Epidemiology and aetiology of moderate to severe diarrhoea in hospitalised HIV-infected patients ≥5 years old in South Africa, 2018-2021: a case-control analysis

Introduction

Diarrhoeal diseases pose a significant health burden across the globe, causing over 1.6 million deaths annually amongst all ages¹. The majority of these deaths occur in low- and middle-income countries (LMICs), specifically in south Asia and Sub-Saharan Africa (SSA)¹. Despite a decrease in global diarrhoeal mortality rates over the past two decades, decreases have not been uniform across age groups or settings². Unsafe water, poor sanitation and hygiene (WaSH)¹, as well as malnutrition and compromised immunity³ are the main risk factors associated with diarrhoeal morbidity and mortality. Other risk factors identified in LMICs include residence in a slum, use of communal toilets and exposure to animals⁴. Children under the age of 5 years are at the highest risk, however there is a significant burden in older age groups, specifically in LMICs^1,5. Despite global improvements in WaSH⁶, much of the remaining burden of diarrhoeal disease may be attributed to foodborne pathogens⁷. Food safety is of specific concern in LMICs where there is increased consumption of unsafe foods, use of effluent in agriculture and changes in food distribution networks with bulk production and increased distance between production and consumption as well as large, often poorly regulated, informal sectors⁷.

Studies from high-income countries (HICs) have shown the majority of diarrhoeal diseases in adults seeking healthcare to be due to Campylobacter spp., Salmonella spp., norovirus and rotavirus^8,9.

Aetiology differs in SSA, with a meta-analysis amongst all ages identifying the main causative pathogens as Escherichia coli, Cryptosporidium, Cyclospora, Entamoeba and Shigella spp.¹⁰ There are several reasons for these differences in aetiology including different exposures and risk factors associated with poorer living conditions and poor underlying health. One important consideration is the high prevalence of human immunodeficiency virus (HIV) in SSA¹¹. HIV is a known risk factor for diarrhoea, with episodes in HIV-infected individuals more likely to be severe, prolonged and result in hospitalisation with higher mortality rates than in HIV-uninfected individuals^12,13. The aetiology of diarrhoea in HIV-infected individuals is known to shift with HIV disease progression and treatment¹⁴.

During early disease stages, while CD4+ counts are high and viral loads low, diarrhoea is often related to HIV seroconversion. As disease progresses, diarrhoea is frequently caused by opportunistic pathogens related to decreased immune function¹⁴. These pathogens include bacterial infections, such as Mycobacteria, parasitic infections, such as Cystoisospora belli, Cyclospora, Strongyloides, Cryptosporidia, Microsporidia and viral infections, such as cytomegalovirus (CMV)^14–16. Patients with low CD4+ cell counts are at increased risk for chronic diarrhoea⁴ and polyparasitic infection¹⁷. The HIV infection itself may be responsible for a proportion of the pathogen-negative cases, through HIV enteropathy and permanent damage to the gastrointestinal mucosa¹⁸, as well as changes in gut microbial populations leading to dysbiosis¹⁹. Mucosal damage reduces small bowel villous surface area, causing diarrhoea through malabsorption²⁰ as well as defects to cellular and humoral defence mechanisms in the gastrointestinal tract²¹. Pathogen-negative cases associated with advanced HIV disease may also be related to gastrointestinal tract involvement by Mycobacterium tuberculosis and/or avium²² which is difficult to diagnose. A Dutch study found that 34% of diarrhoeal cases in untreated HIV-infected individuals could not be explained by a known causative agent, and hypothesized that these unexplained cases were due to the HIV-infection itself²³. Initiation of antiretroviral treatment (ART), and accompanying increase in CD4+ cell count, is associated with a shift in the aetiology of diarrhoea from opportunistic infectious causes to non-infectious causes²⁴. Data from both HICs and SSA show that although patients initiated on ART have significantly reduced diarrhoeal incidence compared with those not on treatment, the overall diarrhoeal incidence remains high^18,25. A proportion of the remaining episodes in patients with well-controlled HIV may be drug-related¹⁸.

