Predictors of Candidemia infections and its associated risk of mortality among adult and pediatric cancer patients: A retrospective study in Lahore, Punjab, Pakistan

Methods: A retrospective study was conducted at a tertiary care cancer hospital in Lahore, Pakistan. The data was collected from the medical records of all the patients who were found positive for Candida species between 1st January 2017 and 31st June 2017. Data were analyzed by using Statistical Packages for Social Sciences (IBM SPSS Statistics for Windows, Version 21.0. Armonk, NY: IBM Corp.) and Microsoft Excel (MS Offi ce 2010).


Introduction
The bloodstream infections (BSI) caused by Fungi are one of the major factors responsible for premature death of hospitalized cancer sufferers. Among all the Fungi, the causative agents for most of the nosocomial infections are the members of genus Candida. These microorganisms have worsen the situation by causing an increment up to 50% in mortality rate of cancer patients [1]. It is evident from the previously published studies that candidemia accounts for 10-50% of the mortalities among hospitalized cancer patients [2,3]. Besides mortality rate, these infections not only lengthen the stay in hospitals and intensive care units (ICU) but also make the treatment expensive [4].
The blood culture test predicts that Candida albican is responsible for fungal infections in most of the immune competent patients. Also species other than albican are responsible for causing infections among cancer patients [5]. Although these infections are prevailing in many low and middle income countries (LMIC). But limited data regarding epidemiology of BSI caused by Candida have been published till date. Particularly cancer patients have fall a pray into these infections because of malnutrition, late diagnosis and inadequate strategies of controlling these infections [6]. The number of cancer patients is continuously increasing in many LMIC, so candidemia is assumed to be the prime factor responsible for premature morbidity and mortality among these patients. Therefore, the present study aimed to evaluate the predictors and outcomes of candidemia among hospitalized cancer patients in Pakistan.

Study design and setting
A retrospective study was conducted in a cancer specialty named as Shaukat Khanum Memorial Cancer Hospital & Research Centre (SKMCH&RC), Lahore, Pakistan. Data were collected between 1 st January 2017 and 31 st June 2017. The selected setting was a non-pro it organization providing referral base cancer specialty to the patients from all over the Pakistan. The characteristics of selected study setting have been summarized in Table 1 [7]. The SKMCH&RC is a well-equipped hospital with electronic database. All the medical records of patients can be assessed online by using International Classi ication of Diseases versions 9 or 10 (ICD-9 / ICD-10) codes or keywords.

Study population and sample size
The records of 6,397 cancer patients were securitized. The patients were included in the study if their blood culture tests were found positive for any microorganism, were not on the last stage of malignancy, had not been found to be the victim of any infectious disease like acquired immunode iciency syndrome (AIDS) and not administered with immunosuppresants. The records of 927 patients were ful illed our inclusion criteria and consequently selected for the study.

Data collection
A data collection tool was used to collect the data. The questionnaire was structured and consisted of 4 parts: 1) characteristics of the patients, 2) diagnosis detail (detection of candida specie and the type of cancer), 3) therapeutics detail (use of total parenteral nutrition (TPN), dose and type of administered steroids, antibiotics and chemotherapeutic agents in the preceding three months, and use of central venous catheter, 4) health status of patients (neutropenia status, and outcomes of hospitalization). If the same patient hospitalized due to candida infections at different time intervals then each occasion was treated as a separate episode of hospitalization. The time period was divided into three categories. First category is from January to February, second is from March to April and third is from May to June. The blood culture test was performed on each patient who were suspected for candidemia and bacteremia. The conventional manual blood culture system was used for processing the blood samples and the reporting was done according to standard microbiological techniques. If the culture was found positive for yeast then it was further investigated through the germtube test and testi ied as candida albican or candida non-albican. Only two antifungal agents Amphotericin B deoxycholate and luconazole were available during the study period.

