Targets for global TB controlMILLENNIUM DEVELOPMENT GOALS"to have halted and begun to reverse

incidence.."

Implementation (DOTS) Target yearCase detection 70% 2005Treatment success 85% 2004/5

ImpactPrevalence 50% of ≈ 300/100K 2015Deaths 50% of ≈ 30/100K (<1m) 2015Incidence <1 per million 2050

1.Measuring and estimating

TB incidence

Direct measures of TB burdenincidence and prevalence

Korean civil servantsTubercle and Lung Disease 76, 534 (1995)

• Prevalence PTB 1990 241/100K • Incidence PTB 1989-90 84/100K/yr

• Estimated duration = 241/84 = 2.9 years (bigger ratio for older age groups)

02468

101214161820

1980 1985 1990 1995 2000 2005

Rep

orte

d TB

cas

es/1

00,0

00/y

r USANetherlandsUKNorway

Where case notifcations = true incidence

Four indirect measures of TB incidence

durationprevalenceTB incidence 2. =

detected proportiononsnotificatiTB incidence 1. =

ratio Styblo infection incidenceTB incidence 3. ×=

rate)fatality (case dying cases proportion

deaths TB incidence 4. =

Incidence, prevalence, deaths derived by rearranging 4 equations

4 steps to check accuracy and completeness of surveillance data

1. Inventory of, and cross-check, data from all possible sources, removing duplications

2. Capture-recapture techniques to estimate case detection from lists of patients that have been "captured" in different ways

3a. Consistency of case reports: spatial and temporal variation, to check for inconsistencies

3b. Consistency of case reports: norms of TB epidemiology and natural history

Completeness of case registrations

TB cases notified by compulsory (CSS) and Varese surveillance systems (VSS)

VSS CSSTotal cases notified 143 89Duplicates 8 13Not TB 14 0True cases 121 76

(-37%)Source: Migliori ERJ 8, 1252 (1995)

Capture-recapture methodPetersen 1896, Lincoln 1930 (ducks)

Take 2 independent samples from an unknown population of N (TB patients):

Method 1 e.g. lab registerPresent Absent Total

Method 2 Present 20 5 25e.g. hospital register Absent 30 c b

Total 50 a N

N = 50×25/20 = 62.5a = 62.5 - 50 = 12.5 CDR lab = 50/62.5 = 0.8b = 62.5 - 25 = 37.5 CDR hosp = 25/62.5 = 0.4 c = 62.5 - 55 = 7.5

Indonesia: why does the fraction of TB cases among suspects vary between provinces?

0.00

0.10

0.20

0.30

0.40

0.50

MALUKU U

TARABANTENMALU

KUJA

BARKALTENG

JAWA TENGAH

PAPUARIA

U

DKI JAKARTAKALSEL

LAMPUNGKALBARSULS

ELD.I.

ACEH N

T T BALI

JAWA TIM

URSUMUTSULT

RASUMSEL

NTB

KALTIMSULU

TSUMBARSULT

ENGBABELD. I

. Y.

GORONTALOJA

MBI

BENGKULUTB

cas

es/s

uspe

cts

Morocco: consistent 40-60% smear+ across 16 regions

0.0

0.2

0.4

0.6

0.8

Oued Ed

Laâyo

une Guelm

imSou

ssGharb

Chaouia

Marrak

ech

Oriental

Casab

lanca

Rabat

Douka

laTad

laMek

nès FèsTaz

aTan

gerPr

opor

tion

smea

r-pos

itive

1b. Assessing trends in TB

incidence, and the impact of control

Cambodia: higher case rates among older people imply long-term epidemic decline

0

200

400

600

800

1000

1200

0–14 15–24 25–34 35–44 45–54 55–64 65+

Age group (yrs)

Rep

orte

d TB

cas

es/1

00k

in 2

006 Male

Female

0

5

10

15

20

25

30

-10 -8 -6 -4 -2 0 2 4 6 8 10 >10

Annual change incidence (up to % marked)

Num

ber c

ount

ries

High incomeSE Asia & W PacificLatin AmericaE MediterraneanC&E EuropeSub-Saharan Africa

TB incidence is falling slowly in 95 of 135 countries, 1996-2005

Morocco:PTB incidence projected to 2015

0

10

20

30

40

50

60

70

1980 1985 1990 1995 2000 2005 2010 2015

Inci

denc

e ra

te/1

00K

on 1994 age-structure

on aging population

31

32

33

34

35

1980 1985 1990 1995 2000

Mea

n ag

e of

adu

lt pa

tient

s (y

r)TB patients in Morocco

Women: 0.15 ± 0.02

Men: 0.14 ± 0.01

Year

0

1

2

3

4

5

6

Men Women Men Women

All forms Smear+

Dec

line

repo

rted

cas

es (%

/per

son/

yr)