There are limited aetiological data on HIV-related diarrhoeal diseases in SSA, with the majority of published studies focusing on children under the age of five years, with available data in older children and adults being limited by diagnostic technology¹⁴. A review of laboratory records in Botswana indicated that only 14% of stool specimens submitted for routine diagnostic testing for all ages had a pathogen detected, of which 8% were bacteria and 6% parasites²⁶. Shigella spp. And Salmonella spp. were the most commonly detected bacteria while Cystoisospora spp. and Cryptosporidium spp. were the most commonly detected parasites. Viral testing was not done. Although HIV status was not available for the participants in this analysis, Botswana is a high HIV prevalence setting with an estimated 23.9% of the population between 15-49 years old being HIV-infected during the study time period²⁶. Contrary to these findings, a longitudinal cohort study of Zambian adults with a 31% HIV seroprevalence (higher than the estimated population HIV seroprevalence of 22%) found that pathogens could be detected in 99% of stool samples collected (diarrhoeal and non-diarrhoeal patients combined), with Cryptosporidium spp., Cystoisospora spp. and Citrobacter spp. being significantly more common in HIV-infected than HIV-uninfected individuals⁵. They found that HIV-infected adults in the pre-ART era were 2.4 times as likely to suffer from diarrhoea than HIV-uninfected adults and that this increased risk lasted throughout the infection rather than being limited to those with low CD4+ cell counts⁵. Meta-analysis data indicates a high burden of Cryptosporidium, microsporidia and Cystoisospora in SSA specifically¹⁶. Another recent meta-analysis identified South Africa as having the highest global Cryptosporidium prevalence (57.0%, CI 95%: 24.4-84.5%), although they recognised that these estimates were based on limited data²⁷. A study in rural South Africa found that 60% of patients suffering from chronic diarrhoea were HIV-infected and that the majority of these infections were due to Campylobacter spp. (20%), Plesiomonas shigelloides (17%), Aeromonas spp. (13%), Shigella spp. (10%), Salmonella spp. (10%) and E. coli spp. (10%)²⁸. Testing did not include parasites or viruses and results were not stratified by ART status. Other studies in African children have identified HIV-infection as a significant risk factor for rotavirus infection^13,29. Very few studies in SSA in adults included screening for viruses, despite data from HICs showing that viruses (specifically norovirus and rotavirus) were responsible for as much as 44% of diarrhoea in adults presenting to emergency departments⁹. Patients with advanced HIV are at increased risk for diarrhoea due to CMV infection³⁰. As CMV is mainly diagnosed on endoscopic biopsy³¹, prevalence is likely underestimated.

South Africa currently has the highest HIV burden in the world, representing 19% of the global HIV-infected population³². It is estimated that 86% of HIV-infected adults in South Africa have been diagnosed, with 57% of these on ART (50.8% - 72.7% by province)³³. Women, elderly patients, those with inadequate access to WaSH facilities and those not yet initiated on ART, have the highest risk for developing diarrhoea amongst HIV-infected South Africans³⁴. The majority of data on HIV-related diarrhoea comes from HICs with low HIV prevalence¹⁴. Since diarrhoeal aetiology is likely to vary between settings³⁵, there is a need for more targeted epidemiological studies¹⁸, specifically in LMICs with high HIV prevalence. There are few studies in HIV-infected populations which investigate a comprehensive range of pathogens or infections with multiple pathogens¹⁹. These data are important for informing public health interventions and updating diagnostic and treatment guidelines, especially since treatment is often based on guidelines rather than individual patient diagnostics³⁶. This study aimed to determine the pathogenic causes for diarrhoeal admissions in HIV-infected patients compared to controls without diarrhoea at three South African hospitals between July 2018 and November 2021.

