Antifungal prophylaxis in liver transplant patients: A systematic review and meta-analysis

Authors


  • Presented in part at the 44th Interscience Conference on Antimicrobial Agents and Chemotherapy, Washington, DC, October 30-November 2, 2004.

Abstract

We performed a meta-analysis to determine whether antifungal prophylaxis decreases infectious morbidity and mortality in liver transplant patients. We searched for randomized trials dealing with prophylaxis with systemic antifungal agents. We used a fixed effect model, with risk ratio (RR) and 95% confidence interval (CI); we assessed study quality for heterogeneity and publication bias. Six studies (5 double-blind), for a total of 698 patients, compared fluconazole, itraconazole, or liposomal amphotericin to placebo (5 studies) or oral nystatin. Prophylaxis reduced colonization (RR, 0.45; CI, 0.37-0.55), total proven fungal infections (RR, 0.31; CI, 0.21-0.46), which included both superficial (RR, 0.27; CI, 0.16-0.45) and invasive (RR, 0.33; CI, 0.18-0.59) infections, and mortality attributable to fungal infection (RR, 0.30; CI, 0.12-0.75). Prophylaxis did not affect overall mortality (RR, 1.06; CI, 0.69-1.64) or empiric treatment for suspected fungal infection (RR, 0.80; CI, 0.39-1.67). The beneficial effect of antifungal prophylaxis was predominantly associated with the reduction of Candida albicans infection and mortality attributable to C. albicans. Compared to controls, however, patients receiving prophylaxis experienced a higher proportion of episodes of non–albicans Candida, and in particular of C. glabrata. No beneficial effect on invasive Aspergillus infection was observed. In conclusion, our analysis shows a clear, though limited, beneficial effect of antifungal prophylaxis in liver transplant patients. Concerns about the selection of triazole-resistant Candida strains, however, are realistic, and the potential disadvantages of prophylaxis should be weighed against the established benefits. Liver Transpl 12:850–858, 2006. © 2006 AASLD.

Systemic fungal infections are a significant cause of morbidity and mortality in solid-organ recipients.1–3 The incidence of fungal infections after liver transplantation has been reported in the range of 7-42%, with Candida spp. and Aspergillus spp. as the most common pathogens responsible for infections.2–16 Actually, Candida spp. accounts for 35-91% of all invasive fungal infections in liver transplant recipients, followed by Aspergillus spp., responsible for 9-34% of infections.3, 5, 6, 8, 9, 17, 18

Invasive fungal infections after liver transplantation have been associated with overwhelming outcome, with attributable mortality rates reported as high as 92-100% for invasive aspergillosis and 70% for invasive candidiasis.9, 14, 16, 18–22

Use of broad-spectrum antibiotic therapy, technical difficulties of surgical procedures, return to surgery and retransplantation, cytomegalovirus disease, absence of the protective effect of anaerobic bacteria toward the overgrowth of Candida in the gut, and overall severity of illness in the patient (e.g., prolonged period of dialysis, access to intensive care unit, use of central venous catheter, total parenteral nutrition, and mechanical ventilation) are major risk factors for the development of invasive fungal infections.6, 16, 23–25 Among the other risk factors is immunosuppression, which may be induced by major surgery, bacterial sepsis, diabetes, steroids, chemotherapy, and immunosuppressive treatment after transplantation. As yet, however, there are few data showing a correlation between measure of immune system function (e.g., ratios of T-helper to T-suppressor lymphocyte) and risk of fungal infection in liver transplant patients.20

Due to the high incidence and mortality rate of invasive fungal infections, the use of a successful antifungal prophylaxis in liver transplant patients is very attractive, and there have been several studies pertaining to prophylaxis in this field.26–41 These studies have been recently analyzed in a Cochrane review on antifungal agents for preventing fungal infections in solid-organ (heart, liver and kidney) transplant recipients.42 We have addressed some questions that were not considered in this review (e.g., mortality attributable to fungal infections, need for empiric antifungal treatment) and restricted the analysis to liver transplant patients, with their distinctive predisposition to develop fungal infections. Moreover, to evaluate the effect of systemic antifungal prophylaxis unaffected by the choice of control regimen, we chose to include only trials controlled with placebo, no prophylaxis, or minimal prophylaxis (oral nonabsorbable agents).

The aim of this study was to systemically identify and summarize the quality of the randomized trials available and the effects of antifungal prophylaxis in liver transplant patients.

