Am. J. Trop. Med. Hyg., 75(1), 2006, pp. 49–54
Copyright © 2006 by The American Society of Tropical Medicine and Hygiene
COMPARISON OF ALTITUDE EFFECT ON MYCOBACTERIUM TUBERCULOSIS
INFECTION BETWEEN RURAL AND URBAN COMMUNITIES IN PERU
MAYUKO SAITO, WILLIAM K. PAN, ROBERT H. GILMAN* CHRISTIAN T. BAUTISTA, SAPNA BAMRAH,
CHRISTOPHER A. MARTÍN, SIMON J. TSIOURIS, D. FERMÍN ARGÜELLO, AND GABRIELA MARTINEZ-CARRASCO
Department of International Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland; Asociación Benéfica,
Proyectos en Informática, Salud, Medicina y Agricultura (A. B. PRISMA), Lima, Peru; Department of Microbiology, Universidad de
Cayetano Heredia, Lima, Peru; US Military HIV Research Program and the Henry M. Jackson Foundation, Rockville, Maryland
Abstract. The mechanism of high altitude effect on tuberculosis (TB) infection has not been fully established. We
previously reported a lower positive tuberculin skin test (TST) prevalence in high altitude villages compared with sea
level communities in Peru. In this study, four additional communities were tested to assess whether decreased TB
transmission was also in urban environments at high altitude. TST results from 3,629 individuals in nine communities
were analyzed using generalized estimating equations to account for community clustering. Positive TST prevalence was
not significantly different between the urban highland and the urban non-highland communities after adjusting for age,
household contacts with a TST-positive person or a TB case, and presence of a Bacillus Calmette-Guérin vaccination
scar. The effect of population concentration and increased contact with active TB overwhelmed the protective effect of
altitude in urban highlands. Highland cities require the same preventive efforts against TB as non-highland communities.
with ∼520 inhabitants located 150 km northeast of the capital
city of Lima.5
Urban highland communities. Huascahura is a peri-urban
village with ∼1,760 inhabitants located near Ayacucho (population: 140,500). San Jeronimo is a peri-urban shantytown
with 19,120 habitants located in the southern part of the center of Cusco (population: 342,000).
Rural non-highland communities. Buen Villa Pastor is a
rural jungle village with ∼270 inhabitants in the Amazon basin.5 San Carlos is a rural jungle village with 176 inhabitants
similar to Buen Villa Pastor.5
Urban non-highland communities. Las Pampas is an urban
desert shantytown with > 40,000 inhabitants in southern
Lima.5 Sachaca is an urban town with 16,570 inhabitants near
the southern part of Arequipa, the second largest city of Peru
(population: 677,000). Cerro Colorado is an urban shantytown (population: 70,000) located in the northern area of
Arequipa.
Residents in the shantytowns are impoverished, earning
money from temporary small businesses, selling seasonal agricultural products, or working for construction companies.
Their houses have few windows and are made of stones or
adobe. Populations living in these nine sites are nearly all
Mestizo, descendants of mixed European and Native American ancestry. Climate in the highlands is relatively dry and
cold, with a rainy season lasting for 3 mo/yr. The jungle areas
are humid. The geographic characteristics of each site are
shown in Table 1 and Figure 1.
Field operations. The participants in Vichaycocha were obtained from a health post, and the data do not include all
family information.5 The sample households were chosen in
Quilcas and Las Pampas from computerized census data using
random sampling function of FoxPro ver.2.6 (Microsoft, Redmond, WA), whereas the whole village population was invited to participate in San Carlos and Buen Villa Pastor. In
Sachaca, Cerro Colorado, Huascahura, and San Jeronimo, a
part of the community was selected, and all houses were included.
