This study was approved by the Institutional Ethics Committee and Institutional Review Board of International Health Sciences University: approval number IHSU-REC/0046. All participants provided written informed consent before the enrolment, and for participants younger than age 18 years, written informed consent was provided by a parent or legal guardian.

This was a cross-sectional study carried out at the Diabetes Clinic of St. Francis Hospital Nsambya, Kampala, Uganda. Two hundred and eighty DM patients attending the hospital’s diabetes treatment clinic were assessed by a medical officer for clinical presentation consistent with diarrhoea.

Clinical data were obtained from the consenting participants’ medical records for demographic characteristics such as age, sex, type of diabetes and date of diabetes diagnosis plus information on use of antibiotics, corticosteroids, anti-diabetes and antifungal medication. Stool specimens and blood samples were collected from the recruited participants according to the United States Centers for Disease Control and Prevention 2014 guidelines.¹⁶ In brief, a sterile, wide-mouthed spoon fitted with a graduated stool container was labelled for each participant and participants were instructed to produce about 10 mL of stool. The specimens were immediately delivered to the laboratory for subsequent analysis. One 4-mL fasting blood specimen was collected from each participant into fluoride/oxalate tubes and the plasma separated from the cells within 30 min after collection.

Potassium hydroxide wet mounts and smears for Gram staining were prepared from each stool specimen to examine them for Candida blastoconidia and pseudohyphae. Specimens were cultured on Saboraud dextrose agar (Laboratorios Conda, Madrid, Spain) for colony counts and Candida differential agar (Himedia Laboratories, Mumbai, India) for species identification; cultures were incubated at 37 °C for 24–72 h and checked for growth.¹⁶ The Germ tube test and growth test at 45 °C were performed to distinguish between C. albicans and C. dublinensis. Colonies on Sabouraud dextrose agar were enumerated and counts above 10⁵ CFU/mL were interpreted as overgrowth indicative of infection, based on reviewed literature.^2,16.

Following species identification of isolates, susceptibility testing was performed using the Kirby-Bauer disk diffusion test on a Mueller Hinton agar supplemented with 2% glucose and methylene blue (Himedia Laboratories, Mumbai, India). Inoculum was prepared by picking five distinct colonies of approximately 1 mm from 24-h-old culture grown on Sabouraud dextrose agar. Colonies were suspended in 5 mL of sterile 0.85% saline and turbidity adjusted to 0.5 McFarland standard which corresponds to an approximate yeast density of 1 × 10⁶ to 5 × 10⁶ cells/mL. The surface of the Muller Hilton agar was dried and seeded with the yeast suspension using a sterile cotton swab by streaking the entire agar surface of the plate with the swab three times, turning the plate 60 degrees between each streaking. The inoculum was allowed to dry for 5 min – 15 min with the lid in place, the antifungal discs were placed on the agar surface aseptically and then the plates were incubated at 35 °C ± 2 °C within 15 min after the discs were applied. The plates were examined for susceptibility after 20–24 h of incubation or at 48 h when insufficient growth was observed after 24 h incubation. Inhibitory zone diameters were measured in millimetres at the transitional point where growth abruptly decreased, as determined by a marked reduction in colony sizes.²⁰

Antifungal susceptibility disks were used to test for susceptibility to two triazoles (fluconazole and itraconazole), two imidazoles (clotrimazole and miconazole) and one polyene (nystatin) (Laboratorios Conda, Madrid, Spain). Fasting blood glucose levels were tested using a point-of-care Accu-chek glucose meter (Roche Diabetes Care, Inc., Ängelholm, Sweden) and glycated hemoglobin levels were determined using an automated glycated hemoglobin 501 analyser (HemoCue AB, Ängelholm, Sweden).

Growth testing for all culture media was performed using reference Candida strains (Candida albicans ATCC 10231, Candida glabrata ATCC 15126, Candida krusei ATCC 24408, Candida tropicalis ATCC 750) and negative control strains (Escherichia coli ATCC 25922 and Staphylococcus aureus ATCC 25923) as recommended by the manufacturers. Susceptibility testing was performed in adherence to National Committee for Clinical Laboratory Standards guidelines.²⁰ Procedures for Gram staining, serial dilutions and culture techniques were performed according to protocols developed by the American Society for Microbiology.²¹ Blood glucose and glycated hemoglobin determination were performed on calibrated devices following the manufacturer’s manual.

Data were entered in an Excel spreadsheet and analysed using STATA special edition, Version 10.0 (StataCorp, College Station, Texas, United States). Data regarding study population characteristics, proportion of patients with gastrointestinal candidiasis, species distribution and susceptibility profiles were analysed using frequency distributions and 95% confidence intervals. Bivariate analysis through the use of chi-square tables, and multivariate logistic regression were used to analyse associations between risk factors and gastrointestinal candidiasis. Statistical significance was assumed to exist if p-values were less than 0.05. The risk factors studied were age of the participants, sex, glycaemic control (‘good’ control = glycated haemoglobin level < 7.0%; ‘poor’ control = glycated haemoglobin > 7.0%), history of antibiotic therapy, history of corticosteroid therapy and type of DM. Participants were analysed in four age groups: children (< 18 years), youth (18–40 years), adult (41–65 years) and elderly (> 65 years). Diabetes type was classified based on diagnostic information on the patient’s form into type 1 and type 2 DM. The history of therapies was classified as Yes if the patient had treatment for more than 2 weeks but less than 1 month earlier or No if the patient had therapies either for less than 2 weeks or more than 1 month earlier.

