The World Health Organization (WHO) estimated that there were 8.8 million new cases of tuberculosis (TB) in 2005.1 Twelve of the 15 countries with the highest TB incidence are in Africa2 where it plays a prominent and intermingled role with the human immunodeficiency virus (HIV); TB is the leading cause of death for those living with HIV/AIDS globally.2,3 In children, the issue is particularly acute as HIV increases both a child’s likelihood of TB exposure and the risk of developing active disease.4,5
The Gold Standard for diagnosis of TB consists of identification of mycobacterium tuberculosis in culture.6 Use of this Gold Standard in children is limited by several factors. First, the paucibacillary nature of the illness in children means that only up to 60% of TB cases can be confirmed with culture techniques,7 not taking account of any impact that HIV disease might have on culture growth. Second, specimen collection in pediatric populations is very difficult, as it requires invasive techniques and personnel and facilities are not well equipped to obtain them.6 Further, in resource-constrained environments, emphasis to-date has been on smear diagnosis capacity at country level because culture capacity has been perceived as too expensive and too technically difficult, but children have difficulty providing sputum samples.6,8 The low organism burden disease in children and the lack of culture capacity in most resource-constrained settings have resulted in profound challenges for clinicians in the verification of specific TB disease in pediatric populations. Scoring systems were consequently developed in attempt to overcome these diagnostic limitations, but without a functional Gold Standard these scoring systems remain unvalidated for sensitivity or specificity.9 Thus, the “real-life” scenario for the HIV pediatrician is that there is no reliably specific laboratory test available and only a nonvalidated scoring system to evaluate a sick child who, if they have active TB, may die rapidly without treatment.10
In spite of these obstacles, there is an urgent need for data on the clinical burden of pediatric TB,6 especially in the context of HIV infection. It is imperative to elucidate both risk and protective factors associated with TB diagnosis in the HIV-infected child. The objectives of this study are therefore: (1) to describe the clinical burden of pediatric TB diagnosed in an HIV-infected population in sub-Saharan Africa; (2) to describe the social and clinical factors associated with a TB diagnosis; and (3) to describe the effect of combination antiretroviral treatment (cART) on the probability of being diagnosed with TB, among these HIV-infected children.
This study used a retrospective observational cohort design with prospectively collected data. The Moi University School of Medicine Institutional Review and Ethics Committee and the Indiana University Institutional Review Board approved this study and waived informed consent as part of a general approval for conducting retrospective analyses with de-identified data collected in the course of routine care.
Based in Eldoret, Kenya, the Academic Model Providing Access to Healthcare (AMPATH) was initiated in 2001 as a joint partnership between Moi University School of Medicine, the Indiana University School of Medicine, and the Moi Teaching and Referral Hospital. With financial support from USAID, the USAID-AMPATH Partnership was established in 2004. The initial goal of AMPATH was to establish an HIV care system to serve the needs of both urban and rural patients in western Kenya and to assess the barriers to, and outcomes of, combination antiretroviral therapy. Details of the development of this program have been described elsewhere.11 Briefly, the first urban and rural HIV clinics were opened in November 2001. To date the program has enrolled over 70,000 HIV-infected adults and children in 18 Ministry of Health facilities around Western Kenya. All HIV and TB-related care and treatment are free at the point of service for patients through AMPATH and the Kenyan Department of Leprosy, TB, and Lung Disease.
Sources of Data
Clinicians complete standardized forms capturing demographic, clinical, and pharmacologic information at each patient visit. These data are then hand-entered into the AMPATH Medical Records System, a secure computerized database designed for clinical management.12,13 Data entry is validated by random review of 10% of the data entered. At the time of registration, patients are provided with a unique identifying number, and all data were stripped of identifying information prior to analysis.
This study included patients aged 0 to 13 years, with confirmed HIV infection, at least one follow-up visit, and who were enrolled at one of the AMPATH clinics between December 2001 and January 2007.