Methods

Results

Enrolment and patient characteristics

A total of 378 specimens for patients ≥5 years were collected during the study period, including 277 cases and 101 controls (Figure 1). Amongst cases, 164 (59.2%) were HIV-infected, 79 (28.5%) were HIV-uninfected and 34 (12.3%) had unknown HIV status. Amongst controls, 23 (22.8%) were HIV-infected, 50 (49.5%) were HIV-uninfected and 28 (27.7%) had unknown HIV status. Cases with unknown HIV status (n=34) were excluded (due to limited testing capacity), hence a total of 344 specimens were included in the analysis.

• Download figure
• Open in new tab

Figure 1:

Specimens included in the analysis

^aDue to limited TAC tests available, cases with unknown HIV status were excluded from the analysis. Controls with unknown HIV status were included to increase sample size for the overall case-control analysis, but were not included in the HIV specific analysis.

The median age of included patients was 34 years (IQR of 23-47), with cases being slightly older (median 36; IQR 26-51) than controls (median 31; IQR 17-38) (Table 1). The majority of patients were female (59.4%) and from the rural sites (67.4%), with proportionally more controls coming from the rural site than cases (82.2% versus 61.3%, p<0.001). Cases and controls were similar with respect to household crowding and dietary habits. Both cases and controls had good access to electric or gas cookers (86.1%), refrigerators (88.4%) and improved WaSH (94.8% access to improved water source and 96.4% access to private latrines/flush toilets). CD4+ cell counts were available for 99 (60.4%) of the 164 HIV-infected cases. Eight (8.1%) had counts >500 cells/μl, 21 (21.2%) were between 200-500 cells/μl and the majority (n=70, 70.7%) had <200 cells/μl. Information regarding ART was available for 156 (95.1%) of the HIV-infected cases, of which 130 (83.3%) were on treatment. Mean CD4+ cells counts were similar between those on ART (229.23 cells/μl) and those not on ART (169.7 cells/μl), (p=0.418). Information regarding cotrimoxazole treatment was available for 151 HIV-infected patients, of which 63 (41.7%) were on treatment.

View this table:

• View inline
• View popup

Table 1:

Comparison of patient characteristics for cases and controls

Discussion

Despite advances in HIV treatment, diarrhoea amongst HIV-infected patients remains a significant public health challenge¹⁹. Our study found that the majority (67.5%) of patients 5 years and older with known HIV-status admitted to the study sites with diarrhoea were HIV-infected, highlighting the importance of HIV-related illnesses among South African adults hospitalised with diarrhoea. The Joint United Nations Programme on HIV/AIDS (UNAIDS) 90-90-90 treatment target aims to ensure that 90% of the HIV-infected population are diagnosed, 90% of those diagnosed are initiated on ART and 90% of those on ART should achieve viral suppression⁴¹. A recent South African analysis estimated that 70.7% of diagnosed HIV-infected individuals have initiated ART, and 87.4% of these are thought to be virally suppressed⁴¹. Proportions reported in the current analysis reflect these numbers, as 83.3% of HIV-infected patients with known ART status were on treatment, however a large proportion of HIV-infected patients included in this analysis had CD4+ cell count <200 cells/μl (70/99, 70.7%) indicating that they are unlikely to be virally suppressed. Mpumalanga (home province to the majority of the patients included in this analysis) is the only South African province in which the third-90 indicator value (the percentage of those on ART that are virally suppressed) is below 85%⁴¹. It is also likely that the study design was skewed towards detection of poorly suppressed HIV cases as they are most likely to be hospitalised for diarrhoea. Diarrhoea in patients with viral suppression will likely be seen at a clinic level. Further studies at clinic level are recommended in order to include a broader spectrum of cases. The clinical presentation of diarrhoea in HIV-infected cases differs to that of HIV-uninfected cases. HIV-infected diarrhoeal cases were more likely to have symptoms for a longer duration before admission and to suffer from weight loss and nausea. HIV-uninfected cases were more likely to suffer from dysentery. This finding has not been reported in other studies, however it is likely related to aetiological differences between HIV-infected and HIV-uninfected diarrhoeal patients. Amongst HIV-infected cases, patients on cotrimoxazole were more likely to experience fatigue, weight loss and chills and were more likely to have parasites detected, specifically Cystoisospora spp., compared to those not on cotrimoxazole. This is an interesting finding since cotrimoxazole is the prescribed treatment for this pathogen. Since all patients with CD4+ <200 cells/μl should receive cotrimoxazole prophylaxis, it can be used as a surrogate marker for advanced HIV disease. However this finding may also point to changes in Cystoisospora sensitivity for cotrimoxazole or may indicate that these patients are carriers or take longer to clear the pathogen. The possibility of drug resistance should be further investigated in future studies.

