Abstract
Although echinocandins are generally well tolerated, there is little information on the frequency with which renal and hepatic adverse effects occur during use of micafungin or other parenteral antifungal (PAF) agents in clinical practice.
MYCOS is a multicentre cohort study of adult and paediatric patients who received micafungin or other PAFs between 2005 and 2012 at seven tertiary care hospitals from six centres in the USA. PAF cohort controls were selected through propensity score (PS) matching to micafungin recipients using clinical characteristics, other treatments, procedures and hospital service where PAF treatment was initiated. Analysis was restricted to patients without chronic liver and kidney conditions at the time of cohort entry. Treatment-emergent hepatic and renal injury was documented by changes in liver enzymes or estimated glomerular filtration rate through 30 days following completion of PAF treatment. Comparisons were quantified using the HR from a proportional hazards analysis.
There were 2970 micafungin recipients PS matched to 6726 recipients of comparator PAFs. Balance was achieved in all baseline covariates between treatment groups. There were similar rates of hepatic injury (micafungin, 13 events per 100 patients and other PAF, 12 per 100; HR = 0.99; 95% CI 0.86–1.14) and lower rates of renal injury (micafungin, 63 events per 100 patients and other PAF, 65 per 100; HR = 0.93; 95% CI 0.87–0.99) for micafungin recipients versus PAF comparators.
For a wide spectrum of underlying conditions, we observed no increase in liver injury by micafungin and possibly a reduced risk of renal dysfunction in comparison with other PAF medications.
Introduction
Echinocandins have a profile of lower renal toxicity compared with older antifungal agents, particularly amphotericin B, but they have shown hepatic toxicity in clinical trials and observational studies.1–4 Recent studies of recipients of micafungin estimated the risk of hepatic injury to be 12% in Japanese patients undergoing HSCT and a 6% risk in other patients.5 Most studies have included patients with existing liver conditions, making it difficult to isolate treatment-emergent liver injury.6 There is little information on the frequency of either renal or hepatic effects in routine care with micafungin in comparison to other parenteral antifungal (PAF) agents in other populations.7–9
Tumours observed in rat models led to a black box warning by the EMA for one echinocandin, micafungin.10,11 Human carcinogenicity has not been observed or investigated.
The reported analysis is embedded in a long-term multicentre pharmacoepidemiology cohort study to assess the risk of hepatocellular cancer. The analysis focuses on the 30 day risks of hepatic and renal injury among persons who received PAF medications, including micafungin, in the USA from 2005 to 2012 as part of the long-term pharmacoepidemiology study. Although liver injury was the focus of this report, we also studied the incidence of renal injury as a general drug safety marker of interest.
Methods
Study design
This multicentre cohort study is based on electronic medical records (EMR) and associated health data systems of seven tertiary care hospitals (six centres as two hospitals operate under the same umbrella organization) in the USA.
Ethics
The study received approval from the New England IRB and from the institutional review boards of the participating centres. The study was registered at ClinicalTrials.gov (NCT01686607) and at ENCePP.eu (2858), where the study protocol was deposited. This study was sponsored by Astellas Pharma Europe, the manufacturer of micafungin. The investigators had full control over the protocol development, data collection and analyses and publication.
Data sources
Cohort eligibility requirements, baseline covariates and 30 day outcomes were identified from the hospitals' inpatient and outpatient EMR and data systems. During the study period, US hospitals used the International Classification of Diseases, 9th revision (ICD-9), for recording diagnoses and the Current Procedure Terminology, 4th revision (CPT-4), for recording medical procedures in the structured portions of their databases. Various coding systems were used for recording medication dispensing.
The study used a rigorous approach to data extraction at each hospital, with: secure transfer of de-identified data; data cleaning and standardization according to a detailed data dictionary; data plausibility and validity checks; and double programming of all data manipulation and most analytical procedures. The pre-defined study protocol was further elaborated in a statistical and epidemiological analysis plan and a specification document that served as the instructions for the study analysts.