The AMPATH clinical care protocols for managing HIV infection are consistent with WHO recommendations and have been described in detail elsewhere.14,15
All HIV-infected children are routinely screened at clinic enrollment for tuberculosis with a symptom screen, physical examination, and chest x-ray. Tuberculosis is diagnosed using the modified Kenneth-Jones scoring system.9,16,17 These criteria are consistent with the tuberculosis case definition for children under 15 years recommended by the WHO.18Table 1 summarizes the scoring system with the original and modified weights for each criteria. A score of ≥5 triggers TB treatment initiation. Isoniazid Preventive Therapy (IPT) is prescribed only to children who are household contacts of a smear positive pulmonary case of TB and in whom the screen for active disease is negative. Neither mycobacterial culture nor tuberculin skin testing were available on site, during the time of this clinical cohort.
The primary end point for this analysis was a recorded diagnosis of tuberculosis (pulmonary or extrapulmonary) or the initiation of antituberculous drugs for a reason other than IPT.
Both sociodemographic and clinical characteristics were considered as independent variables and were ascertained identically for all patients. Included in the analysis were age, gender, the child’s orphan status at enrollment (having lost one or both parents vs. none), weight-for-height at enrollment (Z score calculated with EpiInfo, with severely low weight-for-height defined as a Z score of ≤319), whether the child had ever attended school at the time of enrollment (school in Kenya includes nursery school for very young children and kindergarten), CD4 cell percentage of total lymphocyte count (CD4%) closest to enrollment, the use of IPT, and whether the child was attending an urban or a rural clinic. Combination antiretroviral treatment (cART) was analyzed in descriptive bivariable analyses as to whether children had ever been prescribed cART. However, crucial to this analysis was the establishment of a temporal relationship between the use of cART and the incidence of TB diagnosis, ie, the use of cART precede an incident diagnosis. Therefore, we created a binary variable, where the cART data is coded as 1 if children began using cART prior to event or censoring in the analysis; otherwise the data were coded as 0.
The threshold for severe immune suppression depends on the age of the child. A binary age-specific variable for severe immune suppression was created as per country-adopted WHO recommendations.18 Severe immune suppression is defined for children aged ≤18 months as a CD4% <25%; CD4% <20% if the child is aged >18 months but <5 years; and CD4% ≤15% if the child aged ≥5 years.
Completeness of data were examined for each variable and compared for differential proportions of missing data between the incident TB and no-TB groups. We explored whether differential missing data would bias the outcome by examining Kaplan-Meier plots, and by conducting sensitivity analyses excluding those with missing data.
Normally and non-normally distributed categorical and dichotomous variables were analyzed using the χ2 and Kruskal-Wallis tests, respectively. The medians and interquartile ranges of continuous variables were analyzed with the Wilcoxon Rank Sum test. Factors independently associated with prevalent tuberculosis at the time of enrollment were assessed using multivariable logistic regression.
The Kaplan-Meier method was used to estimate the incidence of and the time to TB diagnosis. Patients were censored at the time of the event, or at the time of their last clinic visit. Incidence was calculated from the date of clinic enrollment in the AMPATH program and is presented per 100 child-years (CY) of follow-up. We conducted a subanalysis by timing of diagnosis postenrollment, and calculated incidence rates between months 0 and 2.0; 2.1 to 6.0, and 6.1 months or more post enrollment to identify the time of highest risk. Survival curves were compared using the Wilcoxon Log Rank test. Cox Proportional Hazards were used to calculate unadjusted hazard ratios, adjusted hazard ratios (AHR) and 95% confidence intervals (CI) in predictive models. We examined assumptions about proportionality of the hazards ratios by graphically displaying the log normal of time plotted against the log normal of the survival probability. Variables were entered into the final model, if they were statistically significant at an alpha of 0.05 or if they were believed to be potential confounders.
We assessed, in subanalyses, whether there was any bias arising in our hazard ratios related to the timing of the diagnosis or to the number of clinic visits a child had, by restricting these analyses to (a) children who were diagnosed at least 2 months after enrollment, and (b) children who had more than 2 visits and rerunning the Cox Proportional Hazards models.
We did determine that age and orphan status, as well as age and having ever attended school, were colinear (as determined through calculation of R-squared and use of Cuzick test for trend20). Because age is already a well-established risk factor for TB disease, the variables of orphan status and school attendance were a priori selected for inclusion in the final multivariable model, and age was dropped. To confirm that age was not acting as a confounding variable in the analysis, we conducted a post hoc, matched, case-control analysis (matched for age categories 0–2, 3–4, 5–9, 10+ years).