A household was defined as persons who share common
living space such as a kitchen, living room, and bathroom. All
family members excluding infants < 6 months of age and
people who were temporally staying in the house (< 6
INTRODUCTION
Worldwide, tuberculosis (TB) affects eight million people
annually, yet 95% of TB cases and 98% of TB deaths occur in
resource-limited countries.1 TB and multi-drug–resistant Mycobacterium tuberculosis (MTB) have re-emerged because of
increases of human immunodeficiency virus (HIV) infection
in developing countries.2
Peru is one of the resource-limited countries in which TB is
endemic. The estimated prevalence of active TB in Peru is 233
cases per 100,000 inhabitants in 2004,3 despite a low prevalence of HIV infection of 0.5% among adults.4
In a previous study, we observed a lower TB transmission
in high altitude rural communities compared with both urban
and rural sea level communities in Peru.5 This study tested
four additional communities at varying altitudes to examine
the effects of high altitude and urban environment on tuberculin skin test (TST) positivity.
MATERIALS AND METHODS
Study design. A previous study sampled five study sites,
two rural highland, one urban lowland, and two rural lowland
sites,5 but did not include urban highland communities. In this
study, two urban highland communities and two urban midhighland communities were added to determine if positive
TST prevalence decreased with altitude in urban highland
communities
Study sites. Nine communities were studied (Figure 1). Five
of these have been described in detail in a previous study5.
Two urban non-highland (altitude < 3,000 m) shantytowns,
Sachaca, and Cerro Colorado, and two highland (altitude
ⱖ 3,000 m) peri-urban shantytowns, San Jeronimo and Huascahura, were newly studied.
Rural highland communities. Quilcas is a rural Andean village with ∼1,280 inhabitants located 45 km from Huancayo
(population: 327,000)5. Vichaycocha is a rural Andean village
* Address correspondence to Robert H. Gilman, Department of International Health, The Johns Hopkins School of Pubic Health, 615
North Wolfe Street, Room W5515, Baltimore, MD 21205. E-mail:
[email protected]
49
50
SAITO AND OTHERS
FIGURE 1. Study sites and natural geographic regions (desert, highland, and jungle) of Peru. Small dots and squares with two capital letters
represent study sites (BP, Buen Villa Pastor; SC, San Carlos; VC, Vichaycocha; QC, Quilcas; PM, Las Pampas; HU, Huascahura; SJ, San
Jeronimo; CC, Cerro Colorado; SA, Sachaca). Large dots represent major cities near the study sites. The square represents the capital city of
Lima.
months) were invited to the study. Written informed consent
was obtained from all adult participants (ⱖ 18 years of age)
and from the parents or guardians of children. All participants were interviewed with structured questionnaires regarding recent (within 5 years) exposure to active TB cases
and general conditions of health.
Trained field workers performed the TST using an intradermal injection of 5 tuberculin units (Tubersol; Connaught
Laboratories, Ontario, Canada) in 0.1 mL on the volar surface of the forearm. Tuberculin vials (1 mL) were kept refrigerated and carried in a cooler box with ice packs. Induration produced by injections was measured 48–72 hours after
administration using the pen method.6 Induration size ⱖ 10
mm was considered to be TST positive, according to the Peruvian Ministry of Health7 and American Thoracic Society
recommendations for people born in countries with high
prevalence of TB.8 Although different field workers worked
in different study sites, all fieldworkers underwent standardized training for reading the TST by an experienced research
nurse, and the field work was supervised by one of the authors
(M.S.). Participants with a positive TST were evaluated for
evidence of active TB as described in the previous study.5
This study was approved by the ethical review boards of
Asociación Benéfica PRISMA, Lima, Peru, and the Johns
Hopkins Bloomberg School of Public Health, Baltimore, MD.