Abbreviations

CI, confidence interval; RR, risk ratio; ARR, absolute risk reduction; df, degree of freedom.

MATERIALS AND METHODS

Search Strategies

The search was carried out on MEDLINE (1966-March 2004), EMBASE (1980-March 2004), and the Cochrane Database of Systematic Reviews (Issue 1, 2004). MeSH terms used were “antifungal agents/antifungal prophylaxis & liver transplantation/organ transplant.” The computer search was supplemented by consulting the bibliographies from the articles retrieved.

We included only randomized controlled trials evaluating the efficacy of antifungal prophylaxis in liver transplant patients. We required that studies compared prophylactic regimens based on systemic antifungal agent to a control arm in which subjects were given placebo, no treatment, or minimal treatment with oral nonabsorbable antifungal agents (e.g., nystatin).

Outcome Measures

We extracted data on overall mortality, mortality attributable to fungal infection, overall fungal infections (including defined and presumed fungal infections), superficial and invasive infections, fungal colonization, need for empiric antifungal treatment, adverse reactions to study drugs, and need for discontinuation of treatment. An invasive infection was defined as the histopathologic evidence of infection, or microbiologic evidence of fungi in tissue culture, or yeast from normally sterile body cavity or organ. A superficial infection was defined as isolation of a fungus from skin, oropharynx, vagina, gastrointestinal tract, or urine in association with symptoms and signs of inflammation, ulceration, plaques, or exudates. Where possible, we also extracted time-to-event data.

Quality Assessment

We assessed the methodology of each trial with a scale developed by Jadad and colleagues that scores (from a low of 0 to a high of 5) the randomization, double-blinding, and reports of dropouts and withdrawals.43 Each trial was independently scored by 2 of us and any areas of disagreement arbitrated by a third person.

Statistical Analysis

A conventional meta-analysis was performed with use of the Mantel-Haenszel fixed-effects model, applying the DerSimonian and Laird random effects model only in cases where the heterogeneity test give a P value <0.1.44, 45 We calculated both the study-specific and the common 95% confidence interval (CI) by the method of Woolf.46 We used risk ratio (RR) as measure of the effect size, and the procedure to combine the 2 × 2 tables was the Mantel-Haenszel-like method by Greenland and Robins.47, 48 Measures of efficacy that we have used were the absolute risk reduction (ARR, defined as the difference in the event rate in intervention group and in control group), the relative risk reduction (defined as the ARR divided by the event rate in control group), and the number of patients needed to be treated to prevent 1 event (expressed as the reciprocal of the absolute risk reduction) and related 95% CI.49, 50 Sensitivity analysis was performed for determining if quantitative results differed with the exclusion of individual studies.

Assessment of Publication Bias and Heterogeneity

Graphical funnel plots were generated to visually inspect for publication bias.51 The statistical methods for detecting funnel plot asymmetry were the rank correlation tests of Begg and Mazumdar and the regression asymmetry test of Egger et al.51, 52 The heterogeneity of study results was assessed by Cochran Q test and by a test of inconsistency (I2).53, 54

RESULTS

As shown in the flow diagram (Fig. 1), 16 potentially relevant controlled clinical trials were identified as being appropriate for inclusion in our analysis.26–41 Of these, 6 studies were excluded because there was evidence of duplication of data.35–40 Data was not usable from a study published as an abstract.41 Three other trials were excluded because they tested other interventions or because the type of outcome measure was different from those specified in our protocol; these studies included a randomized study by Ruskin et al. comparing oral nonabsorbable agents (nystatin and clotrimazole), a randomized study by Winston et al. comparing 2 systemic antifungal agents (itraconazole and fluconazole), and a study by Tortorano et al. aimed at evaluating only the efficacy of fluconazole compared to oral amphotericin B in preventing fungal colonization.32–34 Therefore, we included in the meta-analysis data retrieved from 6 randomized clinical trials (5 double-blind) for a total of 698 patients.26–31

Figure 1.

Meta-analysis profile summarizing trial flow. RCT, randomized clinical trial.

