Long-Term Trends in Esophageal Candidiasis Prevalence and Associated Risk Factors with or without HIV Infection: Lessons from an Endoscopic Study of 80,219 Patients

Background The prevalence of candida esophagitis (CE) might be changing in an era of highly active antiretroviral therapy (HAART) among HIV-infected patients or today’s rapidly aging society among non-HIV-infected patients. However, few studies have investigated long-term CE trends, and CE risk factors have not been studied in a large sample, case-control study. This study aimed to determine long-term trends in CE prevalence and associated risk factors for patients with or without HIV infection. Methods Trends in CE prevalence were explored in a cohort of 80,219 patients who underwent endoscopy between 2002 and 2014. Risks for CE were examined among a subcohort of 6,011 patients. In risk analysis, we assessed lifestyles, infections, co-morbidities, immunosuppressants, and proton-pump inhibitors (PPIs). All patients were tested for HIV, hepatitis B or C virus, and syphilis infection. For HIV-infected patients, sexual behavior, CD4 cell count, history of HAART were also assessed. Results CE prevalence was 1.7% (1,375/80,219) in all patients, 9.8% (156/1,595) in HIV-infected patients, and 1.6% (1,219/78,624) in non-HIV-infected patients. CE prevalence from 2002-2003 to 2012-2014 tended to increase in non-HIV-infected patients (0.6% to 2.5%; P<0.01) and decrease in HIV-infected patients (13.6% to 9.0%; P=0.097). Multivariate analysis revealed increasing age (odds ratio [OR], 1.02; p=0.007), HIV infection (OR, 4.92; p<0.001), and corticosteroid use (OR, 5.90; p<0.001) were significantly associated with CE, and smoking (OR, 1.32; p=0.085) and acetaminophen use (OR, 1.70; p=0.097) were marginally associated. No significant association was found with alcohol consumption, hepatitis B or C virus, syphilis, diabetes mellitus, cardiovascular disease, cerebrovascular disease, chronic kidney disease, liver cirrhosis, anticancer, or PPIs use. In HIV-infected patients, CD4 cell count <100/μL (OR, 4.83; p<0.001) and prior HAART (OR, 0.35; p=0.006) were independently associated with CE, but sexual behavior was not. Among corticosteroid users, CE was significantly associated with higher prednisone-equivalent dose (p=0.043 for trend test). Conclusions This large, endoscopy-based study demonstrated that CE prevalence increased in non-HIV-infected patients but decreased in HIV-infected patients over 13 years. Risk analysis revealed that increasing age, HIV infection, and corticosteroids use, particularly at higher doses, were independently associated with CE, but alcohol, other infections, diabetes, anticancer drugs, and PPIs use were not.


Introduction
Candida esophagitis (CE), one of the most common opportunistic infections of the esophagus [1], is classed as an invasive fungal disease [2]. Severe CE may cause esophageal hemorrhage or progress to stricture or fistula formation, accompanied by reduced quality of life. Early diagnosis and treatment with systemic antifungals is therefore important [3]. Endoscopy is an essential diagnostic tool for detecting fungal white plaque and performing biopsy [4], but it is too costly and invasive to be used in an unselected population. Therefore, high-risk patients must be identified to prevent delays in diagnosis and poor outcomes.
Several studies have reported that proton-pump inhibitors (PPIs), immunosuppressants, and HIV infection are common risk factors for CE, although most of these studies did not have a case-control design, had a small samples, or were restricted to immunocompromised patients [5][6][7][8]. In today's rapidly aging society, it is unclear whether these risk factors remain the same given that comorbidities in general are expected to rise with rapid aging of the population [9] and the suggested risk factors of PPIs and immunosuppressant drugs [6,8,10] are now in widespread clinical use. In contrast, it is known that HIV-related opportunistic infections have decreased dramatically following the introduction of highly active antiretroviral therapy (HAART) [11,12]. Moreover, few studies have investigated CE risk factors in the recent decade and long-term trends in CE prevalence have not been studied to date [7,10]. In light of such societal changes, the risk factors for CE should be reconsidered and trends in CE prevalence investigated to facilitate timely diagnosis and treatment of the disease.
