Secondary Logo

Journal Logo


Schistosoma mansoni Infection Is Associated With a Higher Probability of Tuberculosis Disease in HIV-Infected Adults in Kenya

McLaughlin, Taryn A. PhDa; Nizam, Azhar MSb; Hayara, Felix Odhiambo BSc; Ouma, Gregory Sadat BSc; Campbell, Angela MAd; Khayumbi, Jeremiah MSc; Ongalo, Joshua BSc; Ouma, Samuel Gurrion MBChBc; Shah, N. Sarita MDd,e; Altman, John D. PhDa,f; Kaushal, Deepak PhDg; Rengarajan, Jyothi PhDa,h; Ernst, Joel D. MDi; Blumberg, Henry M. MDh; Waller, Lance A. PhDb; Gandhi, Neel R. MDd,e,h; Day, Cheryl L. PhDa,f; Benkeser, David PhDb

Author Information
JAIDS Journal of Acquired Immune Deficiency Syndromes: February 1, 2021 - Volume 86 - Issue 2 - p 157-163
doi: 10.1097/QAI.0000000000002536



Mycobacterium tuberculosis (Mtb) is a serious public health concern, infecting a quarter of the world, leading to 10 million cases of tuberculosis (TB) disease and more than 1 million deaths annually.1 Mtb infection results in a spectrum of clinical states ranging from clearance, to latent TB infection (LTBI), to subclinical and active TB disease.2 The mechanisms that underlie these disparate outcomes are not well understood; however, coinfections, such as with HIV, and comorbidities, such as diabetes, are known risk factors.1 Helminths are coendemic with Mtb and known to be immunomodulatory.3,4 As such, determining the impact of helminths on Mtb infection outcomes is important and may identify potential risk factors for the development of TB disease.

Although the negative impact of helminths on immunity to Mtb infection, by either suppressing or skewing type 1 immunity, is well documented,5 the consequence of helminth infection on the development of TB disease is less clear. Some studies have reported a higher prevalence of intestinal helminth infections in individuals with active TB as compared with healthy controls (HCs).6–9 Others, however, have found no significant differences in the prevalence of helminth infections between individuals with TB and HCs.10 In addition, 2 longitudinal studies found no increase in TB incidence among individuals infected with intestinal helminths in either HIV-uninfected11 or HIV-infected individuals.12 This variation is particularly evident in a recent meta-analysis of epidemiological studies of individuals coinfected with Mtb and intestinal helminths, which found that the prevalence of coinfection as well as measured associations between helminth and Mtb infections varies greatly between studies, as well as between helminth species.13 These findings may indicate that associations for a single helminth species are not broadly applicable to all helminth species. Moreover, it is not clear that these results apply when considering helminths that do not reside in the intestine, such Schistosoma mansoni (SM).14

Infection with SM or other schistosome worms leads to the clinical disease schistosomiasis, which affects 240 million people globally.14 In sub-Saharan Africa, approximately one quarter of the population is estimated to be infected with SM.15 Schistosome infections are estimated to cause 280,000 deaths in sub-Saharan Africa per year.16 Despite this high disease burden, few studies have measured the association between SM and TB disease. Some evidence suggests that SM-infected individuals are at a higher risk of developing TB disease and that SM infection may impact the clinical manifestations of TB disease.9,12 However, others have reported decreased odds of TB disease in SM-infected individuals.6 Importantly, there has not been a comprehensive analysis of SM infection on Mtb infection outcomes that includes both LTBI and active TB.

There are also limited data on the impact of SM and HIV coinfection on Mtb infection outcomes. This is of particular interest, not only because HIV is the greatest risk factor for TB disease17 but also due to the bidirectional effects of SM and HIV. Schistosomiasis is suspected to play a role not only in susceptibility to HIV but also in higher viral load set point and worse HIV disease progression.18 HIV in turn has been shown to reduce egg shedding from SM adult worms, which makes detecting infection more difficult and can result in increased pathogenesis because of higher internal egg burden.19 Thus, the combination of these 3 infections could have distinct clinical outcomes and treatment implications.