Statistical analysis. Demographic characteristics of the
study population and TST results were compared among
study sites. Differences in continuous variables were evaluated by Kruskal-Wallis and Mann-Whitney U tests. ␹2 tests
were used to test differences in categorical or binary variables. Educational level was categorized as no education (< 6
years), primary education (6 years completed), and secondary
education (ⱖ 12 years completed). Body mass index (BMI)
was calculated for six communities (data in San Carlos, Huascahura, and Vichaycocha were not available). The positive
TST prevalence was compared between rural highland, rural
non-highland, urban highland, and urban non-highland communities using a Wald ␹2 test computed from a generalized
estimating equations (GEE) model to account for the family
clustering effect.9
Characteristics of the study population were compared between individuals living in the highland (ⱖ 3,000 m) and nonhighland (< 3,000 m) communities. Populations were also divided into rural or urban based on distance and ease of travel
51
ALTITUDE EFFECT ON TUBERCULOSIS INFECTION IN PERU
TABLE 1
Study site description, demographic characteristics of the study population, and positive TST prevalence
Desert
Tropical jungle
Mid-highland
Highland
Community
PM
BP
SC
SA
CC
HU
SJ
QC
VC
Altitude (meter)
Urban/rural
Number of households
participated
Number of participants
tested (% targeted)
Female/male ratio
Median age in years‡
Number of children < 18
years old (%)‡
Education level in adults‡
No education (%)
Primary (%)
Secondary or more (%)
BMI of adults [range]‡§
Median number of people
in the household
TST positive prevalence
(%) [95%CI]‡¶
140
Urban
307*
110
Rural
40
110
Rural
31
2,240
Urban
77
2,650
Urban
68
3,000
Urban
56
3,240
Urban
117
3,330
Rural
100*
3,500
Rural
NA†
1238 (85)
186 (70)
149 (85)
308 (81)
226 (72)
241 (80)
458 (81)
453 (67)
370 (72)
1.2
15
690 (56)
1.0
19
89 (48)
1.0
15
80 (54)
1.1
18
151 (49)
1.5
20
104 (46)
1.3
17
122 (51)
1.1
16
251 (55)
1.1
13
266 (59)
1.1
16
189 (51)
40/544 (7)
217/544 (40)
287/544 (53)
24 [22–27]
5 [4–6]
8/96 (9)
57/96 (59)
31/96 (23)
23 [21–25]
5 [4–8]
6/66 (9)
43/66 (65)
17/66 (26)
—
5 [4–7]
26/148 (17)
81/148 (55)
41/148 (28)
24 [22–27]
5 [4–6]
23/122 (19)
40/122 (33)
59/122 (48)
25 [23–28]
4 [3–5]
61/89 (51)
36/89 (30)
22/89 (19)
—
5 [3.5–7]
70/207 (34)
69/207 (33)
68/207 (33)
24 [22–27]
5 [3–6]
27/187 (15)
111/187 (59)
49/187 (26)
23 [21–25]
4 [3–6]
21/172 (12)
105/172 (61)
46/172 (27)
—
NA†
25 [23–28]
33 [26–42]
31 [25–38]
26 [22–32]
22 [17–29]
23 [17–29]
24 [20–29]
7 [4–10]
6 [4–9]
PM, Las Pampas; BP, Buen Villa Pastor; SC, San Carlos; SA, Sachaca; CC, Cerro Colorado; HU, Huascahura; SJ, San Jeronimo; QC, Quilcas; VC, Vichaycocha.
* Households are randomly selected.
† NA, not available. Participation in the study was requested at the health post.
‡ P < 0.05.
§ Weight and height were not available in San Carlos, Huascahura, and Vichaycocha; [range] ⳱ percentile range.
¶ 95%CIs were corrected for family clustering effect using generalized estimating equations (GEE) model.
to the city center. Additional background information such as
experience of living with other person with active TB, presence/absence of Bacillus Calmette-Guérin vaccination (BCG)
scar, and living in an urban area or rural area was also compared. The Wald ␹2 test computed from GEE was used to test
differences to account for community clustering.
To find factors associated with positive TST prevalence,
univariate analyses were performed on the entire study
sample for sex, age, history of living with a person with active
TB, presence/absence of BCG scar, number of people per
household, and living in rural highland compared with living
in urban highland, living in rural non-highland, and living in
urban non-highland. Educational level and BMI were analyzed only for adults ⱖ 18 years of age. Multiple analyses
were performed on 3,161 people (1,466 adults and 1,695 children) from eight communities except Vichaycocha, because
data from Vichaycocha lacked household information and
was thus excluded from the regression model. Model estimation was performed using GEE to account for community
clustering. Adjusted odds ratios (AORs) were compared between each of the different types of communities (rural highland, rural non-highland, urban highland, and urban nonhighland).
Non-parametric tests, ␹2 test, and GEE models for family
clustering were performed using Stata 8.0 (Stata Corp., College Station, TX). GEE models for community clustering
were computed using SAS statistical software 9.1 (SAS Institute, Cary, NC).
