Factors associated with time free of oral candidiasis in children living with HIV / AIDS , São Paulo , Brazil

In clinical practice, recurrence of thrush is common in children living with HIV/AIDS. The aim of this study was to determine the factors associated with time spent free of oral candidiasis using survival analysis for recurrent events. A retrospective cohort study was carried out with 287 children treated between 1985 and 2009 at a reference center in the city of São Paulo, Brazil. The Prentice, Williams and Peterson model for recurrent events was used for the investigation of factors associated with the time free of oral candidiasis. The following factors were associated with the time patients were free of oral candidiasis: moderate immunodepression (HR = 2.5; p = 0.005), severe immunodepression (HR = 3.5; p < 0.001), anemia (HR = 3.3; p < 0.001), malnutrition (HR = 2.6; p = 0.004), hospitalization (HR = 2.2; p < 0.001), monotherapy (HR = 0.5; p = 0.006), dual therapy (HR = 0.3; p < 0.001) and triple therapy/highly active antiretroviral therapy (HR = 0.1; p < 0.001). The method analyzed in the present study proved useful for the investigation of recurrent events in patients living with HIV/AIDS. Oral Candidiasis; HIV; Acquired Immunodeficiency Syndrome; Child; Survival Analysis Resumo A recorrência da candidíase oral em crianças vivendo com HIV/AIDS é um acontecimento muito comum na prática clínica. O objetivo foi verificar os fatores associados ao tempo livre de candidíase oral, utilizando técnica de análise de sobrevida para eventos recorrentes. Estudo de coorte retrospectivo com 287 crianças, atendidas entre 1985 e 2009, em um serviço de saúde de São Paulo, Brasil. Foi utilizado o modelo marginal para eventos recorrentes de Prentice, Williams e Peterson para investigação dos fatores associados ao tempo livre de candidíase oral. Imunodepressão moderada (HR = 2,5; p = 0,005) ou grave (HR = 3,5; p < 0,001), anemia (HR = 3,3; p < 0,001), desnutrição (HR = 2,6; p = 0,004) e internação (HR = 2,2; p < 0,001), monoterapia (HR = 0,5; p = 0,006), terapia dupla (HR = 0,3; p < 0,001) e terapia tripla/HAART (HR = 0,1; p < 0,001) foram associados ao tempo livre de candidíase oral. A metodologia apresentada neste artigo pode ser bastante útil em pesquisas na área de HIV/AIDS, quando pretende-se estudar eventos com comportamento de recorrência. Candidíase Bucal; HIV; Síndrome de Imunodeficiência Adquirida; Criança; Análise de Sobrevida 2197 ARTIGO ARTICLE http://dx.doi.org/10.1590/0102-311X00069213 Kornstantyner TCRO et al. 2198 Cad. Saúde Pública, Rio de Janeiro, 29(11):2197-2207, nov, 2013 Introduction Persistent oral candidiasis (thrush) is the most frequent fungal condition among children living with human immunodeficiency virus/acquired immunodeficiency syndrome (HIV/AIDS) 1. On the systemic level, this population is prone to the development of thrush due to the immunodeficiency caused by the harmful action of HIV on CD4+ T lymphocytes. Locally, an increase in the permeability of the epithelium and change in humoral immunity (IgA) of the oral mucosa stemming from the action of HIV facilitate infection by Candida 2. Other factors commonly found in children living with HIV/AIDS, such as anemia, hospitalization, antiretroviral drugs and broadspectrum antibiotics, have also been implicated in the pathological mechanism of thrush 3,4. Persistent oral candidiasis is considered an important condition in the context of HIV/AIDS, as it affects quality of life and is an indicator of the progression of HIV infection 5,6,7,8,9. Consequently, a large amount of information has been published on the incidence of thrush and associated factors in periods prior to and following the advent of highly active antiretroviral therapy (HAART) using a single episode of thrush per individual as the unit of analysis 10,11,12. However, the recurrence of this condition is a very common event in clinical practice, as children living with HIV/AIDS often experience numerous episodes throughout their lives. Thus, the question arises as to whether the study of factors associated with the occurrence of thrush should consider only a single episode or multiple episodes throughout patient follow-up. Few studies have employed survival analysis techniques developed for recurrent events, such as marginal models, in the investigation of this condition 13,14,15,16,17. Thus, the aim of the present study was to determine factors associated with the time spent free of oral candidiasis in children living with HIV/AIDS using survival analysis for recurrent events.