Here we report the results of a large, endoscopy-based, cross-sectional, case-control study conducted over 13 years that aimed to identify trends in annual CE prevalence in patients with or without HIV infection and determine through risk analysis which lifestyle factors, medication, and comorbidities are associated with CE.

Study design, setting, and participants
We conducted two studies at the endoscopy unit of the National Center for Global Health and Medicine (NCGM; Tokyo, Japan) between January 2002 and April 2014, one to investigate CE prevalence and the other to examine risk factors for CE. This study was approved by the ethics committee of the National Center for Global Health and Medicine Center (Nos. 711, 1424) and was implemented in accordance with the provisions of the Declaration of Helsinki. NCGM has 900 beds and is the largest referral center for HIV/AIDS in Japan. All patients underwent serological testing for HIV, hepatitis B virus (HBV), hepatitis C virus (HCV), and syphilis before endoscopy [13]. Written informed consent was obtained from all patients prior to endoscopy.
In Study 1 to identify overall CE prevalence and trends in annual CE prevalence, we reviewed a hospital-based, prospectively collected endoscopic database of 80,360 consecutive patients who underwent upper endoscopy at NCGM between January 2002 and April 2014. Indications for endoscopy were as follows: (i) continual or intermittent upper GI symptoms; (ii) asymptomatic with anemia; (iii) asymptomatic requiring examination for specific diseases due to abnormal findings on tumor marker or fecal occult blood testing or on abdominal imaging; (iv) screening required for gastric cancer; and (v) therapeutic endoscopy required. In Japan, which has a high incidence of gastric cancer [14], endoscopy is frequently performed for gastric cancer screening [15]. We excluded patients who were 18 years of age or who did not provide informed consent. All inclusion and exclusion criteria were fullfilled before the patients were enrolled.
In Study 2 to identify CE risk factors, we conducted a hospital-based, case-control study at the endoscopy unit of the NCGM between September 2009 and April 2014. Inclusion criteria were Japanese nationality and the aforementioned indications for endoscopy. Exclusion criteria were as follows: (i) no informed consent obtained; (ii) unknown use of medication; (iii) not independent in activities of daily living (ADL); (iv) inability to understand written documents; (v) use of any anti-fungal drugs within 1 month before endoscopy; (vi) urgent or early endoscopy for acute GI bleeding; and (vii) second endoscopy required during the study period. All inclusion and exclusion criteria were fullfilled before the patients were enrolled. The control group comprised 5,800 patients undergoing endoscopy during the same period who did not meet our criteria for the diagnosis of CE.

Data sources and measurement
For risk analysis, histories of HBV, HCV, and syphilis infection were collected before endoscopy and defined as presence of antibody against hepatitis B or C surface antigen and positive results on a Treponema pallidum hemagglutination test, respectively, as previously reported [13].
A detailed questionnaire was also completed at the endoscopy unit on the same day as before endoscopy [16]. Well-trained medical researchers asked patients about alcohol consumption, smoking status, comorbidities, and medication use. Researchers also checked prescriptions and medical records in addition to the information provided by the patients to avoid omissions. Alcohol consumption was defined as consumption of alcohol at least once a week.