The inability to perform randomized controlled trials of SM infection makes it challenging to definitively evaluate the impact of SM on Mtb infection outcomes. Nevertheless, we used data from an observational cohort coupled with modern causal inference and machine learning methodology to estimate a covariate-adjusted association between SM infection and Mtb infection outcomes. This allowed us to estimate the distribution of Mtb infection outcomes in an observed population under counterfactual conditions, that is, if all individuals in the observed population were infected with SM compared to if no individuals were infected with SM. We further evaluated whether HIV infection modified this association. Our analysis suggests that SM infection is associated with a higher probability of active TB but not LTBI and that this is exacerbated by HIV.


Study Population

This study was part of a larger case–control study analyzing cellular immune responses to Mtb. Individuals aged ≥18 years residing in Kisumu County in western Kenya were recruited from 2 community-based health clinics located in Kisumu city and in Kombewa.20 Participants were screened and enrolled between 2014 and 2017 based on Mtb infection status and then evaluated for HIV and SM infection. Participants were excluded only if they were missing data in either the outcome or exposure variables. Of 1109 participants screened for this study, 941 had complete Mtb, HIV, and SM study records and were therefore included in the analysis (Fig. 1).

Study participant flow diagram. a20 TB participants had missing or negative Mtb culture results.

Laboratory Results

Mtb infection status was defined as active TB disease (TB), latent TB infection (LTBI), or Mtb-uninfected HC based on the following criteria. TB participants were identified as referrals from satellite public health clinics in Kisumu County who had an initial positive sputum smear or GeneXpert MTB/RIF result. On enrollment in the study, another sputum sample was collected and evaluated for Mtb by fluorescence microscopy, GeneXpert MTB/RIF, HAIN MTBDR, and MGIT culture. Individuals with active TB were evaluated within the first 7 days of initiating the standard 6-month course of TB treatment, which was provided according to the Kenyan national health guidelines. Healthy asymptomatic individuals with no previous history of TB disease or treatment were identified as community contacts from index TB cases. They were evaluated by QuantiFERON-TB Gold (QFT) assay using the cutoffs and designations specified by the manufacturer (Qiagen) and as previously described.21 LTBI were defined as those with a positive QFT result (TB Ag-Nil ≥0.35 IU IFNγ/mL). HCs were defined as those with a negative QFT result (TB Ag-Nil <0.35 IU IFNγ/mL). Chest x-rays were performed on individuals in the LTBI and HC groups.

Helminth infection was determined using standard Kato–Katz microscopy. Two thick Kato–Katz smears were prepared from stool samples collected on 2 separate days. Slides were analyzed by experienced laboratory technicians who recorded the presence of eggs belonging to SM, hookworm, Ascaris lumbricoides, and Trichuris trichiuria. SM-infected individuals were defined as having at least 1 egg present in at least 1 slide examined. The number of SM eggs were counted and recorded as eggs per gram (epg) for SM-infected individuals. All helminth-infected individuals were referred for treatment, according to the Kenya Ministry of Health guidelines.

Serologic testing for HIV antibodies was performed for all individuals using the rapid diagnostic kit for HIV (1 + 2) Antibody V2 (KHB Shanghai Kehua Bio-engineering Co., Ltd). Plasma HIV viral load and CD4 T-cell counts were measured for HIV-infected individuals. All HIV-infected individuals were referred for antiretroviral therapy (ART) according to the Kenya Ministry of Health guidelines.

Women of childbearing age were tested for pregnancy using test strips for detection of human chorionic gonadotropin in urine. Individuals were tested for malaria using a malaria rapid diagnostic test (SD BIOLINE Malaria Ag P.f/Pan test from Abbott) and referred for treatment when positive.

Ethics Statement

This study was conducted in accordance with the principles expressed in the Declaration of Helsinki. All participants provided written informed consent for the study, which was approved by the KEMRI Scientific and Ethics Review Unit, in addition to the Institutional Review Boards at Emory University and the US Centers for Disease Control and Prevention.