All P values were 2-sided. All analyses were done with STATA Version 9 (College Station, TX).
There were 6,535 HIV-infected children aged 0–13 years eligible for analysis, 50.1% of whom were female, with a median (interquartile range) age at enrollment of 1.0 year (0.2–4–6). Of these, 234 (3.6%) were diagnosed with TB at enrollment. In bivariable analysis, those diagnosed with TB at enrollment were more likely than those not diagnosed with TB to: be orphaned (46.4% vs. 22.5%, P < 0.001), have ever attended school (55.7% vs. 24.4%, P < 0.001), have an HIV-infected sibling (5.3% vs. 2.9%, P = 0.008), be severely low weight-for-height (33.0% vs. 18.5%, P < 0.001), be older (median 5.5 vs. 1.0 years, P < 0.001), and have a lower median CD4% at or near enrollment in all age categories although this was not always statistically significant due to the low number of subjects: ≤18 months, 17.5 vs. 20.5%, P = 0.203 (n prevalent TB cases per age = 10); greater than 18 months and less than 5 years, 15% versus 19%, P = 0.042 (n cases = 42); and greater than 5 years, 10 versus 17%, P < 0.001 (n cases = 92). There was no difference in proportion of males (53% vs. 49%, P = 0.181). In multivariable analysis, factors independently associated with diagnosis of TB at enrollment were, having ever been in school (adjusted odds ratio, AOR: 3.27, 95% CI: 2.21–4.84), being severely immune suppressed (AOR: 2.06, 95% CI: 1.44–2.95), being an orphan (AOR: 1.70, 95% CI: 1.17–2.46), and being of severely low weight-for-height (AOR: 1.61, 95% CI: 1.34–1.94).
New TB Diagnoses
There were 797 new TB diagnoses postenrollment, of whom 765 had time to event information. There were 4368.0 CY of follow-up, with a corresponding incidence rate of 17.5 (16.3–18.8)/100 CY. The incidence rate was much higher in months 1 to 2 post enrollment at 355.0 (321.1–392.3) per 100 CY, compared with 46.0 (40.1–52.8) per 100 CY, during months 3 to 6, and 4.7 (4.0–5.4), after month 6 (Table 2).
Factors Associated With an Incident Diagnosis of TB
Clinical and sociodemographic characteristics comparing those with incident diagnosis versus those without are summarized in Table 3. Completeness of data was similar in both groups for each variable of interest with the exception of CD4%, where there was a greater proportion missing in the nonevent group.
Those with an incident diagnosis were not different in terms of gender (51% male vs. 49%, P = 0.205), but were more likely to be older (median 4.5 years vs. 0.81, P < 0.001), to be orphans (38% vs. 18%, P < 0.001), to have severely low weight for height at enrollment (28% vs. 17%, P < 0.001), to be attending an urban versus rural clinic (59% vs.54%, P = 0.007), to have ever attended school (46% vs. 20%, P < 0.001), and to have a lower median CD4% in each age category. Children with an incident TB diagnosis were proportionately less likely to have ever received cART (16% vs. 19%, P = 0.055). There were 276 children who were initiated on IPT (11% in the incident diagnosis group vs. 3% in the non-TB group, P < 0.001). However, there were only 11 cases where the IPT preceded the TB diagnosis (ie, the majority of children received IPT for a new TB exposure following a previous episode of TB treatment); therefore IPT was not considered further.
Time to Event and Multivariable Models
Figure 1 graphically demonstrates that the incidence rate of TB diagnosis per 100 CY was higher for orphans versus nonorphans (28.1 vs. 14.2); for those who ever attended school versus nonattenders (31.5 vs. 13.1); and for those of severely low weight-for-height at enrollment versus those of moderately low to normal weight-for-height (41.8 vs. 14.8). The diagnosis incidence rate was much lower for those on cART versus those not receiving cART (7.2 vs. 22.2), with a corresponding incidence rate ratio (95% confidence interval) for cART use of 0.32 (0.26–0.40).