Description of Studies and Quality Assessment

Table 1 summarizes the main characteristics of included studies. Duration of antifungal prophylaxis varied in the different studies, ranging from 5 days in the study by Tollemar et al.26 to 4-10 weeks in the remaining trials. The diagnostic criteria for fungal infection were fully consistent with our definition in all study but 1; in the definition of fungal infection the study by Biancofiore et al. included patients with positive cultures from multiple (≥3) peripheral sites.30

Table 1. Main Characteristics of Randomized Clinical Studies Included in the Analysis
 First Author, Year of Publication (reference).
Tollemar, 199526Lumbreras, 199627Meyers, 199728Winston, 199929Biancofiore, 200230Sharpe, 200331
Intervention (number of patients evaluated)Liposomal AmB, 1mg/kg (40) vs. placebo (37); duration: 5 days. Double-blindFluconazole, 100 mg po (76) vs. nystatin (67); duration: 4 weeks. UnblindFluconazole (23) vs. placebo (24) and, in both groups, oral clotrimazole and nystatin vaginal suppositories; duration: 10 weeks. Double-blind. Fluconazole dose not specified.Fluconazole 400 mg, IV then po (108) vs. Placebo (104); duration: 10 weeks.Double-blindLiposomal AmB, 1mg/kg/d for 7 d, then itraconazole, 200 mg po (42) vs. fluconazole 400 mg IV for 7 d, then itraconazole, 200mg po (43) vs. placebo, IV and po (44); duration: 4 weeks. Double-blindItraconazole 2.5-5 mg/kg/d, po (25) vs. placebo (37); duration: 8 weeks. Double-blind
Number of randomized participants (age in treated patients is compared to age in controls)85 liver transplant patients, adults and children. Median age, (range): 40 (1-63) yr vs. 42 (1-67) yr. Males: 50%. Fungal colonization at entry: 64 vs. 70%.143 liver transplant patients, adults and children. Age: 40 ± 16 yr vs. 42 ± 16 yr Males: 60%. Fungal colonization at entry: 20.7 vs. 17.9%. Child-Turcotte-Pugh C score: 17 vs. 15%.55 liver transplant patients. Mean age: 49.6 vs. 48.2 yr. Males: 63%.236 liver transplant patients, adults. Median age (range):49 (15-75) vs. 53 (19-74) yr. Males: 55%. Fungal colonization at entry: 60 vs. 70%.131 liver transplant patients. Age (mean):46.2 vs. 50.3 vs. 51.5 yr. Males: 63%. Fungal colonization at entry: 42.6%. Child-Turcotte-Pugh C score: 7.2 vs. 154 vs. 4.6%, respectively.71 liver transplant pts, adults. Age (mean): 46 yrMales: 59%. Fungal colonization at entry: 44%. Child-Turcotte-Pugh C score: 65%.
SettingSweden and Finland, 2 centersSpain, 3 centersUnited States, 1 centerUnited States, 1 centerItaly, 1 centerCanada, 1 center
OutcomesMortality (overall and attributable to fungal infection), invasive and superficial infection, empiric antifungal treatment, side effects and discontinuation because of side effectsMortality (overall and attributable to fungal infection), invasive and superficial infection, colonization, empiric antifungal treatment, side effects and discontinuattion because of side effectsMortality (overall and for fungal infection), invasive and superficial infection, empiric antifungal treatment, discontinuation because of side effectsMortality (overall and attributable to fungal infection), invasive and superficial infection, colonization, empiric antifungal treatment, side effects and discontinuation because of side effectsMortality (overall and attributable to fungal infection), invasive and superficial infections, colonization, side effects and discontinuation because of side effectsMortality (overall and attributable to fungal infection), invasive and superficial infection, empiric antifungal treatment, side effects and discontinuation because of side effects

The most common underlying disease requiring liver transplantation was hepatic cirrhosis (mostly postviral and alcoholic liver disease), while a lower proportion of patients had liver transplantation following fulminant hepatic failure and hepatocellular carcinoma. All the patients received immunosuppressive treatment (prednisone, azathioprine, cyclosporine or tacrolimus, and OKT3 or antithymocyte globulin for rejection). Fungal colonization at entry varied in the different studies (Table 1).

Study drugs were fluconazole, itraconazole, liposomal Amphotericin B, and liposomal Amphotericin B followed by itraconazole or fluconazole. These drugs were compared to placebo (5 studies) or to nystatin (1 study).

