INT J TUBERC LUNG DIS 14(2):247–249 SHORT COMMUNICATION

© 2010 The Union

National estimate of HIV seroprevalence among

tuberculosis patients in India

P. K. Dewan,* D. Gupta,† B. G. Williams,‡ R. Thakur,§ D. Bachani,§ A. Khera,§ D. F. Wares,* S. Sahu,*

D. C. S. Reddy,* N. Raizada,¶ L. S. Chauhan†
* Office of the World Health Organization Representative to India, New Delhi, † Central Tuberculosis Division,
Directorate General of Health Services, Ministry of Health and Family Welfare, Government of India, New Delhi, India;
‡ Consultant, Geneva, Switzerland; § National AIDS Control Organization, Ministry of Health and Family Welfare,

Government of India, New Delhi, ¶ Foundation for Innovative New Diagnostics, New Delhi, India

SUMMARY

The national estimate for human immunodeficiency vi- estimate that among the 1.96 million incident TB cases
rus (HIV) prevalence among tuberculosis (TB) patients in 2007, 4.85% (95%CI 4.12–5.73) or 95 240 (95%CI
in India has previously been estimated indirectly from 80 730–112 478) were HIV-infected. With these esti-
global data. To derive an improved national estimate mates from local data, the national programme can bet-
from local data, we correlated district-level HIV surveil- ter plan TB-HIV collaborative activities and monitor ef-
lance data from antenatal clinics and TB diagnostic forts to detect HIV infection in this large population.
centres, and applied this correlation to state-level HIV K E Y W O R D S : HIV; tuberculosis; surveillance; India
prevalence estimates for the antenatal population. We

THE ADVERSE INTERACTION between human sites to derive a national estimate for HIV prevalence
immunodeficiency virus (HIV) infection and tubercu- in TB patients.
losis (TB) poses difficult challenges to public health
programmes.1 Detection of HIV infection among TB
METHODS
patients offers the opportunity to deliver prompt HIV
care, such as cotrimoxazole prophylaxis and antiretro- Data on HIV prevalence among TB patients were
viral treatment, which can reduce suffering and death. taken from a 15-district survey conducted by the
Precise estimation of the burden of HIV among TB national TB and AIDS control programmes in 2006–
patients is thus critical to effectively monitor TB-HIV 2007. For this survey, all districts in India were strati-
interventions. fied into three groups according to ANC HIV surveil-
India has one of the world’s highest burdens of lance results (0–0.5%, 0.5–1% and >1%), with five
both TB (1.96 million cases annually) and HIV infec- districts per strata being randomly selected.4 Surveys
tion (2.31 million prevalent cases).2,3 Based on the enrolled all new notified TB patients, including smear-
observed relationship between HIV prevalence in the negative and extra-pulmonary TB patients, for HIV
general population and among TB patients in a vari- testing. Data from ANC site HIV sentinel surveillance
ety of settings worldwide, the World Health Organi- and state-level ANC HIV prevalence estimates were
zation (WHO) estimated that 1.2% of incident TB taken from published surveillance reports.5 The Na-
cases in India in 2006 were HIV-infected.2 Until now, tional AIDS Control Organization (NACO) conducts
however, national data have not been available to ANC HIV surveillance at 646 sites across 476 (77%)
validate this estimation. Recent HIV sentinel surveil- of India’s 622 districts, with 1–2 sites per district. We
lance in TB diagnostic centres in India provides an used 3-year averages (2004–2006) of district and
opportunity to use local empirical data to develop a state ANC HIV prevalence estimates to minimise the
more accurate national estimate of the prevalence of influence of outlier values. TB notification data for
HIV among TB patients. We evaluated the relation- 2007 for states and national level were taken from
ship between HIV prevalence in TB patients at sentinel published TB programme reports,6 and national TB
surveillance sites and the prevalence of HIV among incidence estimates were taken from the WHO global
women attending antenatal clinic (ANC) surveillance report.2