Statistical Analysis

Participants' demographic and clinical characteristics were summarized using descriptive statistics based on SM and Mtb infection status. Groups of individuals were compared based on their SM infection status using the Kruskal–Wallis test to evaluate differences for continuous measures, with pairwise comparisons adjusting for multiple testing using the Benjamini–Hochberg method in the case of 3 or more groups. Comparisons of categorical data were made using either a χ2 or Fisher exact test.

Our primary analysis examined differences in Mtb infection outcomes across the entire study population and separately among HIV-infected and HIV-uninfected individuals. To control for potential confounders of SM and Mtb outcomes, we adjudicated a prespecified list of potential confounders (Table 1). We used these data to estimate the covariate-adjusted probability of outcomes under each SM infection state using targeted minimum loss estimation (TMLE). TMLE is a doubly robust method that requires estimates of the probability of SM infection as a function of covariates (a propensity score) and the probability of Mtb infection outcome as a function of SM infection status and covariates.22 These estimates were both obtained using super learning. Super learning entails prespecifying multiple candidate regression models and uses cross-validation to build a combination of the various models that provide the best estimated fit to the observed data.23 Prespecified regressions and covariates are indicated in Figure 1, Supplemental Digital Content, Wald tests with a significance level of 0.05 were used to test the null hypothesis of no difference in proportion of Mtb outcomes between SM groups. Standard error estimates were obtained using estimated influence functions and the delta method.

TABLE 1. - Characteristics of Study Population
SM n = 764 SM+ n = 177
Age: median (IQR)* 30 (23–40) 30 (23–36)
Sex: n (%)*
 Female 472 (61.8%) 95 (53.7%)
 Male 292 (38.2%) 82 (46.3%)
Race: n (%)
 Black African 764 (100%) 177 (100%)
Recruitment site: n (%)*
 Kisumu city (urban) 426 (59%) 103 (60%)
 Kombewa (rural) 296 (41%) 68 (40%)
Hemoglobin g/dL: median (IQR)* 12.7 (11.4–13.9) 13·0 (11.7–14.1)
Pregnant: n (%)* 1 (0.23%) 0 (0%)
Malaria: n (%)* 2 (0.37%) 0 (0%)
Any Helminth (not including SM): n (%)* 77 (10.1%) 19 (10.7%)
 Ascarisa 13 (1.7%) 1 (0.6%)
 Trichurisa 16 (2.1%) 5 (2.8%)
 Hookworm 52 (6.8%) 14 (7.9%)
Helminth species (not including SM): n (%)*
 0 687 (89.9%) 158 (89.3%)
 1 73 (9.6%) 18 (10.2%)
 2 4 (0.52%) 1 (0.56%)
SM features
 epg: median (IQR) NA 48 (24–144)
 Light intensity (1–99 epg): n (%) NA 122 (68.9%)
 Moderate intensity (100–399 epg): n (%) NA 41 (23.2%)
 Heavy intensity (400+ epg): n (%) NA 14 (7.9%)
HIV features
 Positive: n (%)* 301 (39.4%) 67 (37.9%)
Among positive
 Viral load: median (IQR)* 30,058 (1775–150,826) 18,879 (2442–101,875)
 Suppressed viral load: n (%) 25 (8.3%) 5 (7.5%)
 CD4 count (cells/μL): median (IQR) 395 (184–600) 467 (299–646)
On ART: n (%)
 Yes 50 (16.6%) 6 (9.0%)
 No 203 (67.5%) 55 (82.0%)
 Unknown 48 (15.9%) 6 (9.0%)
Tuberculosis features
 QFT IFNγ IU/mL (TB Antigen-nil): median (IQR)
  HC 0.00 (0.00–0.05) 0.00 (0.00–0.04)
  LTBI 7.00 (2.12–9.57) 7.07 (1.76–9.53)
 Abnormal chest x-ray: n (%)
  HC 3 (1.3%) 0 (0%)
  LTBI 8 (2.5%) 2 (2.7%)
*Included in propensity score and outcome regression models.
IQR, interquartile range.