Cultures from 37 (15.4%) patients had colony counts consistent with gastrointestinal candidiasis (≥ 1 × 10⁵ colony-forming units/mL of stool). Among infected patients, 11 (29.7%) had type 1 DM, and 26 (70.3%) had type 2 DM (p = 0.706). About one-third of patients with gastrointestinal candidiasis (25, 67.6%) were female, and 12 (32.4%) were male (p = 0.009).

The majority of the isolates were C. albicans (n = 23, 62.2%), followed by C. glabrata (n = 7, 18.9%), C. tropicalis (n = 3, 8.1%), Candida dublinensis (n = 2, 5.4%) and C. krusei (n = 2, 5.4%). Among female patients, only C. albicans and C. dubliniensis were isolated (Figure 1). C. glabrata, C. krusei and C. tropicalis were isolated exclusively from male patients.

A total of 25 of the 37 isolates (67.6%) were susceptible to fluconazole and 3/37 (8.1%) isolates were resistant to fluconazole. Itraconazole susceptibility was observed in 33 (89.2%) isolates; 15 (40.5%) isolates were susceptible to clotrimazole, while 7 (18.9%) were resistant and 18 (48.6%) isolates were resistant to miconazole, while 8 (21.6%) were sensitive. All isolates were susceptible to nystatin (Table 2).

There were 12 (34.4%) male, culture-positive patients and 25 (67.6%) female, culture-positive patients (p = 0.01). Twenty-five (67.6%) of the culture-positive patients reported prior use of antifungal drugs, whereas the rest had no history of antifungal drug use in the past weeks (p < 0.001); 26 (70.3%) culture-positive patients had used antibiotics and 11 (29.7%) had not used antibiotics in the past weeks; 14 (37.8%) culture-positive patients had used corticosteroids and the rest had not used corticosteroids in the past weeks (p < 0.001) (Table 3). Thirty-five had poor glucose control, and 20 were hyperglycaemic. Based on glycated haemoglobin results, 2 (5.4%) culture-positive patients had good glycaemic control and the remaining 35 (94.6%) culture-positive patients had poor glycaemic control (p < 0.001).

C. albicans accounted for the majority of isolates in the current study. This is consistent with other studies that identified it as the most common isolate from clinical materials,^27,28 and could be attributed to the fact that C. albicans is highly adapted to the human mucosal surfaces and possesses virulence factors such as protease production and biofilm formation.^3,29 These enhance its chances of survival in the gastrointestinal tract. Omrani et al.⁹ in their review of African studies also found a predominance of C. albicans. Non-albicans species in general accounted for a larger number of isolates in studies by Banerjee et al.³⁰ in India, as well as in Brazil and Chile.³¹ No mixed species infections were encountered in this study, unlike the studies in Poland² and India.³⁰ Among non-albicans species in this study C. glabrata was predominant, accounting for half (7/14) of all non-albicans isolates. This was followed by C. tropicalis (3/14), C. krusei (2/14) and C. dubliniensis (2/14). Species diversity was greater in male patients and type 2 DM patients; however, the latter could simply be a reflection of the fact that the type 2 DM sub-group accounted for the majority of patients. The relationship could perhaps be better established with a cohort study design.

There are relatively fewer options for the treatment of mycoses compared with antibiotics. Fluconazole is a narrow spectrum fungistatic azole with good activity against yeasts; however, from the mid-1990s concerns about resistance have persisted.³ This study found that 87% of C. albicans isolates were fluconazole-susceptible. Similarly, the study in Poland² and a study conducted in 2013 in India³² also reported fluconazole susceptibility in over 80% of C. albicans isolates. Fluconazole resistance among non-albicans species was found to be 21%. This was considerably lower than observations from the Polish study² where 56% of non-albicans species were susceptible to fluconazole. This has implications for clinical therapy since the guidelines for the treatment of non-albicans infection in Uganda require fluconazole.

Intrinsic resistance in non-albicans species has been documented in species such as C. krusei, and reduced susceptibility in C. glabrata and C. guilliermondii has been reported.^33,34 Candida tropicalis has been known to exhibit fluconazole resistance up to 31.3%²⁸; however, this could be due to the use of a panel that mostly comprised high-potency drugs normally reserved for systemic infections, such as amphotericin B and voriconazole. In this study C. krusei, was also the most fluconazole-resistant strain. This strain has been known to exhibit intrinsic fluconazole resistance; Sanguinetti et al.²⁸ reported an estimated global fluconazole resistance of 78% in C. krusei.