The median Jadad score for quality was 3.5 (range, 2-5); the mean score was 3.7 ± 1.2. Two studies clearly reported concealment of treatment allocation.26, 31

Results of the Meta-analysis

Table 2 shows RR and related 95% CI, crude rates, absolute and relative risk reduction, and the number of patients needed to be treated to prevent 1 event. Figures 2-4 {FIG 2-4} show RR and related 95% confidence interval (95%CI) for individual studies for the most relevant outcomes. Visual inspection of Figures shows a statistically significant reduction of total episodes of fungal infection and superficial and invasive infection, as well as mortality attributable to fungal infections in patients receiving prophylaxis with systemic antifungal agents. Prophylaxis, however, did not affect overall mortality (RR, 1.06; 95% CI, 0.69-1.64; P = 0.77) and need for empirical antifungal treatment (RR, 0.80; 95% CI, 0.39-1.67; P = 0.55).

Table 2. Main Cumulative Meta-Analysis Data
OutcomeRisk Ratio* (95% CI)P ValueCrude RatesARR*-RRR (%)NNT* (95% CI)
   TreatedControls  
  • Abbreviations: RRR, relative risk reduction; NNT, number of patients needed to be treated to prevent 1 event; ARR, absolute risk reduction.

  • *

    Risk Ratio, ARR, and NNT are weighted estimates.

  • RRR-ARR/crude rate in controls.

Total fungal infections0.31 (0.21-0.46)<0.000133/365 (9.0)88/314 (28.0)19.4-69.25.1 (3.9-7.2)
Invasive infections0.33 (0.18-0.59)<0.000115/365 (4.1)39/320 (12.1)8.4-69.411.8 (8.0-23.0)
Superficial infections0.27 (0.16-0.45)<0.000119/365 (5.2)60/314 (19.1)13.9-72.77.1 (5.3-10.9)
Empiric treatment for suspected fungal infections0.80 (0.39-1.67)0.5513/280 (4.6)16/270 (5.9)1.1-18.688.9 (21.1-40.2)
Side effects1.38 (1.04-1.83)0.0283/342 (24.2)57/290 (19.6)−7.3/−37.2−13.6 (−87.1/−7.3)
Mortality attributable to fungal infections0.30 (0.12-0.75)0.0106/365 (1.6)17/316 (5.3)3.8-71.625.7 (14.7-100.7)
Figure 2.

Pooled RR estimates and their 95% CI for the outcomes invasive and superficial fungal infections. Studies are identified by first author. Size of squares is proportional to the weighted RR. *Cannot be computed because the presence of frequencies = 0.

Figure 3.

Pooled RR estimates and their 95% CI for the overall outcomes for mortality and mortality attributable to fungal infections. Studies are identified by first author. Size of squares is proportional to the weighted RR. *Cannot be computed because the presence of frequencies = 0.

Figure 4.

Pooled RR estimates and their 95% CI for the outcomes total fungal infections (superficial and invasive) and need for empirical antifungal treatment. Studies are identified by first author. Size of squares is proportional to the weighted RR. *Cannot be computed because the presence of frequencies = 0.

Side effects assessed as being related, or possibly related, to drug were more common in antifungal prophylaxis recipients (RR, 1.38; 95% CI, 1.04-1.83; P = 0.02), but discontinuation of treatment due to side effects did not differ between study groups and controls (RR, 1.15; 95% CI, 0.65-2.04; P = 0.63). In the large majority of cases, adverse events consisted of mild gastrointestinal intolerance in fluconazole and itraconazole recipients, and neurologic events (headache, seizures, tremors) in fluconazole recipients; 3 patients receiving liposomal amphotericin B had back pain, 1 had a transient episode of thrombocytopenia, and 1 suffered from suspected nephrotoxicity. Abnormalities in laboratory tests had similar rates of occurrence in treatment groups and controls. One study reported elevated cyclosporine levels in fluconazole recipients.29 Fungal colonization rates were reported in 3 studies.27, 29, 30 Prophylaxis significantly reduced colonization (RR, 0.45; CI, 0.37-0.55; P < 0.0001).