Correspondence to: Puneet Dewan, World Health Organization, Southeast Asia Regional Office, World Health House,
Indraprastha Estates, Mahatma Gandhi Marg, New Delhi 110011, India. Tel: (+91) 11 4259 5600 ext 23253. Fax: (+91)
11 2338 2252. e-mail: [email protected]
Article submitted 7 June 2009. Final version accepted 2 September 2009.
248 The International Journal of Tuberculosis and Lung Disease

To determine the correlation between HIV in TB When this OR was applied to state-level ANC HIV
patients and HIV in ANC sentinel surveillance, we prevalence estimates, the estimated HIV prevalence
first assessed the linearity of the relationship for the among TB patients varied substantially by state (range
15 surveyed districts by sum-of-squares linear regres- 0.49–13.46%) (data not shown). Summed, this yielded
sion of the logit-transformed data. We then used three a national HIV prevalence among notified TB pa-
separate procedures, linear regression, the Mantel- tients of 4.85% (95%CI 4.12–5.73). This translated
Haenszel method and a random effects model, to ex- to 62 849 (95%CI 53 274–74 223) HIV-infected TB
press the relationship as an odds ratio (OR). In linear patients among the 1.294 million incident TB cases re-
regression, the y-intercept of the logit-transformed ported to the national programme in 2007, or 95 240
data was used to calculate the summary OR. The OR HIV-infected TB patients (95%CI 80 730–112 478)
derived from each approach was applied to NACO- among the larger population of 1.962 million inci-
estimated state-level ANC HIV prevalence to estimate dent TB cases estimated by the WHO.
the percentage of HIV among TB patients by state.
State-level estimates of HIV burden among notified
DISCUSSION
TB patients were summed to generate a national esti-
mate of HIV prevalence among notified TB patients. These estimates suggest that the 2007 HIV preva-
The analysis was conducted using Excel (MicroSoft, lence among TB patients in India was 4.85%. This is
Redmond, WA, USA) and StatsDirect 2.7 (StatsDirect four times higher than previously estimated, and sug-
Ltd, Cheshire, UK). gests that TB patients may be an efficient source for
HIV case-finding efforts in India. These results have
been used to guide the state-by-state intensification
RESULTS
of joint TB-HIV collaborative activities across the
The 15 surveyed districts included a total of 5995 TB country and to monitor the effectiveness of HIV case-
patients tested for HIV in 2006–2007, and 28 396 finding among TB patients.
women tested for HIV in 30 ANC surveillance sites This analysis showed that the use of local surveil-
over 2004–2006. A linear model of logit-transformed lance data can improve estimates of TB-HIV burden.
data (Figure) indicated that the slope was not signifi- In 2009, the WHO refined estimates of HIV among
cantly different than one (slope 0.97, 95% confidence TB patients using routinely reported data from multi-
interval [CI] 0.66–1.29, R2 = 0.77), which supported ple countries;7 the revised WHO estimate of 5.25%
the interpretation of linearity. With the slope con- for India was much closer to our estimate, supporting
strained to one, the linear model yielded an estimated the validity of our approach. India has now imple-
y-intercept of 2.26 (standard error 0.22); exponenti- mented a policy of provider-initiated HIV counselling
ated, this yielded an OR of 9.62 (95%CI 7.76–11.93). and testing for all TB patients in higher HIV preva-
Calculation of the summary OR by the Mantel- lence states.8 If HIV testing uptake is high, routine re-
Haenszel technique (9.78, 95%CI 8.29–11.55) and a porting of HIV status for all TB patients would pro-
random effects model (9.76, 95%CI 7.94–12.00) vide even better information on which to base future
yielded similar results. For the purposes of disease bur- disease burden estimates.
den estimation, the Mantel-Haenszel OR was used. Our analysis is subject to a number of limitations.
We derived these estimates from ANC results, which
intrinsically have a number of weaknesses. ANC sur-
veillance is conducted at just 1–2 sites per district and
shows substantial year-to-year variation; our use of
3-year averages may have helped to reduce the influ-
ence of outlier values. ANC HIV prevalence is known
to be higher than the general population, but general
population HIV prevalence is not available at the dis-
trict level; we thus consistently used ANC HIV preva-
lence from district and state levels. These estimates of
HIV prevalence in TB patients are expected to be
valid only for the national and state levels, and should
not be applied to the district level for programme
monitoring due to the expected heterogeneity of HIV
prevalence in the general population.