UrbanRural

Morocco: TB falling slowly in women, very slowly in men

Trends in case notificationrates in Viet Nam, 1997-2004

-8

-4

0

4

8

15-24 25-34 35-44 45-54 55-64 65+ total

Ann

ual p

erce

ntag

e ch

ange

Men Women Total

020406080

100120140

1990 1995 2000 2005Rep

orte

d TB

cas

es/1

00K

/yr

Average age of men 15-54 yrs with TB is falling in some countries

37

38

39

40

41

42

1996 1998 2000 2002 2004

Ave

rage

age

(yr)

Viet Nam

Sri Lanka

Myanmar

China

TB trends: New Caledonia

All TB 9.4%/yr

Smear+ 8.2%/yr

0

10

20

30

40

50

60

70

80

90

1990 1995 2000 2005

Not

ifica

tion

rate

(per

100

000

) Notifications(new &relapse)

Notificationsnew smear-positive

0

50

100

150

200

250

1980 1985 1990 1995 2000 20050

500

1000

1500

2000

2500

Pul

mon

ary

TB c

ases

/100

k/ye

ar

Slid

es e

xam

ined

(k/y

r)

Suarez, P. G. et al. The dynamics of tuberculosis in response to 10 years of intensive control effort in Peru. J Inf. Dis 184, 473-8 (2001)

DOTS

Increasingdiagnostic effort

Decliningnotification rate

Impact of DOTS in Peru

2. Measuring and estimating

TB prevalence

When to do a prevalence survey

• High burden (e.g. among 22 HBC)• Uncertain burden, in part because surveillance

is weak • Potential for collecting other data e.g. about

where patients are diagnosed and treated • Logistically feasible - terrain, population

density, staff security, mobile X-ray, culture etc

• Potential source of funds: $100,000 - $1m• Potential for doing 2 or more surveys (to

measure change)• Participatory population

Country Design Number examined

Number active pulmonary cases (prevalence ±

95%CL, per 100,00)

Number culture-positive cases

(prevalence ±95%CL, per

100,000)

Number smear-

positive cases (prevalence ± 95%CL, per 100,000)

Republic of Korea 1995 28

Stratified, cluster randomized; Age ≥ 5 years;

X-ray (miniature) screen

64 713 668 (1032 ± 80)[1] 142 (219 ± 37)[2]

(culture+ and/or smear+)

60 (93 ± 24)

Philippines 1997 31

Stratified, cluster randomized; Age ≥ 10 years;

X-ray screen

21 960 537 (4200 ± 330)[3] 124 (810 ± 88) 47 (310 ± 99)

China 2000 29 Stratified, cluster randomized;Age ≥ 3 months;

Tuberculin/symptom/fluoroscopy screen

365 097 1340 (367 ± 28) 584 (160 ± 16) 447 (122 ±14)

Cambodia 2002 50

Stratified, cluster randomized; Age ≥ 10 years;

X-ray/symptom screen

22 160 580 (1916 ± 300) 271 (899 ± 165)(culture+ and/or

smear+)

81 (269 ± 66)

Indonesia 2004 30

Stratified, cluster randomized; Age ≥ 15 years; Symptom screen

50 154 N/A 48 (186 ± 49)[4] 80 (104 ± 38)

Eritrea 2004 120

Stratified, cluster randomized; Age ≥ 15 years;

No screen

18 152 N/A N/A 15 (50 ± 30)

The 6 national TB prevalence surveys carried out since 1995

DOTS reduces prevalence of TB by 37% in less than a decade in China

0

50

100

150

200

250

1990 2000

Prev

alen

ce c

ultu

re+

TB/1

00,0

00

DOTSOther

Progress towards MDGs in Indonesia prevalence rate fell 4%/yr 1980-2004?

61%

45%

74%

53%

0

100

200

300

400

500

600

Sumatera Java-Bali KTI (East) National

Smea

r+ p

reva

lenc

e/10

0K

1980 regional surveys

2004 national survey

% fall 1990-2004

"Model DOTS Project" reduces TB prevalence in south India

source: TRC Chennai

0

1000

2000

68-70

71-73

73-75

76-78

79-81

81-83

84-86

99-01

01-03

Year

Pre

vale

nce/

100K

Male C+Male S+Female C+Female S+

fall ~10%/yr in MDP

Incidence cannot reliably be estimated from prevalence e.g. Cambodia?

prevalence ss+ (survey) 269/100K in 2002

weighted duration DOTS 0.65 @1.0 yearnon DOTS 0.10 @1.5 yearsuntreated 0.25 @2.0 years