As shown in Table 4, multivariable Cox regression analysis suggests that being severely immune suppressed at enrollment (AHR: 4.44, 95% CI: 3.62–5.44), having ever attended school (AHR: 2.64, 95% CI: 2.15–3.25), being an orphan (AHR: 1.57, 95% CI: 1.28–1.92), being severely low weight-for-height (AHR: 1.46, 95% CI: 1.32–1.62), and attending an urban clinic (AHR: 1.39, 95% CI: 1.16–1.67) are all independent risk factors for incident TB diagnosis among these HIV-infected children. Receiving cART was associated with a profound reduction in the risk of being diagnosed with TB disease after adjustment for other factors (AHR: 0.15, 95% CI: 0.12–0.20).
To confirm that age was not acting as a confounding variable in the analysis, we conducted a post hoc, matched, case-control analysis (matched by age group). This reduced the sample size to 4064 (from 6301) but all effect estimates remained stable. Specifically, the AORs (95% CIs) were: use of cART AOR: 0.17 (0.12–0.23), being severely immune suppressed AOR: 4.22 (3.31–5.37), having ever attended school AOR: 1.68 (1.18–2.39), being an orphan AOR: 1.56 (1.22–1.99), attending an urban clinic AOR: 1.47 (1.20–1.82), and extremely low weight for height AOR: 1.37 (1.21–1.55).
The first aim of this study was to describe the frequency with which tuberculosis is diagnosed in HIV-infected children cared for by the USAID-AMPATH Partnership. Our finding that 4% of children presented at their initial clinical encounter with symptoms consistent with active TB, but that nearly 20% were subsequently treated for TB indicate an important clinical burden of presumed TB among HIV-infected children in East Africa. This finding is consistent with rates reported from other sub-Saharan African programs,21 including those from South Africa22 where the incidence of TB is considered the highest in the world.23
A second aim of this study was to identify risk and protective factors associated with a TB diagnosis. We have identified that the use of cART reduced the probability of a TB diagnosis by 85%, after adjustment for immune suppression, weight for height, and key social factors. The use of cART to reduce incident TB in adults has been well established in places such as South Africa and Brazil, where rates of HIV-TB coinfection are high.24,25 Our findings support other retrospective data from Cote d’Ivoire and South Africa that have demonstrated a lower burden of tuberculosis among HIV-infected children receiving cART, compared with those from the pre-HAART era.26,27
Whether cART is protective against development of active TB in HIV-infected children, or development of HIV-related conditions such as diffuse lymphadenopathy which may be misinterpreted as TB is unknown from our study. However, this distinction does not detract from the implications arising from such a profound reduction in the risk of a TB diagnosis: for the child, it means avoidance of the unnecessary use of anti-TB medications resulting in fewer drug-drug interactions, side effects, invasive laboratory monitoring, and time spent in the clinic. Less obvious but still critically important may be the benefits derived from early cART in reducing the nonspecific symptoms that lead to a diagnosis of TB in the absence of TB diagnostics with high specificity. These findings call for a more careful examination of the benefits of early cART initiation in children living in the developing world, where prevalent coinfections (TB, malaria, diarrhea illnesses) play a more significant role. Further research about the effectiveness of cART in preventing tuberculosis disease in HIV-infected children is needed.
The age of the children in this analysis who were diagnosed with TB was much greater than is typical of the natural history of TB in children. We believe this finding to be fitting because (1) these children are HIV-infected, and therefore experience a risk of developing active TB similar to that of infants28 and (2) the infants presenting to the USAID-AMPATH Partnership clinics for the first time are likely to be the infants of our HIV-infected female patients. As a result, they are more likely to be followed in the clinic early—before they become symptomatic. In contrast, the children who present at older ages may be more likely to be sick at presentation (their illness being the motivating factor behind seeking healthcare). Third, infants are generally infected with tuberculosis by members of their family; these mothers are already cared for in our HIV care program where screening for TB is routine, as is IPT, in all adults enrolled, and thus are less likely to be TB source cases for their children.