In clinical practice, recurrence of thrush is common in children living with HIV/AIDS.The aim of this study was to determine the factors associated with time spent free of oral candidiasis using survival analysis for recurrent events.A retrospective cohort study was carried out with 287 children treated between 1985 and 2009 at a reference center in the city of São Paulo, Brazil.The Prentice, Williams and Peterson model for recurrent events was used for the investigation of factors associated with the time free of oral candidiasis.The following factors were associated with the time patients were free of oral candidiasis: moderate immunodepression (HR = 2.5; p = 0.005), severe immunodepression (HR = 3.5; p < 0.001), anemia (HR = 3.3; p < 0.001), malnutrition (HR = 2.6; p = 0.004), hospitalization (HR = 2.2; p < 0.001), monotherapy (HR = 0.5; p = 0.006), dual therapy (HR = 0.3; p < 0.001) and triple therapy/highly active antiretroviral therapy (HR = 0.1; p < 0.001).The method analyzed in the present study proved useful for the investigation of recurrent events in patients living with HIV/AIDS.

Introduction
Persistent oral candidiasis (thrush) is the most frequent fungal condition among children living with human immunodeficiency virus/acquired immunodeficiency syndrome (HIV/AIDS) 1 .On the systemic level, this population is prone to the development of thrush due to the immunodeficiency caused by the harmful action of HIV on CD4+ T lymphocytes.Locally, an increase in the permeability of the epithelium and change in humoral immunity (IgA) of the oral mucosa stemming from the action of HIV facilitate infection by Candida 2 .Other factors commonly found in children living with HIV/AIDS, such as anemia, hospitalization, antiretroviral drugs and broadspectrum antibiotics, have also been implicated in the pathological mechanism of thrush 3,4 .
Persistent oral candidiasis is considered an important condition in the context of HIV/AIDS, as it affects quality of life and is an indicator of the progression of HIV infection 5,6,7,8,9 .Consequently, a large amount of information has been published on the incidence of thrush and associated factors in periods prior to and following the advent of highly active antiretroviral therapy (HAART) using a single episode of thrush per individual as the unit of analysis 10,11,12 .However, the recurrence of this condition is a very common event in clinical practice, as children living with HIV/AIDS often experience numerous episodes throughout their lives.Thus, the question arises as to whether the study of factors associated with the occurrence of thrush should consider only a single episode or multiple episodes throughout patient follow-up.
Few studies have employed survival analysis techniques developed for recurrent events, such as marginal models, in the investigation of this condition 13,14,15,16,17 .Thus, the aim of the present study was to determine factors associated with the time spent free of oral candidiasis in children living with HIV/AIDS using survival analysis for recurrent events.