Data analysis
The characteristics of the study population were summarized using descriptive statistics. Chi-square, Fisher's exact tests and multivariate logistic regression analyses were used to compare categorical variables and to test the association between characteristics and outcomes of mortality. The inal model for estimation of adjusted odds ratios (AOR) was developed through forward selection. All statistical tests were 2-sided, with a type 1 error level of 0.05. Data were analyzed by using Statistical Packages for Social Sciences (IBM SPSS Statistics for Windows, Version 21.0. Armonk, NY: IBM Corp.) and Microsoft Excel (MS Of ice 2010).

Defi nitions
Bacteremia: If the bacteria was identi ied in blood during culture test.
Candidemia: If the candida specie was identi ied in blood during blood culture test.

Shock:
The state of medical emergency when systolic blood pressure (BP) was <90mmHg, diastolic BP was <60mmHg and / or patient was administered with ionotropic agents to maintain BP.

Duration of stay:
The time period from the irst day when blood culture test was found to be positive for candida specie till the day of discharge or death.
Index hospitalization: When a patient was irst time admitted to the hospital.

Ethics approval
Ethical approval was obtained from the Medical Research Ethics Committee (MREC) of Nishtar Medical College, Multan (Reference: 10-2016/REC, dated October 22, 2016). The permission to conduct this study was also obtained from the administrators of the SKMCH&RC.

Results
Among all the cases a total of 135 blood cultures were found positive for candida species. Overall, 927 cases of cancer patients were found positive for any microorganisms while 2.1% (n=19) cases were found positive for candida specie. In every 100 cultures which were positive for any microorganism there were 2 cultures found to be positive for candida specie. The characteristics of study participants are summarized in Table 2.
No signi icant difference was observed across the three time periods. The type of cancers were categorized as solid or hematological for patients for whom data was available. Also, a decrease in proportion of Candida albican positive cultures was observed with the passage of time while the proportion of C.tropicalis positive cultures were reported. Although the proportion of C.albican positive culture was found to be constant across both solid and hematological cancers, but C.tropicalis was signi icantly more likely to be positive in hematological malignancies Table 3.
The investigation about treatment of patients and its outcomes were also made during hospitalization. Antifungal agents were administered to majority of the patients. Also, the mortality rate was signi icantly high during index hospitalization Table 4. The candidemia associated mortality rate was found to be higher in patients with hematological malignancies (AOR: 2.1; 95% con idence interval [CI] 1.0-4.3) as compared to patients suffering from solid tumors. The mortality of patients positive for C.albican was signi icantly less (AOR 0.5; 95% CI = 0.1-0.7) as compared to those who were found positive for non-albican specie or a mixture of non-albican and albican species. The cancer patients who suffered from shock were at higher risk of candidemia associated mortality (AOR: 9.1; 95% CI = 4.1-19.3) as compared to those who were not in the state of shock. The patients who were not suffered from neutropenia were at signi icantly lower risk of candidemia associated mortality (AOR:     0.6; 95% CI = 0.1-1.1) as compared to neutropenic cancer patients. Furthermore, those patients who were administered with antifungal upon diagnosis were signi icantly at lower risk of dying with candidemia (AOR: 0.2; 95% CI = 0.1-0.6) as compared to those who did not administer an antifungal agent Table 5. The empiric therapy of antifungal agents before a positive blood culture was not signi icantly associated with higher rate of mortality (p<0.05).