Patients
Patients were identified by their first PAF use during a hospitalization between 1 January 2005 and 31 December 2012. Patients were excluded if they had: earlier recorded PAF use; started use of micafungin and a different PAF on the same day; had pre-existing chronic hepatic or renal disease; had no ALT or AST test results recorded prior to PAF initiation or had elevated values; had no serum creatinine test result recorded; or had reduced estimated glomerular filtration rate (eGFR) values prior to PAF initiation (Table 1 for details).
Patient identification for the 30 day outcome analysis (first use)
| Exclusion criteria | Remaining patientsa | |
|---|---|---|
| Patients with cohort entry date between 1 May 2005 and 31 December 2012 | 41 427 | 100.0% |
| 1) Drop patients with multiple age or gender or hospital admission date is 60 days more than index hospital start | 41 154 | 99.3% |
| 2) Drop if any PAF use in past 6 months in participating hospital | 40 848 | 98.6% |
| 2b) Drop patients with only 1 day of PAF | 40 848 | 98.6% |
| 3) Drop patients initiated micafungin and a comparator additional PAF on the same day (‘dual PAF initiators’) | 40 500 | 97.8% |
| 4a) Drop patients with evidence of pre-existing chronic hepatic disease: as a discharge diagnosis at any hospitalization in the 183 days BEFORE the index hospitalization | 37 902 | 91.5% |
| 4b) Evidence of pre-existing chronic hepatic disease: as a visit diagnosis at any outpatient visit in the 183 days the BEFORE index hospitalization | 36 287 | 87.6% |
| 4c) Evidence of pre-existing chronic hepatic disease: as a discharge diagnosis DURING the index hospitalization | 32 045 | 77.4% |
| 5a) Evidence of pre-existing chronic kidney disease: as a discharge diagnosis at any hospitalization in the 183 days BEFORE the index hospitalization | 29 742 | 71.8% |
| 5b) Evidence of pre-existing chronic kidney disease: as a visit diagnosis at any outpatient visit in the 183 days BEFORE the index hospitalization | 29 089 | 70.2% |
| 5c) Evidence of pre-existing chronic kidney disease: as a discharge diagnosis DURING the index hospitalization | 25 929 | 62.6% |
| 6) Neither pre-treatment ALT nor pre-treatment AST results available: during 30 days before the index hospitalization but before PAF initiation | 23 948 | 57.8% |
| 7) Significantly elevated liver enzymes during 30 days before the index hospitalization but before PAF initiation: ALT >5× the hospital ULN (or >300 U/L) or AST >5× ULN (or >200 U/L) | 22 273 | 53.8% |
| 8) No pre-treatment creatinine results available: during 30 days before the index hospitalization but before PAF initiation | 22 257 | 53.7% |
| 9) Significantly reduced pre-treatment eGFRb: <30 mL/min | 17 690 | 42.7% |
| Exclusion criteria | Remaining patientsa | |
|---|---|---|
| Patients with cohort entry date between 1 May 2005 and 31 December 2012 | 41 427 | 100.0% |
| 1) Drop patients with multiple age or gender or hospital admission date is 60 days more than index hospital start | 41 154 | 99.3% |
| 2) Drop if any PAF use in past 6 months in participating hospital | 40 848 | 98.6% |
| 2b) Drop patients with only 1 day of PAF | 40 848 | 98.6% |
| 3) Drop patients initiated micafungin and a comparator additional PAF on the same day (‘dual PAF initiators’) | 40 500 | 97.8% |
| 4a) Drop patients with evidence of pre-existing chronic hepatic disease: as a discharge diagnosis at any hospitalization in the 183 days BEFORE the index hospitalization | 37 902 | 91.5% |
| 4b) Evidence of pre-existing chronic hepatic disease: as a visit diagnosis at any outpatient visit in the 183 days the BEFORE index hospitalization | 36 287 | 87.6% |
| 4c) Evidence of pre-existing chronic hepatic disease: as a discharge diagnosis DURING the index hospitalization | 32 045 | 77.4% |
| 5a) Evidence of pre-existing chronic kidney disease: as a discharge diagnosis at any hospitalization in the 183 days BEFORE the index hospitalization | 29 742 | 71.8% |
| 5b) Evidence of pre-existing chronic kidney disease: as a visit diagnosis at any outpatient visit in the 183 days BEFORE the index hospitalization | 29 089 | 70.2% |
| 5c) Evidence of pre-existing chronic kidney disease: as a discharge diagnosis DURING the index hospitalization | 25 929 | 62.6% |
| 6) Neither pre-treatment ALT nor pre-treatment AST results available: during 30 days before the index hospitalization but before PAF initiation | 23 948 | 57.8% |
| 7) Significantly elevated liver enzymes during 30 days before the index hospitalization but before PAF initiation: ALT >5× the hospital ULN (or >300 U/L) or AST >5× ULN (or >200 U/L) | 22 273 | 53.8% |
| 8) No pre-treatment creatinine results available: during 30 days before the index hospitalization but before PAF initiation | 22 257 | 53.7% |
| 9) Significantly reduced pre-treatment eGFRb: <30 mL/min | 17 690 | 42.7% |
aNumbers refer to individual patients. Few patients contributed twice, as first-use and as second-use PAF users.