Our secondary analysis evaluated whether HIV modified the effect of SM infection on Mtb infection outcomes. We tested the null hypothesis that the difference in the probability of each Mtb infection outcome between SM-uninfected and SM-infected groups was equivalent between HIV-uninfected and HIV-infected groups. We then performed a sensitivity analysis to evaluate the impact of adjusting for CD4 T-cell counts, measured among HIV-infected individuals. We also performed a sensitivity analysis including study participants with indeterminate QFT results. All primary and secondary statistical analyses were performed with R software, using the SuperLearner and drtmle packages.23,24


Of the 941 individuals included in the analysis, 177 (18.81%) tested positive for SM (Table 1). The median egg burden for SM-infected individuals was 48 epg, with 122 individuals categorized as having a light-intensity infection (1–99 epg), 41 a moderate-intensity infection (100–399 epg), and 14 a heavy-intensity infection (>399 epg) according to the WHO standards.14

Participants in each SM group were similar with regard to demographic and clinical parameters (Table 1). Few participants had additional parasitic infections, most of which were hookworm infections. Furthermore, the frequency of both malaria and additional helminth infections was equivalent between SM-uninfected and SM-infected groups. The proportion of HIV-infected individuals in each group was equivalent and had similar clinical features with regard to the viral load and CD4 count between SM-uninfected and SM-infected individuals. There were no differences in the median quantitative QFT response or the percentage of abnormal chest x-rays between SM-uninfected and SM-infected individuals.

Participants in Mtb infection groups were different with regard to both clinical and demographic measures (Table 2). HC individuals were younger than individuals in both the LTBI and TB group. There were more men and more individuals from Kisumu city in the TB group compared with the LTBI and HC groups. In addition, TB participants had lower hemoglobin levels compared with LTBI and HC individuals. The LTBI group had the lowest frequency of HIV-infected individuals. This group also had the lowest median viral load and the highest median CD4 T-cell counts. Importantly, very few HIV-infected HC and LTBI individuals were receiving ART treatment compared with 35.2% of TB individuals who were receiving ART at the time of screening. The prevalence of additional helminth infections was higher in the LTBI and HC than in the TB groups.

TABLE 2. - Study Characteristics by Mtb Infection Status
HC n = 315 LTBI n = 432 TB n = 194 P
Age (yr): median (IQR)* 25.0 (20.0–34.0) 31.0 (24.0–41.2) 33.0 (26.2–40.0) <0.001
Sex: n (%)* <0.001
 Female 212 (67.3%) 276 (63.9%) 79 (40.7%)
 Male 103 (32.7%) 156 (36.1%) 115 (59.3%)
Race: n (%) 1
 Black African 315 (100%) 432 (100%) 194 (100%)
Recruitment site: n (%)* <0.001
 Kisumu city (urban) 150 (50%) 250 (59.2%) 129 (75.4%)
 Kombewa (rural) 150 (50%) 172 (40.8%) 42 (24.6%)
Clinical information
 Hemoglobin (g/dL): median (IQR)* 12.9 (11.8–13.9) 13.2 (11.9–14.2) 11.2 (9.80–12.3) <0.001
 Pregnant: n (%)* 0 (0%) 1 (0.38%) 0 (0%) 1
 Malaria: n (%)* 1 (0.47%) 1 (0.38%) 0 (0%) 1
Any Helminth (not including SM): n (%)* 39 (12.4%) 46 (10.6%) 11 (5.67%) 0.048
 Ascaris* 2 (0.63%) 11 (2.55%) 1 (0.52%) 0.060
 Trichuris* 11 (3.49%) 5 (1.16%) 5 (2.58%) 0.087
 Hookworm* 28 (8.89%) 33 (7.64%) 5 (2.58%) 0.020
No. of Helminth species (not including SM): n (%)* 0.131
 0 276 (87.6%) 386 (89.4%) 183 (94.3%)
 1 37 (11.7%) 43 (9.95%) 11 (5.67%)
 2 2 (0.63%) 3 (0.69%) 0 (0%)
HIV clinical features <0.001
 Positive: n (%)* 131 (41.6%) 132 (30.6%) 105 (54.1%) 0.022
Among positive
 Viral load (copies/mL): median (IQR)* 52,004 (6100–145,968) 14,856 (1810–54,166) 41,728 (355–376,951) 0.007
 Suppressed viral load: n (%)§ 6 (4.6%) 8 (6.1%) 16 (15.2%) <0.001
 CD4 count (cells/μL): median (IQR) 466 (292–646) 518 (356–720) 172 (78–386)
On ART: n (%) <0.001
 Yes 6 (4.6%) 10 (7.6%) 40 (38.1%)
 No 105 (80.2%) 104 (78.8%) 49 (46.7%)
 Unknown 20 (15.3%) 18 (13.6%) 16 (15.2%)
Tuberculosis clinical features
 QFT IFNγ IU/mL: median (IQR) 0.00 (0.00–0.04) 7.07 (2.09–9.57) ND <0.001
 Abnormal chest x-ray: n (%) 3 (1.05%) 10 (2.5%) ND <0.001
*Included in propensity score and outcome regression models.
Fisher exact test.
§<20 copies HIV-1 RNA/mL plasma.
χ2 test.
IQR, interquartile range; ND, not done.