Discussion
The present study aimed to retrospectively analyze the medical records of cancer patients where the blood culture test was found positive for candida species among adult and pediatric populations. Findings revealed that per 100 cultures positive for any microorganism there were 2 cultures positive for Candida specie. Also, the disease burden of candidemia was not persistent throughout the study period. This is because C.albican was not only the common causative agent for candidemia. But in many cases non-albican species especially C.tropicalis was found positive for patients suffering from hematological malignancies. Among all the hospitalized cancer patients, the mortality rate of patients admitted in hospital for the irst time was found in a range of 46%-57%. The study predicted the statistically signi icant association of factors with the higher rate of mortality and these include malignancy of hematological origin, positive blood culture test for non-albican species, absolute neutropenia (neutrophil count < 500), and not receiving timely antifungal agents.
Our indings revealed the outcome of adults and pediatric cancer patients suffering from hematological and solid tumors. Previously published studies in Pakistan were conducted on pediatric cancer patients suffering from [8], non-cancer patients [9][10][11] and immunocompromised patients who underwent stem cell transplantation [12]. The result of present study revealed that C.albican associated infections were found to be signi icantly less common than that of non-albican species especially C.tropicalis. From January to June most of the blood culture tests were found positive for C.tropicalis. Similar results have been reported in other studies among cancer patients [13,14]. This might be because of the fact that luconazole was administered prophylactically to these patients especially sufferers of hematological malignancies. The ef icacy of luconazole was more against C.albican as compared to non-albican species. The increased use of this drug could be the reason of decline in C.albican associated infections. Also, the prevalence of non-albican species e.g., C.tropicalis is more in tropical and subtropical countries like Pakistan [15]. Our indings revealed the mortality rate associated with candida species in between 46% and 57%. The high rate of mortality associated with candida species was also found by previously published systematic reviews of matched case-control studies. Likewise, other studies also predicted the mortality rate of cancer patients due to candida species between 50% and 61% [1,3]. But a Portuguese study found a signi icantly lower rate of candida associated mortality among cancer patients i.e., 31.9% [13].
The results of present study found non-albican species, shock, neutropenia and hematological malignancies as the signi icantly associated factors responsible for higher rate of mortality among cancer patients. Similarly, our this inding C.tropicalis was the most commonly found organisms in the blood samples of cancer patients is in line with the results of the study reported by Pfaller et al., [16]. Other studies also found hematological malignancies and neutropenia as the signi icantly associated risk factors for higher rate of mortality [14,17].
The multivariate analysis revealed that antifungal agents should be used when blood culture test found positive. In this study most of the patients with candidemia were administered with amphotericin B while some of the patients either died before the availability of blood culture test report or received luconazole for C.albican. Most of the patients received amphotericin B, after the diagnosis of candidemia. Hence, it is recommended that all the cancer patients must be screened for candidemia. Also the prompt use of antifungal agents in these populations must be ensured to decrease the risk of mortality associated with candidemia.

Strengths and limitations
As there is limited availability of previously published literature on the risk factors of candida associated mortalities among cancer patients especially in LMICs like Pakistan, so our results will be bene icial in understanding the epidemiology, management and outcome of candidemia in cancer patients from this part of the world. Antifungal agents are not used commonly because of their out of pocket cost but our indings suggest bene its of antifungal agents among cancer patients suffering from candidemia. So, this study emphasize the need of availability of antifungal drugs at lower cost. Like other studies it also has some limitations. First the data were uneven and retrospective in nature. The investigators stratify by time because of unevenness of data. Second, in some cases data related to candida specie were missing. This is because of the fact that blood culture test reports were not available during the study period. . But the investigators did not ind any systematic differences across time periods because sample size and the length of study period is relatively short. Third, the investigators did not evaluate the mortality attributable to candidemia, rather all-cause mortality were evaluated. Fourth, the results cannot be generalized to all over the country because the study was conducted at a single healthcare setting. Fifth, the use of antifungal agents as a prophylaxis therapy was also not assessed. Lastly, as the minimum inhibitory concentrations and antifungal susceptibility were not assessed, so the resistance pattern of candida species against antifungal agents can't be evaluated.

Conclusion
The blood culture tests of cancer patients were found to be positive for different species candida and bacteria simultaneously. There was no statistically signi icant risk factor found to be associated with candidemia. But the possible predictors during index hospitalization were hematological malignancies, shock and non-albican species of Candida.