beGFR was computed via the Mayo quadratic formula.32
Patient identification for the 30 day outcome analysis (first use)
| Exclusion criteria | Remaining patientsa | |
|---|---|---|
| Patients with cohort entry date between 1 May 2005 and 31 December 2012 | 41 427 | 100.0% |
| 1) Drop patients with multiple age or gender or hospital admission date is 60 days more than index hospital start | 41 154 | 99.3% |
| 2) Drop if any PAF use in past 6 months in participating hospital | 40 848 | 98.6% |
| 2b) Drop patients with only 1 day of PAF | 40 848 | 98.6% |
| 3) Drop patients initiated micafungin and a comparator additional PAF on the same day (‘dual PAF initiators’) | 40 500 | 97.8% |
| 4a) Drop patients with evidence of pre-existing chronic hepatic disease: as a discharge diagnosis at any hospitalization in the 183 days BEFORE the index hospitalization | 37 902 | 91.5% |
| 4b) Evidence of pre-existing chronic hepatic disease: as a visit diagnosis at any outpatient visit in the 183 days the BEFORE index hospitalization | 36 287 | 87.6% |
| 4c) Evidence of pre-existing chronic hepatic disease: as a discharge diagnosis DURING the index hospitalization | 32 045 | 77.4% |
| 5a) Evidence of pre-existing chronic kidney disease: as a discharge diagnosis at any hospitalization in the 183 days BEFORE the index hospitalization | 29 742 | 71.8% |
| 5b) Evidence of pre-existing chronic kidney disease: as a visit diagnosis at any outpatient visit in the 183 days BEFORE the index hospitalization | 29 089 | 70.2% |
| 5c) Evidence of pre-existing chronic kidney disease: as a discharge diagnosis DURING the index hospitalization | 25 929 | 62.6% |
| 6) Neither pre-treatment ALT nor pre-treatment AST results available: during 30 days before the index hospitalization but before PAF initiation | 23 948 | 57.8% |
| 7) Significantly elevated liver enzymes during 30 days before the index hospitalization but before PAF initiation: ALT >5× the hospital ULN (or >300 U/L) or AST >5× ULN (or >200 U/L) | 22 273 | 53.8% |
| 8) No pre-treatment creatinine results available: during 30 days before the index hospitalization but before PAF initiation | 22 257 | 53.7% |
| 9) Significantly reduced pre-treatment eGFRb: <30 mL/min | 17 690 | 42.7% |
| Exclusion criteria | Remaining patientsa | |
|---|---|---|
| Patients with cohort entry date between 1 May 2005 and 31 December 2012 | 41 427 | 100.0% |
| 1) Drop patients with multiple age or gender or hospital admission date is 60 days more than index hospital start | 41 154 | 99.3% |
| 2) Drop if any PAF use in past 6 months in participating hospital | 40 848 | 98.6% |
| 2b) Drop patients with only 1 day of PAF | 40 848 | 98.6% |
| 3) Drop patients initiated micafungin and a comparator additional PAF on the same day (‘dual PAF initiators’) | 40 500 | 97.8% |
| 4a) Drop patients with evidence of pre-existing chronic hepatic disease: as a discharge diagnosis at any hospitalization in the 183 days BEFORE the index hospitalization | 37 902 | 91.5% |
| 4b) Evidence of pre-existing chronic hepatic disease: as a visit diagnosis at any outpatient visit in the 183 days the BEFORE index hospitalization | 36 287 | 87.6% |
| 4c) Evidence of pre-existing chronic hepatic disease: as a discharge diagnosis DURING the index hospitalization | 32 045 | 77.4% |
| 5a) Evidence of pre-existing chronic kidney disease: as a discharge diagnosis at any hospitalization in the 183 days BEFORE the index hospitalization | 29 742 | 71.8% |
| 5b) Evidence of pre-existing chronic kidney disease: as a visit diagnosis at any outpatient visit in the 183 days BEFORE the index hospitalization | 29 089 | 70.2% |
| 5c) Evidence of pre-existing chronic kidney disease: as a discharge diagnosis DURING the index hospitalization | 25 929 | 62.6% |
| 6) Neither pre-treatment ALT nor pre-treatment AST results available: during 30 days before the index hospitalization but before PAF initiation | 23 948 | 57.8% |