Methods
A retrospective cohort study was carried out with children treated at a reference center for the treatment of AIDS in the city of São Paulo, Brazil.Data from medical charts were collected by a team of healthcare professionals using a previously tested, standardized chart.
The physicians at the reference center were trained to diagnose persistent oral candidiasis through: (i) a microscopic examination of a sample of the oral mucosa with characteristic findings; or (ii) a macroscopic examination of removable white plaques on erythematous tissue, following the recommendations of the Brazilian Ministry of Health 18 .
The study population was made up of patients with a laboratory confirmation of HIV infection contracted through vertical transmission and less than 13 years of age at the time of admission to the health center.Patients in follow up for less than 90 days were excluded from the study.Children admitted to the health center between January 1, 1985 and October 1, 2009 were included and followed up until December 31, 2009 (end of the study).Different criteria for the definition of HIV infection and clinical conditions were used during the long follow-up period of the cohort, according to year of diagnosis 18,19,20,21 .
The initial population was made up of 452 children, 325 of whom met the inclusion criteria.However, 38 were excluded for having a followup period of less than 90 days.Thus, the final sample was made up of 287 children.
A descriptive analysis (absolute and relative frequency distribution) of the cohort was first performed.The Kaplan-Meier product limit estimator 22 was then used to estimate the cumulative probability curves for the time between admission to the study and the emergence of oral candidiasis in relation to the independent variables (time of HIV diagnosis, gender, age, immune system impairment, viral load, anemia, malnutrition, hospitalization and antiretroviral regimen).The log-rank test was used for the comparison of the curves 23,24 .The (-ln(-ln(S(t)))) method in function of the follow-up time logarithm was used to determine the parallelism of the curves.
The Prentice, Williams and Peterson (PWP) marginal model for recurrent events was used to investigate factors associated with the time free of oral candidiasis.This model considers events in time and assumes that the basal risk of a new event is different from but dependent on the previous event 24,25 .The hazard ratio (HR) was estimated for each independent variable.The Wald test was used to test the hypothesis that HR = 1 25 .The null hypothesis was rejected when the p-value was ≤ 0.05.
For the use of the PWP model, an event was considered to be the occurrence of an episode of oral candidiasis and it was defined that up to three episodes per individual could be counted.Censure occurred in four situations: (i) patients in follow up at the end of the study; (ii) those transferred to another health center; (iii) those who abandoned follow up; and (iv) those who died.
The 287 children were distributed among three risk strata.A particular number of obser-vations were entered in the database for each child based on the number of episodes of oral candidiasis: 198 children who did not develop oral candidiasis made up stratum 1 (each contributed one observation to the database); 44 exhibited one episode of oral candidiasis and made up strata 1 and 2 (each contributed two observations to the database); 26 exhibited two episodes of oral candidiasis and made up strata 1, 2 and 3 (each contributed three observations to the database); and 19 exhibited three episodes of oral candidiasis and made up strata 1, 2 and 3 (each contributed three observations to the database).Thus, the cohort was formed by 421 observations among 287 children (Figure 1).
Time was the dependent variable and was defined based on the order of occurrence of events: • 1 st time interval (stratum 1): period (in years) between the beginning of follow up at the reference center and the date of the first episode of oral candidiasis or censure; • 2 nd time interval (stratum 2): period (in years) between the date of the first episode and the date of the second episode of oral candidiasis or censure; • 3 rd time interval (stratum 3): period (in years) between the date of the second episode and the date of the third episode of oral candidiasis or censure.
The forward stepwise method was employed in the multivariate modeling.For such, variables with a p-value < 0.20 in the univariate analysis of the PWP model were selected.The variables that maintained statistical significance (p ≤ 0.05) (immune system impairment, anemia, malnutrition, hospitalization and antiretroviral regimen) and those that statistically adjusted the parameters of the other variables (time of HIV diagnosis) remained in the final model.Proportionality of risk over time was determined using Schoenfeld's residual analysis, which employs a chi-square statistic with one degree of freedom based on the proportion of observed and expected survival 22    Figure 2 shows the cumulated probability of time free of oral candidiasis according to the independent variables.Significant differences among the curves were found for the following variables: period of HIV diagnosis (p < 0.001), immune system impairment (p < 0.001), anemia (p < 0.001), malnutrition (p < 0.001), hospitalization (p < 0.001) and antiretroviral regimen (p < 0.001).The survival time free of oral candidiasis was greater among the children with a diagnosis of infection following the advent of HAART and those without anemia, malnutrition or hospitalization.In contrast, the survival time was reduced with the exacerbation of immune system impairment as well as with the reduction in the number of drugs in the antiretroviral regimen.