RESULTS
Distribution of demographic characteristics. In this study,
data from a total of 3,629 people from the nine sites were
analyzed. The number of households, participants, and the
participation rates to the target population are shown in the
Table 1. Median age in years, proportion of the child population, educational level, BMI, and the median number of
people per household were significantly different among the
study sites (Table 1). When we compare the background characteristics among the study population from the four highland
communities and five non-highland communities, there was
no significant difference in the demographic factors such as
sex, age, living with other person with active TB or not, educational level, BMI, presence/absence of a BCG scar, number of person within the household, and living in an urban
area or rural area.
Positive TST prevalence. Overall positive TST prevalence
in the study population was 21.1% (95% confidence interval
[CI]: 19.7–22.3). The lowest positive TST prevalence was
found in Vichaycocha (6%, rural highland), and the highest
prevalence was in Buen Villa Pastor (33%, rural nonhighland; Table 1). Positive TST prevalence was lower in rural highland communities compared with rural non-highland
communities (OR ⳱ 0.14, P < 0.001); however, there was no
significant difference between urban highland and urban nonhighland communities (OR ⳱ 0.92, P ⳱ 0.500; Figure 2). In
highland communities, positive TST prevalence was significantly higher in urban communities than in rural communities
(OR ⳱ 2.84, 95% CI: 2.00–4.03, P < 0.001). In contrast, in the
non-highland sites, positive TST prevalence was higher in rural than urban communities (OR ⳱ 1.43, 95% CI: 1.09–1.89,
P ⳱ 0.011; Figure 2).
Low TST positive prevalence in rural highland. In the multiple regression analysis, the positive TST prevalence remained significantly lower in the rural highland communities
than in the urban highland, rural non-highland, and urban
non-highland sites after adjusting for sex, age, living with another person with a positive TST, having lived with another
person with active TB, having a BCG scar, and number of
person within the household (Table 2). This trend was con-
52
SAITO AND OTHERS
FIGURE 2. Prevalence of positive tuberculin skin test reaction by altitude and type (urban or rural) of the community. R, rural community;
U, urban community. GEE model was used to account for family clustering.
TABLE 2
Crude and adjusted ORs for TST positivity in total study popuation*
Variables
OR
AOR
95% CI
Male sex
Age group (in years)
0–6
7–17
18–31
ⱖ 32
Living with other person
with a positive TST
Living or has lived with
other person with
active TB
Presence of BCG scar
Number of person
within the household
Living in rural highland
Living in urban highland
Living in rural
non-highland
Living in urban
non-highland
1.09
1.34†
[1.05–1.72]
Reference‡
4.76†
16.08†
33.37†
1.30†
Reference
4.75†
19.91†
42.17†
1.84†
[2.43–9.28]
[9.40–42.17]
[22.52–78.96]
[1.42–2.38]
1.72†
1.81†
[1.29–2.56]
1.88†
0.98
1.77†
0.99
[1.36–2.30]
[0.95–1.04]
Reference
4.25†
6.60†
Reference
3.52†
5.16†
[3.03–4.09]
[4.72–5.63]
4.63†
3.96†
[3.48–4.37]
* OR and AOR were computed using GEE model. AOR was adjusted for all variables
in the model. Missing data in the regression analysis inlcuded 370 individuals in Vichaycocha, 80 people without BCG scar data, and 18 lacked information on their TB contact
status.
† P < 0.05.
‡ The categories with Reference described the baseline group for OR calculation.
stant when adults ⱖ 18 years of age and children < 18 years
old were analyzed separately (Tables 3 and 4).
Comparison of altitude effect between rural and urban
communities. When we compared between rural non-highland and rural highland communities, the non-highland population had a significantly higher positive TST prevalence than
the highland population (AORs in total population, adults,
and children were 5.16, 5.50, and 5.75, respectively; Tables
2–4). In contrast, among the urban communities, the differences between non-highland and highland were not significant in the total population (difference in the AOR ⳱ 1.12,
95% CI: 0.95–1.33) and the adult population (difference in
the AOR ⳱ 1.13, 95% CI: 0.98–1.30). In the children, however, the AOR of positive TST prevalence was lower in the
non-highland community group than in highland group (difference in the AOR ⳱ 0.82, 95% CI: 0.71–0.94).