Smoking was defined as currently smokes or ever smoked. Comorbidities included diabetes mellitus (DM), cardiovascular disease, cerebrovascular disease, chronic kidney disease (CKD), liver cirrhosis, peptic ulcer disease, post-esophageal resection, and post-gastric resection. DM was considered present in patients taking anti-diabetes drugs. CKD was considered present in patients on hemodialysis or peritoneal dialysis, or with serum creatinine levels 2.0 mg/dl. Patients were asked to indicate which drugs, if any, they had used based on drugs pictured in the questionnaire, as previously reported [17,18]. Use of a drug was defined as intermittent or regular administration within 1 month before the questionnaire interview. Patients were asked about the following medication use: anticancer drugs for solid or hematological tumors; 5 types of systemic corticosteroids (prednisone, methylprednisolone, betamethasone, dexamethasone, and hydrocortisone) and daily corticosteroid dose; 8 types of non-steroidal anti-inflammatory drugs (NSAIDs); acetaminophen; 4 types of PPIs (lansoprazole, rabeprazole, omeprazole, and esomeprazole); and 6 types of H2RAs (nizatidine, roxatidine, famotidine, ranitidine, lafutidine, and cimetidine). We calculated prednisone-equivalent daily corticosteroid dose based on the data that glucocorticoid doses providing anti-inflammatory effects approximately equivalent to 5 mg prednisone are methylprednisolone 4 mg, betamethasone 0.75 mg, dexamethasone 0.75mg, and hydrocortisone 20 mg [19].
For HIV-infected patients, routes of infection were determined using questionnaires administered by medical staff [20] and classified into one of six categories: homosexual, bisexual, heterosexual, drug user, use of untreated blood products, or unknown. Sexual behavior was divided into two categories: men who have sex with men (MSM) and heterosexual. Patients who were neither homosexual nor bisexual were regarded as heterosexual. CD4 cell count values obtained within 1 month before or after endoscopy were obtained from the medical records.

Diagnosis of upper GI disease and candida esophagitis
A high-resolution scope (GIF-H260, Olympus Corp., Tokyo, Japan) was used for diagnosis of upper GI diseases. Well-trained staff performed endoscopy while blinded to the questionnaire results. When abnormal findings were detected on endoscopy, biopsy or endoscopic mucosal resection was performed. All removed specimens were evaluated by expert pathologists (>10 years of experience), and final diagnoses of upper GI diseases were made. Diagnosis of CE was made on the basis that candida white plaques in the esophagus, detected by endoscopy, cannot be washed away [21] (Fig 1A and 1B) and confirmation by pathological assessment with hematoxylin and eosin and periodic acid-Schiff staining ( Fig 1C) or culture for Candida species [17,18].

Statistical analysis
In Study 1, we examined trends in CE prevalence over the 13-year study period in all subjects, HIV-infected patients, and non-HIV-infected patients, assessing changes in prevalence in 2-year bins with the Chi-squared test for linear trends.
In Study 2, we summarized descriptive data for patients with and without CE to identify risk factors. Patient characteristics were compared using the Mann-Whitney U test for quantitative variables or the Chi-square test or Fisher's exact test for qualitative variables. The association between CE and other upper GI diseases was evaluated from both the endoscopy and pathology findings. CE risk factors were identified using a multiple logistic regression model including factors with a significance of P<0.2 on univariate analysis. Odds ratios and 95% confidence intervals were calculated. CE risk factors were also evaluated separately for HIVinfected and non-HIV-infected patients.
A P value of less than 0.05 was considered statistically significant. All statistical analyses were performed using Stata version13 software (StataCorp, College Station, TX).

Study 2: Risk factors for CE
Patient characteristics. Between September 2009 and April 2014, among the 32,827 patients who underwent upper endoscopy, we did not conduct questionnaire interviews with 23,485 patients because they either did not provide informed consent, did not know what medications they were receiving, were not independent in ADL, or could not understand written documents, and we found 3,331 patients met one of the exclusion criteria. This left a subcohort of 6,011 patients who underwent endoscopy and completed the questionnaire interview for analysis of CE risk.
Baseline characteristics are shown in Table 1. The factors significantly associated with CE were alcohol consumption, HIV infection, CD4 cell count, history of HAART, HCV infection, syphilis, and use of corticosteroids, NSAIDs, acetaminophen, or PPIs. As shown in Table 2, the upper GI diseases significantly associated with CE were esophageal ulcer, gastric ulcer, and severe atrophic gastritis.
