
up 790,000 available for CXR 2,500 with
suspicious X-rays 2000 had sputa taken 200
smear positive 300 culture positive (smear
negative)
17
South Korea
ARTI?50
Civil servants
Rate of decline 7.9/yr
18
Measuring the prevalence of disease Same
sampling problems as for incidence.
19
South Korea
Rate of decline 8.0/yr
20
Incidence and prevalence It is easier to reduce
prevalence than incidence. Duration of disease ?
Prevalence/Incidence
21
Incidence and prevalence in South Korea
Prevalence/100k Incidence/100k/yr
Duration of disease/years
Age/years
22
DDR
23
(No Transcript)
24
Smear positive disease in South African gold
miners Incidence (/yr)
Prevalence () Dis.Duration (yr) HIV- 0.48
0.55 1.15 HIV 2.87 0.44 0.15
Ratio 6.01 0.80 0.13
Corbett et al. 2003
25
Changes in the annual risk of TB infection in
Kenya (percent per annum)
Early HIV Late
HIV 1986-1990 3.7 5.9 1994-1996 11.3 6.9
Ratio 3.2 1.2
Early HIV Kisii, South Nyanza, Kakamega, Siaya,
Nairobi Late HIV Nakuru, Muranga, Kitui, Elgeyo
Marakwet, Meru
Odhiambo et al. Am. J. Pub. Health (1999)
26
Changes in the annual risk of TB infection in
Tanzania (percent per annum) 1995
1.10
2000 0.67
Ratio 0.61
Egwaga et al. TSRU meeting, Bagamoyo
India
27
The Impact of HIV
28
The incidence of SS tuberculosis plotted against
the prevalence of HIV for African countries in
2001.
29
Sub-Saharan Africa
30
Nairobi
31
Nairobi
32
Nairobi
33
Williams, B. G. Dye, C. Antiretroviral Drugs
for Tuberculosis Control in the Era of HIV/AIDS
Science (2003) 301 1535-1537.
34
Reduction in the life-time incidence of TB among
HIV positive people
Williams, B. G. Dye, C. Antiretroviral Drugs
for Tuberculosis Control in the Era of HIV/AIDS
Science (2003) 301 1535-1537.
35
We think that DOTS should continue to work for
HIV-negative people and HIV-positive people need
ART. But how do we stop transmission?