Comparison of positive TST prevalence between rural and
urban communities. The positive TST prevalence was significantly higher in urban communities than in rural communities
in the highlands (difference in the AOR ⳱ 3.52, 95% CI:
3.03–4.09), whereas in the non-highland environment, the
positive TST prevalence was higher in rural communities than
in the urban communities. The difference in AOR of these
non-highland communities was significant for the whole
53
ALTITUDE EFFECT ON TUBERCULOSIS INFECTION IN PERU
TABLE 3
Crude and adjusted ORs for TST positivity in adult population (⭌ 18
years old)*
TABLE 4
Crude and adjusted ORs for TST positivity in children (0–17 years old)*
Variables
Variables
OR
AOR
95% CI
Male sex
Age group (in years)
18–31
ⱖ 32
Living with other person
with a positive TST
Living or has lived with
other person with
active TB
Educational level
No education
Primary
> Secondary
BMI by 25th percentile
< 22
22–23.9
24–26.4
ⱖ 26.5
Presence of BCG scar
Number of person within
the household
Number of person within
the household
Living in rural highland
Living in urban highland
Living in rural nonhighland
Living in urban nonhighland
1.65†
1.65†
[1.30–2.10]
Reference‡
2.00†
1.74†
Reference
2.13†
1.69†
[1.51–3.56]
[2.33–1.23]
1.58†
1.54†
[1.23–1.92]
Reference
0.99
1.13
Reference
1.28
1.12
[0.90–1.81]
[0.87–1.62]
Reference
0.98
0.91
1.14
2.04†
1.6
Reference
1.01
0.86
0.90
1.64†
1.00
[0.58–1.35]
[0.55–1.31]
[0.74–1.35]
[1.27–2.04]
[0.94–1.07]
1.06
1.00
[0.94–1.07]
Reference
4.09†
8.13†
Reference
3.36†
5.50†
[3.02–373]
[4.96–6.11]
4.59†
3.79†
[3.39–4.24]
* OR and AOR were computed using GEE model. AOR was adjusted for all variables in
the model. BMI data were not available from 369 adults in three communities (Vichaycocha,
Huascahura, and San Carlos) and 16 adults from other communities. Fifty-one adults were
missing one of the covariates.
† P < 0.05.
‡ The categories with Reference describe the baseline group for OR calculation.
population (1.30, 95% CI: 1.14–1.49) and adults (1.45, 95%
CI: 1.27–1.66), but not for children (1.30, 95% CI: 0.81–2.09).
Other factors associated with positive TST results. In
adults, male sex, older age group (ⱖ 32 years old), living with
another person with a positive TST, having lived with another
person with active TB, and having a BCG scar were all associated with a positive TST in both univariate and multiple
regression analyses (Table 3). In children, univariate analysis
indicated all variables were positively associated with a positive TST except male sex (Table 3). After adjusting for all
variables, having contact with a family member with active
TB and the number of persons within the same family were
not associated.
DISCUSSION
The protective effect of high altitude on positive TST
prevalence in a rural highland population, found in a previous
study, was nearly completely overcome in urban highland
populations. Crowding and increased contact with persons
with active TB infection are the probable causes for the reversal of the low transmission rates of TB infection that is
present in rural highland areas. Rural highland villages also
experience little migration into their population from the lowlands where TB disease is common.
Using GEE to estimate model parameters to account for
community clustering effects, the effects of associated factors
Male sex
Age group (in years)
0–6
7–17
Living with other person
with a positive TST
Living or has lived with
other person with
active TB
Presence of BCG scar
Number of person within
the household
Living in rural highland
Living in urban highland
Living in rural nonhighland
Living in urban nonhighland
OR
AOR
95% CI
1.06
1.03
[0.70–1.53]
Reference†
4.17‡
4.16‡
Reference
4.56‡
3.45‡
[2.37–8.77]
[1.63–7.31]
3.37‡
2.38
[0.91–6.23]
3.04‡
1.09‡
3.58‡
1.01
[1.96–6.56]
[0.96–1.07]
Reference
9.16‡
8.12‡
Reference
5.43‡
5.75‡
[4.21–7.00]
[3.27–10.12]
7.77‡
4.43‡
[3.69–5.31]
* OR and AOR were computed using GEE model. AOR was adjusted for all variables in
the model. One hundred eighty-nine children from Vichaycocha, 11 children without the
data on BCG scar, and 47 children without the data of contact with patients with active TB
were excluded from the multiple regression model.