Discussion
Candidiasis has considerable clinical significance in the prognosis of HIV infection and as an indicator of the non-effectiveness of antiretroviral treatment 26,27 .While some authors state that the use of statistical methods that do not consider the recurrence of episodes of thrush could lead to errors in the estimate of risks 14,28,29,30 , there is debate regarding whether the analysis of a single episode may correspond equally to the joint analysis of multiple episodes in some situations 31 .
In the present study, the investigation of factors associated with the time patients are free of oral candidiasis revealed the following independent predictors of the disease: moderate to severe immunodepression, anemia, malnutrition and hospitalization.Moreover, the use of antiretroviral therapy (monotherapy, dual therapy or triple therapy/HAART) proved to be an independent protection factor.
Survival free of disease was closely linked to the degree of immunodepression, as risk increased with the exacerbation of immune system impairment.This association has been described in previous cohorts of children throughout the world 32,33,34 .However, researchers state that defenses against Candida may be compromised by HIV even before the drop in the levels of CD4+ T lymphocytes 35,36 , which may explain the appearance of thrush in individuals with HIV whose defense cell counts are within the range of normality.
In the present study, anemia was a predictor of oral candidiasis, as the time free of this disease was greater in children without anemia.Different clinical presentations of oral candidiasis, especially the mucocutaneous and angular forms, have previously been associated with anemia 37 .Besides iron deficiency, which is the main cause of childhood anemia, other nutritional disorders have been implicated in the occurrence of oral  candidiasis among children living with HIV/ AIDS.In the present study, the time patients were free of oral candidiasis was significantly shorter among malnourished children.Previous studies have also found malnutrition to be a systemic risk factor for the development of oral candidiasis 38,39 .Hospitalization was investigated as a proxy variable of serious diseases to which children infected with HIV are subject.Hospitalization is a frequent, critical process among such children due to the exacerbation of the clinical situation or the social status of the child 40,41 .In the present study, a significant association was found with hospitalization (regardless of the determinant cause), as survival time free of oral candidiasis was greater in children who had not been hospitalized.
The present study also demonstrated the beneficial effect of antiretroviral drugs, especially HAART, in the prevention of oral candidiasis.This finding has been previously reported for Brazilian children 42,43 .Moreover, the time patients were free of oral candidiasis increased with the addition of new drugs to the antiretroviral regimen.These results are in agreement with findings reported in previous studies, which defend the hypothesis that antiretroviral drugs not only reestablish systemic immunological competence, but also improve the immune response of the oral mucosa against the fungus 44,45 .This may explain situations in which the resolution of oral candidiasis precedes the reestablishment of normal CD4+ T lymphocyte counts 46 .
No associations were found between survival time free of oral candidiasis and the gender or age of children living with HIV/AIDS.These results are in agreement with findings reported in a cohort of American children 33 .
Due to the nearly 100% prevalence rate of a detectable viral load, it was not possible to determine the association between this variable and the risk of developing oral candidiasis.However, high viral loads have been associated with compromised oral health in the pediatric population living with HIV/AIDS 36,47,48 .Moreover, while no analysis was performed of associations between the prior use of antibiotics and/or corticoids and the appearance of oral candidiasis in the present study, these groups of medications have been described as risk factors for this condition due to the fact that they alter the local flora and create a favorable environment for the fungus 37 .
The present study has limitations that should be considered.Due to the cohort design, the reference for the comparisons was the availability of different treatment regimens in different times of the year and with different degrees of adherence.However, this strategy is employed and defended by a number of authors 49,50 .Different criteria for the definition of cases were used during the long follow-up period of the cohort, with an increasingly earlier diagnosis and onset of treatment/prophylaxis over the years.Moreover, the source document for the data collection was the child's medical chart, which was not always filled out satisfactorily for the purposes of the present study.To minimize possible bias, care was taken  to ensure the standardization of the data collection from the charts and a review of all questionnaires.Additionally, the inclusion criteria (acquisition of HIV infection through vertical transmission and age zero to 13 years) minimized difficulties in the comparison of the children analyzed with regard to clinical-epidemiological data and strengthened the homogeneity of the sample.It should be stressed that the choice of the PWP marginal model is considered adequate for modeling recurrent events and the variables of interest are known and available for the analysis of factors associated with the time patients are free of oral candidiasis 30,51 .Despite the sophistication of current statistical packages, which enable the complete execution of this type of modeling in a matter of minutes, the organization and analysis of the databank were performed with caution, since the statistical theory behind this type of model is complex.In conclusion, the present study identified immunodepression, anemia, malnutrition, hospitalization and antiretroviral therapy as factors associated with the time patients are free of oral candidiasis in a population of children living with HIV/AIDS in the city of Sao Paulo, Brazil.Moreover, the method presented in this article can be useful for the investigation of recurrent events in studies addressing HIV/AIDS.

Figure 1 Formation*
Figure 1 Formation of cohort for analysis of factors associated with oral candidiasis in children infected by HIV; reference center in São Paulo, Brazil, 1985-2009.

Figure 2 Cumulative
Figure 2 Cumulative probability of time free of oral candidiasis in children infected by HIV; reference center in São Paulo, Brazil, 1985-2009.

Figure 2
Figure 2 (continued) Contributors T. C. R. O. Konstantyner participated in the design and planning, data collection and analysis and interpretation, elaboration of the draft, and approval of the final version of the article.A. M. Silva participated in the design and planning, data collection, critical review of the content, and approval of the final version of the article.L. F. Tanaka participated in the design and planning, data collection, critical review of the content, and approval of the final version of the article.H. H. S. Marques participated in the design and planning, critical review of the content, and approval of the final version of the article.M. R. D. O. Latorre participated in the design and planning, critical review of the content, and approval of the final version of the article.
. All analyses were performed using the Stata program, version 11 (Stata Corp., College Station, USA).This study received approval from the Human Research Ethics Committee of the School of Public Health, São Paulo University, Brazil (USP) under process number CEP/FSP n. 2033.

Table 1
Distribution of observations * according to demographic, clinical, laboratory and treatment characteristics; reference center in SãoPaulo, Brazil, 1985-2009.
# Most recent regimen before appearance of oral candidiasis in cases of occurrence of disease or last regimen received throughout study for cases in which disease did not occur.Cad.Saúde Pública, Rio de Janeiro, 29(11):2197-2207, nov, 2013

Table 2
Univariate and multivariate analyses of factors associated with time free of oral candidiasis in children infected by HIV;reference center in São Paulo,Brazil, 1985-2009.** Most recent regimen before appearance of oral candidiasis in cases of occurrence of disease or last regimen received throughout study for cases in which disease did not occur.
* Up to three months before or three months after appearance of oral candidiasis in cases of occurrence of disease or at end of study for cases in which disease did not occur; ** Up to three months before appearance of oral candidiasis in cases of occurrence of disease or throughout study for cases in which disease did not occur; *