To estimate a covariate-adjusted association between SM and Mtb infection outcomes, we used machine learning to build regressions for both the propensity score and the outcome regression. These models were built using the combined HIV-uninfected and HIV-infected data sets and prespecified regressions and covariates with the SuperLearner package in R (see Fig. 1, Supplemental Digital Content, We then used TMLE to evaluate the impact of SM infection on Mtb infection outcomes in HIV-uninfected and HIV-infected individuals. This allowed us to estimate and compare the distribution of Mtb infection outcomes—defined as the probability of being Mtb-uninfected (HC), LTBI, or TB—under the counterfactual conditions of all participants being SM-infected or all participants being SM-uninfected.

We found that the estimated distribution of Mtb infection outcomes between SM-infected and SM-uninfected groups was different in HIV-infected individuals compared with HIV-uninfected individuals (P = 0.0002, Fig. 2). The covariate-adjusted estimate of proportion of HIV-uninfected individuals with active TB was modestly higher in the SM-infected group as compared to the SM-uninfected group (SM-infected: 19.7%; SM-uninfected: 15.8%), although the estimated distribution of Mtb infection outcomes was not different between SM groups (P = 0.486, Fig. 2). By contrast, within the HIV-infected group, the estimated distribution of Mtb infection outcomes differed between SM-infected and SM-uninfected groups (P = 0.0018, Fig. 2). The difference was driven by higher estimates of active TB in the SM-infected group when compared with the SM-uninfected group (SM-infected: 41.2%; SM-uninfected: 27.3%). Interestingly, the estimated proportion of HIV-infected individuals with LTBI was similar between groups (SM-infected: 38.3%; SM-uninfected: 36.7%).

Estimated distribution of Mtb infection outcomes. Probabilities are derived from TMLE and are represented as stacked bar plots ± 95% confidence interval. Standard error estimates were obtained using estimated influence functions and the delta method. The significance level was calculated using Wald tests. The estimated distribution of Mtb infection outcomes differs between SM and SM+ HIV+ individuals. The difference in the estimated proportion of Mtb infection outcomes between SM+ and SM groups differs between HIV+ and HIV individuals.

Sensitivity analysis adjusting for CD4 count led to an attenuation of the association of SM on the probability of active TB in HIV-infected individuals. The estimated distribution of Mtb infection outcomes between SM-uninfected and SM-infected groups still differed between HIV-uninfected and HIV-infected individuals (P = 0.0051; see Fig. 2, Supplemental Digital Content, Sensitivity analysis including individuals with indeterminate QFT status did not affect the observed results (data not shown).


In this study we used a robust machine learning–based strategy to evaluate the association of SM infection on Mtb infection outcomes among HIV-infected and HIV-uninfected individuals in western Kenya. Our results suggest that coinfection with SM and HIV may perturb the distribution of Mtb infection outcomes and drive a higher proportion of active TB disease than might be expected through HIV infection alone.