= 1.3 years

Therefore: incidence ss+ = 269/1.3 = 207/100K/yearNB: usually wide range on estimates

duration weighted incidence ss prevalence ×=+

3. Measuring and estimating

TB mortality

Measuring and estimating TB deaths

Three approaches1. Incidence × case fatality (method 4)2. Verbal autopsy (in sample vital

registration)3. Vital (death) registration

rate)fatality (case dying cases proportion

deaths TB incidence :4 Method =

Comparing unknownsCause of death from vital statistics (VSD) and verbal autopsy (VA) of 48 000 adult (> 25) deaths in Chennai, India: 1995-97

Cause of death in VSD

Cause of death based on VA

Other Respiratory 1088 ( 4) 596 ( 3) 1597 ( 6) 855 ( 4)

Cause ofdeath

M (%) F(%) M (%) F(%)

Vascular disease 8319 (30) 5168 (25) 11056 (41) 7435 (37)

1999 (10)

618 ( 3)

5889 (29)

2804 (14)

Nil

20175

Tuberculosis (TB) 1399 ( 5) 372 ( 2) 2231 ( 8) 575 ( 3)

Neoplasm 1163 ( 4) 1002 ( 5) 2344 ( 9)

Infectious (ex Resp/TB) 584 ( 2) 303 ( 2) 1034 ( 4)

Unspec med. 12291 (44) 11511 (56) 4367 (16)

Other spec med. 1899 ( 7) 1045 ( 5) 4414 (16)

Cause NA 983 ( 4) 634 ( 3) Nil

Total deaths - med 27726 20631 27043

Source: Jha, Gajalakshmi et al

Regional trends in TB death registrationsfew data from Asia & Africa, long reporting delays

0

5

10

15

Geo

met

ric m

ean

TB m

orta

lity

(per

100

K)

20

1985 1990 1995 2000 20050

10

20

30

40

TB m

orta

lity

Phili

ppin

es

50

(per

100

K)

ex Soviet UnionCentral EuropeIndustrializedLatin AmericaRep. KoreaPhilippines

4. Measuring and estimating risk of

TB infection

0

5

10

15

20

25

5 10 15 20 25 30Tuberculin reaction (mm)

Num

ber o

f chi

ldre

n

Curve fitting

Chadha, V. K. et al. Annual risk of tuberculous infection in the northern zone of India. Bull World Health Organ 81, 573-80 (2003).

Cut-off

Mirror

0 to 9 years19651995

0 10 20 30 0 10 20 30 Size of induration (mm) Size of induration (mm)

Freq

uenc

y (%

) 10

5

0 Freq

uenc

y (%

) 10

5

0

0 to 20 years19651995

a b

c d

BCG

No BCG

BCG

0 10 20 30 0 10 20 30Size of induration (mm) Size of induration (mm)

Freq

uenc

y (%

)

Freq

uenc

y (%

)4

2

0

15

10

5

0

Tanzania 1983 to 1988 Egypt 1995 to 1997

No BCG

Korea

Styblo's 1:50 rule and endemic (untreated) TB

Prevalence infectious TB

~ sputum smear+, 2yr

Infected with M tuberculosis

Prevalence non-infectious TB ~ sputum smear-

0.5 0.5

10 × p

MTB infection

0.1 0.1

TB incidence

Styblo (incidence × 105)/(risk infection × 102) = 1000/(10 × 2) = 50

Endemic TB 1 smear+ cases gives 10 × 2 × 0.1 × 0.5 = 1 smear+ case

0

5

10

15

20

25

30

ALDEWANIA

AL-NAJA

FKARBALA

AL-MUTH

ANNAWASIT

AL-BASRAHMESAN

BAGHDADBABIL

KURKOKTHE-Q

AR

SALAH ADDEN

NINAWA

DIYALAAL-A

NBAR

Smea

r+ n

otifi

catio

ns/1

00K

20052004

IRAQ Notification rate varies 5-fold among governerates

0.51% infection1.43% infection

Tuberculin skin test responses in houehold contacts of active TB cases

979 children, median age 7yr, Istanbul

01020304050607080

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31

ELISpot -veELISpot +ve

Source: Bakir et al 2006

5. Summary

Measuring TB burden and the impact of control

• Routine surveillance the ultimate tool for evaluating TB epidemiology and control; completeness of reporting to be formally examined in all countries

• Disease prevalence surveys best for measuring prevalence (and change), not incidence

• Tuberculin surveys feasible where ARI high and BCG coverage low; better for comparisons (trends); Styblo'srule defunct

• TB death registrations need to be improved in all countries with high TB burden, and compared with data from NTPs; verbal autopsy needs validation

Measuring tuberculosis burden, trends and the impact of control

programmesLancet Infectious Diseases 2008

"By 2015, every country should be able to assess progress in control by evaluating the time trend in incidence, and the magnitude of reductions in either TB prevalence or deaths."