Univariate and multivariate odds ratios for CE. Risk factors for CE identified by multivariate analysis are shown in Table 3. Increasing age, HIV infection, and corticosteroid use were independently associated with CE and smoker and acetaminophen were marginally associated with CE, but sex, alcohol consumption, HCV infection, syphilis, post-gastric resection, and use of NSAIDs or PPIs were not.   The risk factors identified by multivariate analysis for HIV-infected and non-HIV-infected patients are shown in Table 4. In HIV-infected patients, CD4 cell count<100μL significantly increased the risk of CE and history of HAART significantly decreased the risk. Among HIVinfected patients, 25% of the CE patients had CD4 cell counts >400/μL. In non-HIV-infected patients, increasing age and corticosteroid use were independently associated with CE, and HCV infection, peptic ulcer disease, and post-esophageal resection were marginally associated. Among corticosteroid users, higher prednisone-equivalent daily corticosteroid doses significantly increased the risk for CE (Table 5).

Discussion
In this study, we identified that, over the 13-year period from 2002 to 2014, CE prevalence showed significant increases in non-HIV-infected patients but decreases in HIV-infected patients. Moreover, risk analysis revealed that increasing age, HIV infection, and corticosteroids use were independently associated with CE, but alcohol, other infections, diabetes, anticancer drugs, and antisecretory drug use were not. Among HIV-infected patients, low CD4 cell count and absence of HAART were independent risk factors for CE. In corticosteroid users, higher prednisone-equivalent daily corticosteroid dose significantly increased CE risk.
CE prevalence in non-HIV-infected patients is reported to 0.3-7.3% [5,8,10,22,23], which is consistent with our findings of 1.5% in non-HIV-infected patients. Similarly for HIV-infected  [17,24], with which our finding of 9.8% prevalence in HIV-infected patients, most of whom received HAART (80.3%), concurs. Little information about CE trends has been available to date, and most studies have focused on HIV-infected patients [12,24]. Declining CE incidence has been reported in this patient population by Mocroft et al [12], from 5.7 person-years in 1994 to 0.5 person-years in 2004, and Nkuize et al [24], from 42.8% in 1991-1994, 23.5% in 1999-2002, and 16.7% in 2005-2008. Likewise, CE prevalence decreased from 13.6% in 2002-2003 to 9.0% in 2012-2014 in the present study. This is probably due to the introduction of HAART, which helped avoid immunodeficiency, and the development of effective prophylaxis against opportunistic infection [26]. In contrast, CE prevalence in non-HIV-infected patients showed a significant increase over the same period, likely because, in a rapidly aging society, more people have comorbidities and take immunosuppressants [27,28], which can lead to the development of CE.
Several risk factors for CE have been investigated in case-control studies [6][7][8]10], although these studies had a small number of CE cases (three studies; n = 56 [6][7][8], one study; n = 163 [10]) and retrospective in nature. They identified corticosteroid or PPIs use, heavy drinking, uncontrolled DM, and cancer as CE risk factors. In our larger scale study investigating the very same factors, multivariate analysis similarly identified systemic corticosteroid use but also revealed increasing age and HIV infection as risk factors for CE. In terms of systemic corticosteroid use, an association was found in non-HIV-infected patients in this study. This could be explained by the suppression of macrophages and polymorphonuclear leucocytes by glucocorticoids, which can in turn lead to the development of CE [29]. Moreover, Heidenreich et al [30] showed that glucocorticoids principally affect the capacity of monocytes to control extracellular growth of Candida species by inhibiting tumor necrosis factor alpha secretion. We also found that higher prednisone-equivalent dose of corticosteroids was a risk factor for CE. Among the very few studies to date that have investigated the association between opportunistic infection and steroid dose, Yale et al [31] demonstrated that >30 mg of steroid daily was further associated with worsened Pneumocystis carinii pneumonia (PCP), compared with <30 mg daily, and Porges et al [32] demonstrated that >40 mg daily was associated with PCP development among patients with systemic lupus erythematosus. Our study is the first to investigate the association between steroid dose and CE and our findings suggest that patients who receive high-dose steroids have lower cellular immunity, which can lead to CE development.