Rates of patients free from fungal infection at a specific time were reported in 2 studies.29, 31 The study by Winston et al. showed that starting from day 10 rates of proven fungal infection became significantly lower in treated patients compared to controls, both in all patients and in high-risk patients.29 Likewise, in the study by Sharpe et al. the probability of developing a fungal infection requiring systemic antifungal therapy was significantly lower in treated patients beyond the first 2 weeks of prophylaxis.31

In sensitivity analysis, the exclusion of any single study yielded only minimal change on the effect size for all the outcomes analyzed except for side effects. Actually, the difference in the occurrence of side effects was no longer statistically significant after exclusion of 2 studies with fluconazole.27, 29

Data on fungal isolates according to treatment groups were provided in 5 studies and are summarized in Table 3.26, 27, 29–31C. albicans was by far the most common isolate in control groups, responsible for 64% of deaths attributable to fungal infections. Patients receiving prophylaxis had a higher rate of infections sustained by other species of Candida, including C. glabrata infections, and a higher rate of death due to Aspergillus spp. Invasive aspergillosis was responsible for 3 of the 6 deaths attributable to fungal infection in treated groups, in 2 cases in patients receiving fluconazole and in 1 case in a patient receiving intravenous fluconazole combined with oral itraconazole.27, 29, 30

Table 3. Isolates From 687 Liver Transplant Patients (370 treated, 317 controls) With Defined Fungal Infections
Species (% of total isolates)From Patients With Proven Infections (invasive & superficial)From Patients Who Died From Fungal Infections
OverallTreatmentControlOverallTreatmentControl
  • *

    Including C. krusei, C. pelliculosa, C. tropicalis, C. parapsilosis, and C. pseudotropicalis.

C. albicans64 (55)11 (36)53 (61)12 (52)1 (16)11 (64)
C. glabrata20 (17)8 (26)12 (14)3 (13)1 (16)2 (11)
Other Candida species*25 (21)9 (30)16 (18)1 (4)1 (16)-
Aspergillus spp.6 (5)2 (6)4 (4)6 (26)3 (50)3 (17)
Other1 (0.8)-1 (1)1 (4)-1 (5)
Total (100%)116308623617

Heterogeneity and Publication Bias Assessment

There was not evidence of intertrial heterogeneity for the outcomes analyzed. The Cochrane Q statistic for heterogeneity provided the following results: total fungal infections, Q = 8.97, (degree of freedom) = 5, P = 0.11; invasive infections, Q = 4.43, df = 4, P = 0.35; superficial infections, Q = 5.47, df = 4, P = 0.24; colonization, Q = 1.87, df = 2, P = 0.9; empirical treatment for suspected fungal infections, Q = 2.39, df = 4, P = 0.66; side effects, Q = 1.3, df = 4, P = 0.86; discontinuation of treatment, Q = 1.73, df = 3, P = 0.63; overall mortality, Q = 4.98, df = 5, P = 0.41; attributable mortality, Q = 2.66, df = 3, P = 0.4. There was also no evidence for publication bias with funnel plots (symmetrical appearance) as well as with the Begg and Mazudmar and Egger tests. The Begg and Mazudmar test gave a P value = 1 for the outcomes total fungal infections, invasive infections, superficial infections, colonization, and attributable mortality, and a P value = 0.806 for side effects.

DISCUSSION

In this meta-analysis, we have included data from 6 randomized clinical trials (5 double-blind, placebo-controlled) evaluating the effect of antifungal prophylaxis in 698 patients undergoing orthotopic liver transplantation. A Cochrane review on antifungal agents for preventing fungal infections in solid-organ transplant recipients was recently published by Playford et al.42 We have addressed some questions that were not considered in this review (e.g., mortality attributable to fungal infections, need for empiric antifungal treatment). Moreover, to evaluate the effect of systemic antifungal prophylaxis unconfounded by the choice of control regimen, we chose to include only trials controlled with placebo, no prophylaxis, or minimal prophylaxis (oral nonabsorbable agents). In this regard, the present study provides more firm results than those available from the Cochrane review, which included trials with a variety of control regimens. Moreover, contrary to the Cochrane review, we have not included in our analysis studies dealing with renal and cardiac transplant recipients, who are at lower risk of invasive fungal infections compared to liver transplant recipients.

The results of our meta-analysis show that antifungal prophylaxis had a significant effect on infectious morbidity and on mortality attributable to fungal infections, but not on overall mortality. Compared to controls, patients who received prophylaxis experienced a reduction of overall episodes of fungal infections (69.2% relative risk reduction), as a result of a decrease in invasive infections (69.4% relative risk reduction) and in superficial infections (72.7% relative risk reduction). There was also evidence for a decrease in mortality attributable to fungal infections in patients receiving prophylaxis, with a 71.6% relative risk reduction. By contrast, empiric treatment for suspected fungal infections and overall mortality were not affected by antifungal prophylaxis. Side effects were more commonly reported in recipients of study drugs compared to controls, but discontinuation of treatment did not differ between groups.