Our data provide evidence that SM is associated with a higher probability of active TB disease, after controlling for common covariates and additional helminth infections. This is consistent with findings from previous observational studies that evaluated SM as well as gastrointestinal helminths, despite differences in study design and setting.9,12 The higher probability of active TB in SM-infected individuals is likely due to impaired or altered CD4 T-cell responses, which are necessary for the control of Mtb infection,25 as opposed to innate immune responses, because similar proportions of LTBI were observed between SM-uninfected and SM-infected groups. Although helminths in general have been shown to impair Mtb immunity by compromising TH1 responses,5 this has not been observed for SM in humans.20 As such, the means by which SM might contribute to active TB disease remain unclear. In addition, we cannot determine whether active TB disease in the setting of SM infection is due to primary Mtb infection or reactivation of LTBI. This is further confounded in the setting of HIV, which results in profound CD4 T-cell depletion, including memory CD4 T cells.26 Helminth infections, which induce TH2 responses, may lead to further dysregulation of TH1 immunity and therefore exacerbate HIV infection and perturb Mtb immunity.18,19 Alternatively, HIV has been shown to worsen SM pathology that may contribute to overall disease burden, resulting in susceptibility to the development of TB disease.18,19 Finally, HIV and SM may simply impair Mtb immunity through independent mechanisms that both contribute to increased disease burden. Further studies are needed to ascertain the mechanism by which coinfection with SM and HIV increases the probability of active TB disease.

The strong association of SM with active TB among HIV-infected individuals but not HIV-uninfected individuals suggests that co-infection with HIV and SM may be a greater risk factor than either infection alone. This is consistent with a study conducted in Uganda that reported an increased incidence of active TB among SM-infected HIV-infected individuals compared with SM-uninfected HIV-infected individuals.12 The data regarding the impact of SM infection and treatment on virological outcomes in the setting of HIV monoinfection have been conflicting.18,19,27,28 Testing and treatment of SM among HIV-infected individuals may, however, be useful as a strategy for preventing TB, particularly in high-burden settings.

The main strengths of this study are a large sample size and the use of a robust analytic method. Although this cohort had a large number of coinfected individuals, SM was likely underdiagnosed in the HIV-infected group. HIV is known to reduce egg burden in stool, which was the diagnostic criterion used in this study.29 Despite this, we still found that HIV modifies the association between SM and active TB. Our findings therefore support robust testing to evaluate SM infection among HIV-infected individuals in future studies. Although we found little evidence of confounding, we were underpowered to further stratify our analysis to test for interactions with measured characteristics, such as biological sex. In addition, there is a possibility of bias by unmeasured confounding, such as socioeconomic status, occupational risk, and previous SM infection and/or treatment. This study was also limited by the inability to determine the order or timing of infection, which would be crucial to derive causal conclusions about the direction of effects in this setting. Epidemiologically, people living in western Kenya often become infected with SM in early childhood with prevalence reaching a peak of around 60% in early adolescence.30 Although mass drug administration has been very successful in this region, reinfection does occur in both children and adults.31 This suggests that individuals likely experience repeated SM infections before HIV and Mtb infection. One study in Uganda that reported increased TB incidence in SM-infected individuals compared with SM-uninfected individuals supports the notion that previous SM infection can increase the risk of active TB disease.12 Future studies, however, would benefit from evaluating Mtb infection outcomes over time among SM-infected individuals with and without HIV infection. This would be of particular interest in the setting of anthelmintic treatment and ART.