We have additionally identified several risk factors for a TB diagnosis among these HIV-infected children. Children who are orphaned are especially vulnerable to being diagnosed with TB, supporting other work by our group showing that orphan status is predictive of poorer clinical outcomes among HIV-infected children.14 Children who are severely underweight for their height at enrollment also have a greatly increased risk of being diagnosed with TB disease. This is partially attributable to severe immune suppression, but given the significance of weight for height after adjustment for immune suppression in the multivariable model, is likely also an indicator of malnutrition. Both school and urban clinic attendance were identified as risk factors for being diagnosed with TB in our study. Although both of these variables were only recorded at enrollment, it is known that over-crowding and/or lack of ventilation may promote TB transmission, both conditions being commonly observed in urban homes in the region as well as in schools. These data suggest that places of congregation, including schools, should be targeted as high priority settings for TB screening and prevention initiatives by the public health sector. Advanced immune suppression continues to play a leading, but additional and independent role as a risk factor for being diagnosed with TB, and children who are severely immune suppressed at their initial presentation must be considered at extremely high risk.5,29
The majority of TB diagnoses postenrollment in this analysis were made within the first 6 months following an initial encounter. There are several possible explanations for these findings, both clinical and logistical. First, there is a window period in which patients are developing TB disease yet currently available diagnostic evaluations are insensitive for detecting it.4 Thus, the original examination and chest radiograph taken at the time of enrollment may have missed early (subclinical) disease. Because children, and particularly those who are immune compromised, have only short periods of asymptomatic disease,4,28,30,31 this diagnostic limitation may contribute to the early high incidence. Second, TB disease in this study is defined as initiation of TB therapy. With the combination of severe malnutrition and immune suppression, many children present with early failure to thrive. Faced with this clinical scenario in our area of high TB burden with limited diagnostic confirmatory choices, clinicians often initiate antituberculosis treatment as an empirical trial. This may result in an overestimation of TB disease in those early months. Third, the reason why children presented for HIV care at this time in their life is not recorded. A family member may have recently tested positive for TB or HIV causing other members of the family to be tested. Kenyan TB treatment guidelines recommend HIV testing for all TB suspects. This event could make the child “recently exposed” and in the high risk period for development of disease.28,31 Recent exposure to TB in a child may explain early 2 month incidence as children do not contain the organism well after recent exposure and easily progress to early disease.28 Finally, logistical factors may also have contributed to the early high incident rates. All children receive a chest x-ray (CXR) as part of their initial evaluation. The taking and reading of the CXR are scheduled to occur on the date of first visit; however, it is possible that one or the other did not occur and thus the diagnosis was made and/or recorded (and treatment initiated) at a second visit. This logistical concern could contribute to the early 2 month incidence but would be unlikely to contribute beyond that time period. Importantly, although immune reconstitution syndrome (IRS) may have been responsible for those TB diagnoses that occurred among children who had already begun receiving cART, our data do not suggest that the high early incidence of TB diagnoses in general is related to IRS as our effect estimates indicate a lower risk of diagnosis associated with cART, not a higher risk.
Our study does also have limitations. Diagnostic tests for tuberculosis with high sensitivity and specificity for pediatric patients do not generally exist in resource-constrained environments. We are therefore dependent upon scoring systems which have not been validated in or out of the setting of HIV infection. The lack of adequate diagnostic tests does not, however, preclude children from receiving a diagnosis of TB or undergoing treatment with its resultant clinical implications (adverse side effects, drug interactions, delay of other diagnoses, delay of cART initiation). Although, the WHO, in its guidance document on the management of TB in children states that scoring criteria should not be relied upon to initiate anti-TB treatment, there remains little in the way of alternatives for clinicians working in resource constrained settings where diagnostic interventions might not be feasible or available (eg, lymph node biopsies, gastric aspiration, etc.). Furthermore, given the profound implications of not treating suspect TB in HIV infected children, international guidelines underscore that initiation of antituberculosis therapy is the priority in HIV infected childhood TB suspects.32 The intent of this report is neither to validate nor to advocate the use of scoring systems for the diagnosis of TB in HIV infected children. Rather this report serves to underscore the frequency with which HIV infected children are started on TB medications in a care program where scoring systems are the only diagnostic approach available.
Second, there are limitations inherent to using observational data. Observational data are dependent upon clinician judgment and accurate recording of the information. Therefore, missing data are an issue. In reporting our work, we have used various accepted statistical techniques to test our findings in instances where missing data may have biased our results; our findings remained robust to these tests. Also related to the limitations of observational data, ascertainment bias may have occurred if some of the early incident diagnoses were in fact prevalent cases missed at enrollment because of poor diagnostics or logistical reasons. However, as TB disease in children progresses rapidly and because of the limitations of existing diagnostic tools, these early diagnoses should be (and are) counted as incident ones.33 These cases only prompted medical intervention (ie, treatment) sometime after enrollment, further, supporting their being counted as incident diagnoses.