The use of molecular methods and the expanded panel of pathogens included in testing resulted in a high pathogen detection rate, with pathogen detection being significantly higher in cases compared with controls. Although viruses were commonly detected (18.5% of all specimens tested) and were more likely to be detected in cases as compared with controls, when stratified by HIV status, viruses were not significantly associated with diarrhoea in this study. Although this lack of association may be partly due to asymptomatic carriage of viruses (specifically in HIV-uninfected patients), it is likely that the sample size was insufficient to determine significant differences between cases and controls when stratified by HIV status. Interestingly, norovirus GII was more prevalent in HIV-infected individuals than HIV-uninfected individuals and detection in HIV-infected individuals was limited to cases. Previous findings from South Africa have shown that HIV-infected children are more likely to suffer poor outcomes related to norovirus infection⁴², however detection rates were similar amongst HIV-infected and –uninfected children. This increased detection of norovirus in HIV-infected patients ≥5 years of age has not been previously described. Bacteria was detected in 43.9% of specimens tested and was significantly associated with diarrhoea overall and amongst HIV-uninfected patients. Interestingly, prevalence of bacterial pathogens was high in both cases and controls within the HIV-infected group. This was specifically due to the high prevalence of E. coli spp. (including ETEC, EPEC and EAEC) amongst HIV-infected controls. High E. coli spp. colonisation rates in HIV-infected patients have been documented in other settings⁴³. We found Shigella spp, Salmonella spp. and H. pylori to be significantly associated with diarrhoea in the current study, specifically amongst HIV-uninfected patients. It’s likely that the sample size was insufficient to detect a difference in prevalence between cases and controls in the HIV-uninfected group. Campylobacter spp. and C. difficile were detected more often in HIV-infected patients as compared with HIV-uninfected patients but were not significantly associated with diarrhoea (likely due to small sample sizes).

As expected from literature¹⁴, parasite prevalence was significantly higher amongst HIV-infected compared with HIV-uninfected patients, specifically Cystoisospora, Cryptosporidium spp. and Enterocytozoon spp., although only Cystoisospora was significantly associated with diarrhoea amongst HIV-infected patients. The case-control analysis indicated that Blastocystis and Schistosoma are unlikely to be causative agents of diarrhoea since these pathogens were detected at higher rates in controls than in cases. Prevalence of Schistosoma among healthy individuals has been reported to be as high as 51.2% in the Democratic Republic of Congo and is likely to be higher in villages and rural areas where water is collected from rivers and dam⁴⁴. Blastocystis has been noted as an important cause of diarrhoea amongst immunosuppressed patients⁴⁵, however this was not found in the current study.

Diarrhoeal aetiology differed with HIV status, with Shigella spp. and Salmonella spp. being prevalent amongst HIV-uninfected cases and Shigella spp., Cystoisospora, and Cryptosporidium spp. being prevalent amongst HIV-infected cases. These differences should be considered during development of diagnostic and treatment guidelines. This study highlights the importance of Shigella spp. in diarrhoeal morbidity amongst adults (regardless of HIV status). According to the Global Burden of Disease Study (GBD), Shigella spp. were the leading cause of diarrhoeal deaths amongst individuals over the age of 5 years and the second leading cause of death in young children in 2016¹. While there is currently no approved Shigella vaccine, there are several candidates which show promise for efficacy testing⁴⁶. In order to introduce a candidate vaccine, burden estimates are required. Although a larger sample size is required, this analysis gives an indication of the high burden of Shigella spp. in our setting. Further work is required to determine if HIV-infected individuals should be considered as a risk group for possible target should Shigella vaccine become available.