† The categories with Reference describe the baseline group for OR calculation.
‡ P < 0.05.
were tested. High altitude had strong association with a decreased rate of TST positivity even after adjusting for age,
household active TB contacts, household positive TST contacts, BCG vaccination status, and community clustering. The
finding of greater TB infection risk in rural versus urban at
low altitude, with the opposite finding at high altitude, may be
caused by the high humidity present in the jungle and the
relative lack of health care facilities.
Limitations of this study are that the number of sites was
not large enough to show if there was a linear or threshold
protective effect of high altitude in rural areas. Even though
we accounted for the effect of contact with a known active TB
patient or another person with a TST-positive result in the
same household, it was not possible to account for the risk of
acquiring MTB infection from visitors, temporal immigrants
to the community, or unknown contacts. The non-significant
effect of an active TB contact (AOR ⳱ 2.4) in children was
most likely caused by a small sample size for this variable. In
addition, the relatively wide CIs for the adult TST prevalence
data are also probably related to a low sample size in each
community.
The mechanism of the high altitude effect on TB disease or
MTB infection has not been fully established. The number of
studies on this topic decreased after the discovery of the antiTB drugs in the 1940s. Altitude effects on TB transmission
rates have only been examined in one previous study in Darjeeling10 reported that positive TST prevalence was decreased
in villages above 1,000 m compared with those below this
altitude. That study, however, did not control for confounding
factors. In terms of TB disease, which our study did not examine, two recent studies from Mexico showed an inverse
relationship between altitude and TB disease morbidity11 and
mortality.12 Although the Mexican findings of an association
between TB disease and altitude could be caused by a decreased severity of the TB disease at high altitude, an effect of
altitude on MTB transmission is equally likely.
There are several mechanisms by which MTB infection
might be prevented in high altitude climate, such as lower
54
SAITO AND OTHERS
oxygen tensions13 or stronger exposure to ultraviolet (UV)
light, which may inhibit MTB growth.14 However, houses in
the highland have relatively small and few windows; therefore, UV rays may not reach inside houses, where transmission is most likely to occur. The dry climate in high altitude
may increase the susceptibility of Mycobacterium to UV
light,15 reducing TB transmission. Further detailed studies are
needed to determine which mechanism(s) are primarily responsible for the decreased transmission of TB in the highlands.
The protective effect of high altitude on MTB infection
found in rural highland communities was not found in urban
highland communities. Assuming that high TB transmission
rates reflect high rates of active TB, urban highland communities will require the same measures against TB as nonhighland communities.
Received December 19, 2005. Accepted for publication March 9,
2006.
Acknowledgments: We thank L. Cabrera, R. Montoya, and P. Maguiña for study coordination, M. Varela for data management, and A.
Griffin, J. B. Phu, D. Sarah, and A. Sebastian for technical support.
We also thank the communities of Quilcas, Vichaycocha, Las Pampas
de San Juan de Miraflores, San Carlos, Buen Villa Pastor, Sachaca,
Cerro Colorado, San Jeronimo, and Huascahura for cooperation.
Portions of the data contained in this manuscript were presented at
the ASTMH 53th Annual Meeting, Miami Beach, Florida, November
7–11, 2004.
Financial support: These studies were supported by D43 TW07646-5
Tutorial in Tropical Health at JHU/Peru Overseas Site NIH/NIAID,
D43 TW066581 Infections Diseases Training Program in Peru-Global
Research Fund DHHS/Fogarty International, and Research Support
Grant from St. Luke’s Life Science Institute.
Disclaimer: The opinions and assertions made by the authors do not
reflect the official position or opinion of the US Department of the
Army or of any of the other organizations listed.