A final limitation of this analysis is that we were only able to assess the effects of orphan status and school attendance as binary variables recorded at enrollment, and thus we were not able to measure an exposure gradient effect for time spent in school or length of time, orphaned. Beyond looking at school and type of clinic, we were not able to assess the effects of other places of congregation, such as places of worship. These findings should therefore be interpreted with caution.
In summary, this study highlights the important clinical burden of presumed TB among HIV-infected children, and identifies possible, currently available, strategies for reducing their risk. We have found that the use of antiretroviral medication to prevent the development of TB may assist in reducing the burden of diagnosis in HIV-infected children, while simultaneously preventing the consequences of immune decline from HIV infection. Further studies are urgently needed to better elucidate the mechanisms of this reduction and the optimal timing of cART initiation in all age groups. TB programs should consider focused TB screening, prevention, and treatment campaigns centered in places where children are potentially exposed to tuberculosis such as schools, churches, and urban areas of overcrowding. It cannot be forgotten that pediatric TB is a direct result of uncontrolled adult TB; improved TB control programs that aggressively pursue active case finding and contract tracing can markedly lead to reductions in pediatric TB cases. The expanded use of IPT in pediatric populations may also be effectively used in resource-constrained settings to prevent the development of active TB.22
Tuberculosis is a preventable and treatable disease, catastrophically promoted by HIV infection. Recent recognition of tuberculosis drug resistance, whether multidrug resistant or extremely drug resistant, have highlighted program shortcomings that both promote transmission and delay care, such as the lack of readily available culture capability in the developing world. Improved diagnostics, including both laboratory development for standard cultures and novel diagnostic point of care testing, are essential in reversing the epidemic trends of TB and in decreasing the morbidity and mortality from TB in vulnerable populations including children. Mycobacteriology laboratories presently have the capability of supplying a culture diagnosis in up to 60% of pulmonary cases of TB; children throughout the world should have access. These program capacity issues as well as appropriately designed clinical trials that address the diagnostic and treatment issues related to tuberculosis in children, and particularly those who are HIV-infected, are past due.
The authors thank all the clinicians in all the AMPATH clinics for their dedication in caring for patients, and their attentiveness in accurately recording their patients’ data; and also the data entry technicians, data managers, administrative and clerical staff, for enabling the collection, management, interpretation, and publication of these data. AMPATH and the authors also thank the Rockefeller Foundation for funding the development of the AMPATH Medical Records System, and the Kenyan Department of Leprosy, TB and Lung Disease (formerly the Kenyan National Leprosy and Tuberculosis Program) for their support.
1. World Health Organization. Global Tuberculosis Control—Surveillance, Planning, Financing, WHO Report
2007. Geneva, Switzerland: World Health Organization; 2007.
2. Tindyebwa D, Kayita J, Musoke P, et al; African Network for the Care of Children Affected by AIDS (ANECCA), eds. Handbook on Paediatric AIDS in Africa
. Kampala, Uganda: Family Health International; 2005;124–132.
3. Raviglione MC, Snider DE Jr, Kochi A. Global epidemiology of tuberculosis. Morbidity and mortality of a worldwide epidemic. JAMA
4. Marais BJ, Graham SM, Cotton MF, et al. Diagnostic and management challenges for childhood tuberculosis in the era of HIV. J Infect Dis
. 2007;196(suppl 1):S76–S85.
5. Rekha B, Swaminathan S. Childhood tuberculosis - global epidemiology and the impact of HIV. Paediatr Respir Rev
6. Marais BJ, Gie RP, Schaaf HS, et al. Childhood pulmonary tuberculosis: old wisdom and new challenges. Am J Respir Crit Care Med
7. Marais BJ, Hesseling AC, Gie RP, et al. The bacteriologic yield in children with intrathoracic tuberculosis. Clin Infect Dis
8. Frieden T, ed. Toman’s Tuberculosis: Case Detection, Treatment, and Monitoring - Questions and Answers
. Geneva: World Health Organization; 2002.