Our results are in line with clinical literature which indicates a shift in diarrhoeal aetiology with CD4+ cell count amongst HIV-infected patients. Pathogen detection was high in those with CD4+ cell counts <200 cells/μl, indicating increased opportunistic infections (specifically Cryptosporidium spp.). Pathogens could only be detected in just over a third of cases with high CD4+ cell count (>500 cells/μl), suggesting that these cases are possibly drug-related. This finding may be useful for treatment guidelines. It also highlights the importance of addressing diarrhoea in patients on ART especially since chronic diarrhoea has been identified as an major cause of dropout from ART care services⁴⁷.

Detection of C. difficile in this study was almost exclusive to HIV-infected participants (13/14, 92.9%; including 12 cases and 1 control). Data from another hospital study in Gauteng Province estimated that community acquired C. difficile infections represent only 1.3% of cases (with the remaining 98.7% being hospital acquired)⁴⁸. Our study design excluded cases of hospital-acquired diarrhoea, as enrolment and specimen collection was done within 48 hours of admission. The 5.4% C. difficile prevalence reported here is lower than systematic review estimates of 15.8% amongst symptomatic non-immunosuppressed in-patients from LMIC⁴⁹. This systematic review indicated similar rates amongst immunosuppressed and non-immunosuppressed populations⁴⁹ whereas the current study found that C. difficile was more commonly detected in HIV-infected cases than in –uninfected cases. While not considered an opportunistic infection, C. difficile is associated with exposure to healthcare settings and with the use of antibiotics, both of which are increased amongst HIV-patients⁵⁰. The increased management of HIV patients in an outpatient setting⁵¹ may dispute the inclusion of recent admission as a criterion in the case definition. There is also growing evidence to suggest the possibility of C. difficile being an important community-acquired pathogen⁵¹, and consideration of this should be given in populations with high HIV prevalence.

A major limitation of this study was the small sample size which restricted the analysis (attributable fractions and odds ratios could not be calculated as many categories had zero count). The ANDEMIA study was not specifically aimed at enrolling HIV-infected cases and controls, therefore this analysis relied on incidental HIV enrolments. Enrolment, specifically of controls, was also hampered by COVID-19 restrictions. An extension of this analysis would benefit from enrolling controls from HIV clinics. Viral load was not analysed here, since few recent results were available; thus recent CD4+ cell count was used as a proxy. CD4+ cell counts were not available for all HIV-infected cases at the time of admission. Many of those without CD4+ cell counts available were matched from NHLS data, however these tests were not always performed at the same time as the enrolment. We used a cut off of 12 months on either side of enrolment for CD4+ cell counts as this was deemed to be a reasonable estimation of the CD4+ cell count at admission. In cases where there were significant changes in CD4+ cell counts over time, these may have been incorrectly quantified. There were also limited data available for antibiotic use. Date of last dose was available, however start date for antibiotic course was not collected. This has limitations specifically for the C. difficile analysis, since we could not determine the length of antibiotic courses. Future studies should also investigate the expansion of diagnostic testing for M. avium which was not done in the present study.

In conclusion, this study highlights the importance of HIV-related diarrhoea amongst South African inpatients ≥5 years of age, and underscores the importance of research in a high HIV prevalence setting in order to improve the understanding of aetiology of diarrhoea. These data are important for development of guidelines and treatment protocols as well as for preventative interventions, such as vaccine introduction. Our data specifically highlights the importance of Shigella spp. amongst both HIV-infected and HIV-uninfected diarrhoeal patients. We suggest that research be expanded to primary healthcare levels in order to include a broader spectrum of HIV disease, specifically to investigate diarrhoea in individuals with well-controlled HIV.

Supplementary tables