Figure Log-odds plot of the relationship between the preva- References

lence of HIV in tuberculosis patients and patients attending an- 1 Havlir D V, Getahun H, Sanne I, Nunn P. Opportunities and
tenatal clinic surveillance sites. R2 = 0.77. HIV = human immuno- challenges for HIV care in overlapping HIV and TB epidemics.
deficiency virus; TB = tuberculosis; ANC = antenatal clinic. JAMA 2008; 300: 423–430.
 HIV seroprevalence among TB patients in India 249

2 World Health Organization. Global tuberculosis control: sur- 6 Ministry of Health and Family Welfare, Government of India.
veillance, planning, financing. WHO report 2008. WHO/HTM/ TB India 2008: RNTCP status report. New Delhi, India: Cen-
TB/2008.393. Geneva, Switzerland: WHO, 2008. tral TB Division, Directorate General of Health Services, Min-
3 National AIDS Control Organization. HIV sentinel surveillance istry of Health and Family Welfare, Government of India,
and HIV estimation in India, 2007. A technical brief. New Delhi, 2008.
India: Ministry of Health and Family Welfare, Government of 7 World Health Organization. Global tuberculosis control: epide-
India, 2008. miology, strategy, financing. WHO/HTM/TB/2009.411. Geneva,
4 Raizada N, Chauhan L S, Khera A, et al. HIV seroprevalence Switzerland: WHO, 2009.
among tuberculosis patients in India, 2006–2007. PLoS One 8 Ministry of Health and Family Welfare, Government of India.
2008; 3: e2970. National framework for joint TB/HIV collaborative activities:
5 Ministry of Health and Family Welfare, Government of India. February 2008. New Delhi, India: Ministry of Health and Fam-
HIV sentinel surveillance and HIV estimation, 2006. New Delhi, ily Welfare, Government of India, 2008.
India: National AIDS Control Organization, Ministry of Health
and Family Welfare, Government of India, 2008.

RÉSUMÉ

Antérieurement, en Inde, l’estimation nationale de la tions de prévalence du VIH par état dans la population
prévalence du virus de l’immunodéficience humaine prénatale. Nous estimons que sur les 1,96 millions de
(VIH) chez les patients tuberculeux avait été établie in- nouveaux cas de TB en 2007, 4,85% (IC95% 4,12–
directement à partir des données mondiales. Pour obte- 5,73), soit 95 240 (IC95% 80 730–112 478), étaient in-
nir une amélioration de l’estimation nationale à partir fectés par le VIH. Grâce à ces estimations issues de don-
des données locales, nous avons corrélé les données de nées locales, le programme national peut améliorer la
surveillance des districts dans les polycliniques préna- planification des activités de collaboration VIH-TB et
tales et dans les centres de diagnostic de tuberculose surveiller les efforts de détection de l’infection VIH dans
(TB), et avons appliqué cette corrélation aux estima- cette grande population.

RESUMEN

El cálculo de la frecuencia de infección por el virus de la correlación al cálculo estatal de prevalencia de infección
inmunodeficiencia humana (VIH) a escala nacional en por el VIH en la población de la consulta prenatal. Se
los pacientes con tuberculosis (TB) en la India se ha calculó que en 1,96 millones de casos nuevos de TB en
realizado en forma indirecta a partir de los datos mun- el 2007, 4,85% (IC95% 4,12–5,73), o 95 240 personas
diales. Con el propósito de obtener un dato nacional más (IC95% 80 730–112 478), padecían infección por el VIH.
preciso partiendo de los datos locales, se correlaciona- Con estas cifras a partir de los datos locales, el programa
ron los datos de la vigilancia de la infección por el VIH nacional puede planear mejor las actividades conjuntas
a escala distrital en los consultorios de consulta prenatal contra la TB y el VIH y vigilar los esfuerzos de detección
y los centros de diagnóstico de la TB y se aplicó esta de la infección por el virus en esta vasta población.