Itraconazole has a wider spectrum of activity than fluconazole, and is normally effective against both fluconazole-susceptible and fluconazole-resistant Candida strains.¹⁷ Lesser susceptibility, however, was observed in 2 C. glabrata and both C. krusei isolates. The Polish study,² however, found that only 28% of C. albicans and 11% of non-albicans species were itraconazole susceptible. Higher rates of resistance were observed in the imidazoles; 18.9% of isolates were resistant to clotrimazole and nearly half were resistant to miconazole.

It is worth noting that a number of patients in this study had prior exposure to antifungal medications, especially azoles, which were mostly used to treat dermatomycoses and Candida. Clotrimazole and miconazole are commonly used over-the-counter antifungals, and these could therefore be a driver for the reduced susceptibility and resistance patterns observed in some patient isolates.

All isolates in this study were susceptible to nystatin, which is consistent with reports that document low rates of polyene-class antifungal resistance.^3,17,33 Nystatin carries the additional advantage of being a low-cost medicine and is widely available in many formulations including tablets, suspension and topical preparations. In Uganda, the treatment plan for oropharyngeal candidiasis and gastrointestinal candidiasis may require the use of nystatin preparations, which replaces the expensive and less available medicines such as caspofungins.

There was no significant association between the diabetes type and gastrointestinal candidiasis. Gosiewski et al.³⁴ also found no association, despite a higher prevalence among those with type 2 DM patients. This was contrary to Kumar et al.,³⁵ who found a higher prevalence among type 1 DM.

Gastrointestinal candidiasis was more common among female patients than among male patients (p = 0.01) and among patients with poor glycaemic control (p ≤ 0.001) than among those with good glycaemic control. The use of antifungals (p ≤ 0.001), antibiotics (p ≤ 0.001) and corticosteroids (p ≤ 0.001) was a predisposing factor to gastrointestinal candidiasis. Similar findings were reported by Banerjee et al.,³⁰ in India, where a majority of patients were on antibiotic and corticosteroid therapy. Antibiotics, especially the broad-spectrum variety, are known to interfere with the balance of gut microbiota in the human body.^27,36,37 In doing so, they create a favourable environment for the proliferation of yeasts, and some studies have reported a possible influence on drug resistance profiles. Ben-Ami et al.,³⁷ found a significant association between the use of antibiotics such as carbapenems, clindamycin and colistin and fluconazole resistance, but in this study the class of antibiotics used by diabetic patients was not reported. Use of antifungal therapy could also select for more resistant strains. This was demonstrated by Lortholary et al.,³⁸ using data from a prospective surveillance programme; the authors observed high rates of isolation of fluconazole-resistant Candida species following the recent use of the drug. In this study, poor antifungal susceptibility at baseline level and inappropriate doses of antifungal drugs could be responsible for the association that was observed.

Increased age is usually associated with decreased effectiveness of the immune response.^1,17 The elderly patients and adults accounted for the majority of infections, but there was no significant association between age and gastrointestinal candidiasis. Therefore, this observation could simply be a reflection of the study population, most of which comprised adults and elderly patients. In contrast, Banerjee et al.³⁰ observed a higher isolation rate in children (0–12 years), and attributed it to the weaker immune systems in such populations. Given the adult: paediatric ratio in that study (1:1.9), however, the sampling method may have also had an influence on this result.

Corticosteroid use was also associated with gastrointestinal candidiasis. This has also been observed by Glavey et al.,³⁹ in their cross-sectional study, and Madhumati and Rajendran.⁴⁰ Kakeya et al.⁴¹ and Fardet et al.⁴² reported an increased risk of susceptibility to opportunistic infections in patients exposed to corticosteroids.

Most patients with gastrointestinal candidiasis in this study had poor glucose control. Studies investigating the association between the two have yielded varying results. Findings from this study are similar to the report by Olczak-Kowalczyk et al.,⁴³ but contrary to studies conducted by Suarez et al.,⁴⁴ Kowalewska et al.² and Arslan et al.²⁹ which did not find association between poor glycaemic control and candidiasis.

As a package for the diagnosis and management of gastrointestinal candidiasis, speciation of Candida isolates should be performed, and antifungal susceptibility profiles established to guide patient therapy. The speciation is important, since species within the Candida genus differ widely, both in their ability to cause infection and also in their susceptibility to antifungal agents. Nystatin was shown to have excellent antifungal activity in this study and could be considered for empirical therapy in settings where inadequate resources may inhibit antifungal susceptibility testing.

This study enrolled only participants who complained of gastrointestinal symptoms and several asymptomatic patients could have been left out. This might have reduced the prevalence of gastrointestinal candidiasis. Routine culture and sensitivity testing was conducted and very sensitive techniques such as molecular techniques were not performed. This could have reduced the detection of Candida organisms.

Prevalence of gastrointestinal candidiasis was relatively low among the participants of this study. The infection is associated with female sex, poor glycaemic control and previous use of antifungals, antibiotics and corticosteroids. Nystatin can be a drug of choice in the treatment of gastrointestinal candidiasis, if the suspension or tablet formulations can be made available.