Increasing age was identified as a risk factor for CE in non-HIV-infected patients in this study. It is possible that, in non-HIV-infected patients, increasing age leads to impaired immunity due to defects in the hematopoietic bone marrow and in peripheral lymphocyte migration, maturation, and function [9]. Another reason could be that increasing age causes a decline in cellular immunity of the epithelial layer, which can lead to colonization of Candida species [33].
HIV infection itself was a risk factor for CE in this study. We suggest that impaired oral immunity plays a role in this. Oral immunity may be affected earlier than immunity in other organs [34] and thus facilitate the development of CE early in HIV infection. Indeed, Chih-Ko et al [34] demonstrated that salivary function is impaired early in the course of HIV infection. In the present study, the analysis of HIV-infected subjects showed an association between CD4 cell count <100/μL and CE, although 25% of HIV-infected patients with CE also had a relatively high CD4 count (CD4>400). Buchacz et al found that the median CD4 cell count in CEinfected patients increased from 43/μL in the pre-HAART era (1994)(1995)(1996)(1997) to 100/μL in the HAART era (2003)(2004)(2005)(2006)(2007). These findings suggest that in the HAART era, HIV-infected patients are likely to develop CE even though their CD4 cell count is relatively high. Previous studies have suggested that PPIs use is associated with CE because PPIs reduce gastric acidity, an important barrier for most microorganisms, and this can result in a number of infections [35]; however, the findings of this association are inconsistent. Chocarro et al [6] showed that, among 51 CE patients and 102 controls, PPIs use was independently associated with CE. In contrast, in case-control studies by Weerasuriya et al [7] and Choi et al [10], PPIs use was not associated with CE, consistent with our data. In clinical practice, PPIs are frequently administered when NSAIDs are used for protection from NSAID-induced peptic ulcer disease, but NSAID use was not adjusted for in previous studies that were also small in scale [6,7,10]. Therefore, a large-scale study with adjustment for NSAIDs is required to identify whether PPIs are a risk of CE.
The strengths of this study include its prospective nature, large scale, and evaluation of detailed HIV information, which was obtained by performing HIV testing in all subjects. Its limitations include the following facts. 1) Because sufficient information was not collected on the pre-endoscopy setting, we could not evaluate the potential risk factors of chronic obstructive pulmonary disease and antibiotic or herbal medication use. 2) Selection bias was present because endoscopy could not be performed for critically ill or intubated patients. 3) Endoscopy is not always performed for HIV-infected patients with oral candidiasis or esophageal symptoms who are empirically treated with antifungal therapy. Thus, this study potentially underestimated the number of CE cases. 4) This study has a cross-sectional design, so we can only comment on risk factors as being associated with disease but we cannot truly understand the causal pathways. 5) This case-control study is potentially affected by recall or interviewer bias, particularly for survey items on alcohol consumption, medication use, and smoking. Therefore, we attempted to limit bias by blinding survey results and performing surveys prior to endoscopy. 6) The community setting of the controls (undergoing endoscopy without CE) may have been different from that of cases (patients with CE); thus, unmeasured confounders exist. In particular, co-morbidities have a confounding effect. In addition, the study would have been better if subjects had been matched for age, sex, and co-morbidities rather than just adjusting the model for these factors.
In conclusion, this study identified that CE prevalence over a 13-year period increased in non-HIV-infected patients but decreased in HIV-infected patients. Besides increasing age, HIV infection itself and corticosteroid use increased the risk for CE development, but antisecretory drug use did not. In HIV-infected patients, a low CD4 count and absence of HAART were independent risk factors for CE. In corticosteroid users, CE was associated with higher prednisone-equivalent dose. Given that the elderly tend to have co-morbidities and take drugs, CE prevalence is expected to continue increasing in today's rapid aging society.