The interpretation of our results is influenced by the definitions used for Mtb infection status, which are based on imperfect diagnostic tools. There is currently no gold standard for diagnosing LTBI.32 The test used in this study, QFT, is an IFNγ release assay (IGRA) based on an immune response to Mtb antigens and does not differentiate LTBI from subclinical or active TB.33 The diagnosis of active pulmonary TB relies on a combination of symptoms and microbiological tests, the sensitivity of which can range from 50% to 90%.25 The diagnosis of active TB is even more challenging in HIV-infected individuals who are more likely to have negative microbiological tests.34

In high-pathogen burden regions, such as western Kenya, individuals may be infected with multiple microbes at any given time. It is therefore important to understand the interplay between these infections not only to appropriately treat each individual but also to manage public health initiatives. We provide evidence from machine learning–based approaches that SM infection is associated with significant alteration of Mtb infection outcomes in HIV-infected individuals. Further research is needed to establish the underlying mechanisms by which this occurs and to compare SM-induced immune regulation in HIV-uninfected and HIV-infected individuals. If supported by prospective studies, these findings could have implications to the control of TB in SM-endemic areas of the world. Our findings suggest that treatment of SM with praziquantel should be considered for the clinical prevention of active TB in HIV-infected patients. Furthermore, regular praziquantel treatment of HIV-infected individuals who are at risk for recurring SM infection could be used as an additional low-cost component of TB control programs. Such consideration when designing treatment and prevention strategies could reduce the burden of TB disease at the population level.


The authors thank Dr. Kevin Cain for advice on the recruitment and enrollment of study participants with Mtb infection and TB disease in Kisumu, Kenya. The authors also thank Dr. W. Evan Secor for advice on testing and interpretation of helminth infections in study participants. The authors thank many additional members of the Kenya Medical Research Institute (KEMRI)/Centers for Disease Control and Prevention (CDC) team who helped with enrollment and evaluation of participants and the participants themselves.