The number of patients needed to be treated to prevent 1 event is another established method to express the magnitude of the effect of an intervention on the outcomes considered. In the present meta-analysis, the numbers of patients needed to be treated to prevent 1 event were as follow: for total fungal infections, 5.1; for colonization, 2.5; for invasive infections, 11.8; for superficial infections, 7.1; for empiric treatment of suspected fungal infections, 88.9; for overall mortality, -149; for mortality attributable to fungal infections, 25.7; for side effects, -13.6.

Time to event data was available from 2 studies.29, 31 Data from these trials shows that the probability of developing a fungal infection requiring systemic antifungal therapy was significantly lower in treated patients beyond the first 10-14 days of prophylaxis. To this end, it is relevant to emphasize that in the study by Tollemar et al., prophylaxis was given for only 5 days.26

Some limitations of this meta-analysis need to be acknowledged. First of all, the diagnosis of systemic fungal infections is difficult and open to bias that may affect the evaluation of end points. Moreover, as with all meta-analyses, our conclusions can be only as accurate as the trials upon which they are based. Based on quality scores, the methodological quality of the studies was, on average, more than satisfactory, though it varied considerably in individual studies. There was, however, variability in the design and reporting of individual studies. Study drugs were oral/intravenous fluconazole, 100-400 mg; oral itraconazole, 200 mg or 2.5-5 mg/kg; and liposomal amphotericin B, 1 mg/kg. In 1 study, liposomal amphotericin B was followed by oral itraconazole and compared to intravenous fluconazole followed by oral itraconazole, or to placebo. Duration of prophylaxis varied from 5 days to 10 weeks. All the included studies reported overall mortality, mortality attributable to fungal infections, proven fungal infections, and need for discontinuation of treatment; other outcomes, however, were not uniformly reported in the trials analyzed.

Publication bias and heterogeneity represent significant threats to the validity of a meta-analysis. In the present meta-analysis, evidence of publication bias with graphical and statistical methods was not detectable for any of the outcomes analyzed. Moreover, there was consistency in treatment effects, and no evidence of heterogeneity was detected. This was also partially confirmed in sensitivity analysis. For all the outcomes except for side effects, the exclusion of any single study yielded only minimal change on the effect size.

Overall, the cumulative data from the available randomized studies suggest that among patients undergoing orthotopic liver transplantation, prophylaxis has a beneficial effect on morbidity and mortality attributable to fungal infections, but not on overall mortality. The beneficial effect of antifungal prophylaxis was predominantly associated with the reduction of C. albicans infection and mortality attributable to C. albicans. A reduction of C. albicans infections can be adequately attained with azole prophylaxis; to this end, fluconazole seems to be a suitable agent. Since the majority of Candida infections seem to be acquired in the first 2-6 weeks after transplantation, duration of prophylaxis should be limited, as already suggested by Winston et al., to 6 weeks.29, 31 On the other hand, there is growing evidence showing an important role of fluconazole in a shift toward non–C. albicans species, and several investigators have noted increases in the frequency of C. glabrata isolation in conjunction with the selective pressure exerted by azole use.55–59 This trend was also evident in the trials included in the analysis: Compared to controls, patients receiving prophylaxis experienced a higher proportion of episodes of non-albicans Candida (56% vs. 33% of total proven infections), namely of C. glabrata infections (26% vs. 14%).

In contrast to Candida infections, invasive aspergillosis has a relatively low incidence and is more common in the late posttransplantation period.1, 9, 19, 22 Actually, 55% of these infections now occur after the first 3 months following transplantation, making the implementation of an effective, safe, and cost-effective prophylaxis problematic.9 Therefore, a pre-emptive approach in high-risk patients (e.g., those with acute liver failure) or an empirical treatment for suspected infections with agents that have anti-Aspergillus activity seems more rational.1

The absence of evidence of either heterogeneity among the studies or of any publication bias suggests that the conclusions we have drawn are reasonably generalizable and robust.

In conclusion, our analysis suggests that prophylaxis of fungal infections among liver transplant recipients has beneficial effect on morbidity and mortality attributable to C. albicans. However, the selection of triazole-resistant Candida strains is of concern and needs to be carefully addressed in future trials. These studies should include epidemiological data on the development of resistance over time.

Ancillary