9. Hesseling AC, Schaaf HS, Gie RP, et al. A critical review of diagnostic approaches used in the diagnosis of childhood tuberculosis. Int J Tuberc Lung Dis
10. Lincoln E, Sewell E. Tuberculosis in children
. New York, NY: McGraw-Hill Book Company Inc; 1963.
11. Einterz RM, Kimaiyo S, Mengech HN, et al. Responding to the HIV pandemic: the power of an academic medical partnership. Acad Med
12. Mamlin BW, Biondich PG. AMPATH Medical Record System (AMRS): collaborating toward an EMR for developing countries. AMIA Annu Symp Proc
13. Siika AM, Rotich JK, Simiyu CJ, et al. An electronic medical record system for ambulatory care of HIV-infected patients in Kenya. Int J Med Inform
14. Nyandiko WM, Ayaya S, Nabakwe E, et al. Outcomes of HIV-infected orphaned and non-orphaned children on antiretroviral therapy in western Kenya. J Acquir Immune Defic Syndr
15. Wools-Kaloustian K, Kimaiyo S, Diero L, et al. Viability and effectiveness of large-scale HIV treatment initiatives in sub-Saharan Africa
: experience from western Kenya. AIDS
16. Nair P, Philip E. A scoring system for the diagnosis of tuberculosis in children. Indian Pediatr
17. Stegen G, Jones K, Kaplan P. Criteria for Guidance in the Diagnosis of Tuberculosis. Pediatrics
18. World Health Organization. WHO Case Definitions of HIV for Surveillance and Revised Clinical Staging and Immunological Classification of HIV-related Disease in Adults and Children
. Geneva, Switzerland: World Health Organization; 2007.
19. World Health Organization. Management of Severe Malnutrition: A Manual for Physicians and Other Senior Health Workers
. Geneva, Switzerland: World Health Organization; 1999.
20. Armitage P, Berry G, Matthews JN. Statistical Methods in Medical Research
. Oxford, UK: Blackwell Sciences Ltd; 2002.
21. Elenga N, Kouakoussui KA, Bonard D, et al. Diagnosed tuberculosis during the follow-up of a cohort of human immunodeficiency virus-infected children in Abidjan, Cote d’Ivoire: ANRS 1278 study. Pediatr Infect Dis J
22. Zar HJ, Cotton MF, Strauss S, et al. Effect of isoniazid prophylaxis on mortality and incidence of tuberculosis in children with HIV: randomised controlled trial. BMJ
23. Dye C. Global epidemiology of tuberculosis. Lancet
24. Badri M, Wilson D, Wood R. Effect of highly active antiretroviral therapy on incidence of tuberculosis in South Africa
: a cohort study. Lancet
25. Santoro-Lopes G, de Pinho AM, Harrison LH, et al. Reduced risk of tuberculosis among Brazilian patients with advanced human immunodeficiency virus infection treated with highly active antiretroviral therapy. Clin Infect Dis
26. Kouakoussui A, Fassinou P, Anaky MF, et al. Respiratory manifestations in HIV-infected children pre- and post-HAART in Abidjan, the Ivory Coast. Paediatr Respir Rev
27. Walters E, Cotton MF, Rabie H, et al. Clinical presentation and outcome of tuberculosis in human immunodeficiency virus infected children on anti-retroviral therapy. BMC Pediatr
28. Marais BJ, Gie RP, Schaaf HS, et al. The natural history of childhood intra-thoracic tuberculosis: a critical review of literature from the pre-chemotherapy era. Int J Tuberc Lung Dis
29. Chintu C. Tuberculosis and human immunodeficiency virus co-infection in children: management challenges. Paediatr Respir Rev
30. Chintu C, Mwaba P. Tuberculosis in children with human immunodeficiency virus infection. Int J Tuberc Lung Dis
31. Marais BJ, Gie RP, Schaaf HS, et al. The clinical epidemiology of childhood pulmonary tuberculosis: a critical review of literature from the pre-chemotherapy era. Int J Tuberc Lung Dis
32. World Health Organization. Guidance for national tuberculosis programmes on the management of tuberculosis in children. Geneva, Switzerland: World Health Organization; 2006.
33. Rothman KJ, Greenland S. Modern Epidemiology
. 2nd ed. Philadelphia, PA: Lippincott Williams & Wilkins; 1998.