Authors’ addresses: Mayuko Saito, Will K. Pan, and Robert H. Gilman, Department of International Health, The Johns Hopkins School
of Public Health, 615 North Wolfe Street, Room W5515, Baltimore,
MD 21205, E-mails: [email protected] and [email protected].
Christian T. Bautista (present affiliation), Department of Epidemiology and Threat Assessment, The US Military HIV Research Program, and the Henry M. Jackson Foundation, 1 Taft Court, Suite 250,
Rockville, MD 20850, E-mail: [email protected]. Sapna
Bamrah, Christopher A. Martín, Simon J. Tsiouris, D. Fermín
Argüello, and Gabriela Martinez-Carrasco, Asociación Benéfica,
Proyectos En Informática, Salud, Medicina y Agricultura (A.B.
PRISMA), Carlos Gonzales 251, Urb. Maranga, San Miguel, Lima 32,
Peru.
REFERENCES
1. Zumla A, Mwaba P, Squire SB, Grange JM, 1999. The tuberculosis pandemic–which way now? J Infect 38: 74–79.
2. Murray JF, 1997. Tuberculosis and HIV infection: global perspectives. Respirology 2: 209–213.
3. World Health Organization, 2005. WHO Report 2005: Global
Tuberculosis Control—Surveillance, Planning, Financing.
Geneva, Switzerland: World Health Organization.
4. The Joint United Nations Programme on HIV/AIDS/World
Health Organization/United Nations Children’s Fund, 2004.
Epidemiological fact sheet on HIV/AIDS and sexually transmitted infections: 2004 update. Available at: http://
www.who.int/GlobalAtlas/predefinedReports/EFS2004/
index.asp Accessed September 15, 2005.
5. Olender S, Saito M, Apgar J, Gillenwater K, Bautista CT, Lescano AG, Moro P, Caviedes L, Hsieh EJ, Gilman RH, 2003.
Low prevalence and increased household clustering of Mycobacterium tuberculosis infection in high altitude villages in
Peru. Am J Trop Med Hyg 68: 721–727.
6. Jordan TL, Sunderam G, Thomas L, Reichman LB, 1987. Tuberculin reaction size measurement by the pen method compared
to traditional palpation. Chest 92: 234–236.
7. Dirección del Programa Nacional de Control de Enfermedades
Transmisibles Control de la Tuberculosis, 2000. Actualizacion
en la Tecnica de Aplicación, Lectura e Interpretación de la
Prueba de Tuberculina. Lima, Peru: Ministry of Health.
8. American Thoracic Society and the Center for Disease Control
and Prevention, 2000. Diagnostic standards and classification
of tuberculosis in adults and children. Am J Respir Crit Care
Med 161: 1376–1395.
9. Zeger SL, Liang KY, 1986. Longitudinal data analysis for discrete
and continuous outcomes. Biometrics 42: 121–130.
10. Bates LR, Busk T, Palmer CE, 1951. Research Contributions of
BCG Vaccination Programs: Tuberculin Sensitivity at Different
Altitudes of Residence. Copenhagen, Denmark: World Health
Organization, Tuberculosis Research Office, 1427–1442.
11. Vargas MH, Furuya ME, Perez-Guzman C, 2004. Effect of altitude on the frequency of pulmonary tuberculosis. Int J Tuberc
Lung Dis 8: 1321–1324.
12. Perez-Padilla R, Fronco-Marina F, 2004. The impact of altitude
on mortality from tuberculosis and pneumonia. Int J Tuberc
Lung Dis 8: 1315–1320.
13. Sever JL, Youmans GP, 1957. The relation of oxygen tension to
virulence of tuberculosis bacilli and to acquired resistance in
tuberculosis. J Infect Dis 101: 193–202.
14. Riley RL, Knight M, Middlebrook G, 1976. Ultraviolet susceptibility of BCG and virulent tubercle bacilli. Am Rev Respir Dis
113: 413–418.
15. Ko G, First MW, Burge HA, 2000. Influence of relative humidity
on particle size and UV sensitivity of Serratia marcescens and
Mycobacterium bovis BCG aerosols. Tuber Lung Dis 80: 217–
228.