1. Global Tuberculosis Report 2019. Geneva, Switzerland: World Health Organization; 2019.
2. Lin PL, Flynn JL. The end of the binary era: revisiting the spectrum of tuberculosis. J Immunol. 2018;201:2541–2548.
3. Brooker S, Hotez PJ, Bundy DA. The global atlas of helminth infection: mapping the way forward in neglected tropical disease control. PLoS Negl Trop Dis. 2010;4:e779.
4. Maizels RM, McSorley HJ. Regulation of the host immune system by helminth parasites. J Allergy Clin Immunol. 2016;138:666–675.
5. Babu S, Nutman TB. Helminth-tuberculosis Co-infection: an immunologic perspective. Trends Immunol. 2016;37:597–607.
6. Elias D, Mengistu G, Akuffo H, et al. Are intestinal helminths risk factors for developing active tuberculosis? Trop Med Int Health TM IH. 2006;11:551–558.
7. Hailu AW, Solomon GS, Merid Y, et al. The case control studies of HIV and Intestinal parasitic infections rate in active pulmonary tuberculosis patients in Woldia General Hospital and Health Center in North Wollo, Amhara Region, Ethiopia. Int J Pharma Sci. 2015;5:1092–1099.
8. Tristao-Sa R, Ribeiro-Rodrigues R, Johnson LT, et al. Intestinal nematodes and pulmonary tuberculosis. Rev Soc Bras Med Trop. 2002;35:533–535.
9. Mhimbira F, Hella J, Said K, et al. Prevalence and clinical relevance of helminth co-infections among tuberculosis patients in urban Tanzania. PLoS Negl Trop Dis. 2017;11:e0005342.
10. Abate E, Belayneh M, Gelaw A, et al. The impact of asymptomatic helminth co-infection in patients with newly diagnosed tuberculosis in north-west Ethiopia. PLoS One. 2012;7:e42901.
11. Chatterjee S, Kolappan C, Subramani R, et al. Incidence of active pulmonary tuberculosis in patients with coincident filarial and/or intestinal helminth infections followed longitudinally in South India. PLoS One. 2014;9:e94603.
12. Brown M, Miiro G, Nkurunziza P, et al. Schistosoma mansoni, nematode infections, and progression to active tuberculosis among HIV-1-infected Ugandans. Am J Trop Med Hyg. 2006;74:819–825.
13. Taghipour A, Mosadegh M, Kheirollahzadeh F, et al. Are intestinal helminths playing a positive role in tuberculosis risk? A systematic review and meta-analysis. PLoS One. 2019;14:e0223722.
14. WHO. Schistosomiasis: Progress Report 2001-2011, Strategic Plan 2012-2020. 2019. Available at: Accessed June 14, 2019.
15. Hotez PJ, Kamath A. Neglected tropical diseases in sub-saharan Africa: review of their prevalence, distribution, and disease burden. PLoS Negl Trop Dis. 2009;3:e412.
16. van der Werf MJ, de Vlas SJ, Brooker S, et al. Quantification of clinical morbidity associated with schistosome infection in sub-Saharan Africa. Acta Trop. 2003;86:125–139.
17. Narasimhan P, Wood J, Macintyre CR, et al. Risk factors for tuberculosis. Pulm Med. 2013;2013:828939.
18. Secor WE. The effects of schistosomiasis on HIV/AIDS infection, progression and transmission. Curr Opin HIV AIDS. 2012;7:254–259.
19. Secor WE. Interactions between schistosomiasis and infection with HIV-1. Parasite Immunol. 2006;28:597–603.
20. McLaughlin TA, Khayumbi J, Ongalo J, et al. CD4 T cells in Mycobacterium tuberculosis and schistosoma mansoni Co-infected individuals maintain functional TH1 responses. Front Immunol. 2020;11:127.
21. Mazurek GH, Jereb J, Vernon A, et al. Updated guidelines for using interferon gamma release assays to detect Mycobacterium tuberculosis infection - United States, 2010. MMWR Recomm Rep. 2010;59:1–25.
22. van der Laan MJ, Rubin D. Targeted maximum likelihood learning. Int J Biostatistics. 2006;2:11.
23. van der Laan MJ, Polley EC, Hubbard AE. Super learner. Stat Appl Genet Mol Biol. 2007;6:Article25.
24. Benkeser D, Carone M, Laan MJV, et al. Doubly robust nonparametric inference on the average treatment effect. Biometrika. 2017;104:863–880.
25. Pai M, Behr MA, Dowdy D, et al. Tuberculosis. Nat Rev Dis Primers. 2016;2:16076.
26. Okoye AA, Picker LJ. CD4(+) T-cell depletion in HIV infection: mechanisms of immunological failure. Immunol Rev. 2013;254:54–64.
27. Colombe S, Machemba R, Mtenga B, et al. Impact of schistosome infection on long-term HIV/AIDS outcomes. PLoS Negl Trop Dis. 2018;12:e0006613.
28. Stete K, Glass TR, van Dam GJ, et al. Effect of schistosomiasis on the outcome of patients infected with HIV-1 starting antiretroviral therapy in rural Tanzania. PLoS Negl Trop Dis. 2018;12:e0006844.
29. Karanja DM, Colley DG, Nahlen BL, et al. Studies on schistosomiasis in western Kenya: I. Evidence for immune-facilitated excretion of schistosome eggs from patients with Schistosoma mansoni and human immunodeficiency virus coinfections. Am J Trop Med Hyg. 1997;56:515–521.
30. Sakari SSW, Mbugua AK, Mkoji GM. Prevalence of soil-transmitted helminthiases and schistosomiasis in preschool age children in mwea division, kirinyaga south district, kirinyaga county, and their potential effect on physical growth. J Trop Med. 2017;2017:1013802.
31. Mwandawiro C, Okoyo C, Kihara J, et al. Results of a national school-based deworming programme on soil-transmitted helminths infections and schistosomiasis in Kenya: 2012-2017. Parasit Vectors. 2019;12:76.
32. Salgame P, Geadas C, Collins L, et al. Latent tuberculosis infection--Revisiting and revising concepts. Tuberculosis (Edinb). 2015;95:373–384.
33. Pai M, Riley LW, Colford JM Jr. Interferon-gamma assays in the immunodiagnosis of tuberculosis: a systematic review. Lancet Infect Dis. 2004;4:761–776.
34. Getahun H, Harrington M, O'Brien R, et al. Diagnosis of smear-negative pulmonary tuberculosis in people with HIV infection or AIDS in resource-constrained settings: informing urgent policy changes. Lancet. 2007;369:2042–2049.

HIV; tuberculosis; schistosomiasis; machine learning; causal inference; Kenya

Supplemental Digital Content

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.