| 7) Significantly elevated liver enzymes during 30 days before the index hospitalization but before PAF initiation: ALT >5× the hospital ULN (or >300 U/L) or AST >5× ULN (or >200 U/L) | 22 273 | 53.8% |
| 8) No pre-treatment creatinine results available: during 30 days before the index hospitalization but before PAF initiation | 22 257 | 53.7% |
| 9) Significantly reduced pre-treatment eGFRb: <30 mL/min | 17 690 | 42.7% |
aNumbers refer to individual patients. Few patients contributed twice, as first-use and as second-use PAF users.
beGFR was computed via the Mayo quadratic formula.32
Parental antifungal exposure
Patients were classified as parenteral users of either micafungin or a comparator PAF, which included caspofungin, anidulafungin, fluconazole, itraconazole, voriconazole, amphotericin B deoxycholate or lipid formulations of amphotericin B. In this observational study, embedded in routine care, any PAFs that were used for patients similar to those receiving micafungin were eligible as comparators. Through the stratified propensity score (PS) matching, only those comparator patients (including those receiving azoles) who, according to their health state, could have but did not receive micafungin contributed to the analysis.12
Patients were classified as first-use recipients of micafungin or a comparator PAF based on the first PAF administration during the index hospitalization. Among patients who received parenteral fluconazole during the index hospitalization, the study also included second-use treatments in the analysis. Second-use exposure status was defined based on the subsequent antifungal after parenteral fluconazole during the same hospitalization. Second-use patients were included in the study to increase the number of micafungin patients in the analysis, thereby improving the precision of effect estimates. First- and second-use patients were analysed separately and were only combined when no effect measure modification was observed.
Some patients who started on fluconazole contributed to the first-use as well as the second-use treatment cohort. Participation in two exposure groups is accounted for in the analysis by adjusting standard errors for dependent observations. The study did not consider covariate changes during follow-up, such as shifts in liver function tests that could have triggered the switching.
Patient follow-up and outcomes
Treatment-emergent hepatic injury was defined through the occurrence of at least one ALT or AST measurement greater than five times the hospital's upper limit of normal (ULN) from the day after the index PAF initiation through 30 days after the last day of the index PAF use (see Figure 1).
Baseline and follow-up periods for the 30 day treatment-emergent hepatic and renal injury outcome analysis. PAF, parenteral antifungal.
Treatment-emergent renal injury was defined as a decrease in eGFR by >20% from the latest assessment prior to index PAF treatment, following definitions in earlier studies.13–15 The eGFR value utilized for this definition was the average of the three lowest recorded eGFR values during the interval from the day after the index PAF initiation through 30 days after the last day of the index PAF use. Otherwise eligible patients for whom follow-up measurements of ALT, AST or eGFR were not available after initiation of PAF were removed from the hepatic or renal analyses, respectively.
Patient characteristics
Characteristics that were measured before the index hospitalization (at prior hospitalizations or prior outpatient visits as available) included cancer type (none, haematological cancer or any solid cancer), graft-versus-host disease, solid organ transplantation, bone marrow transplantation, haematopoietic disorders (agranulocytosis, neutropenia or pancytopenia), HIV/AIDS, congenital immunodeficiency, chronic renal disease, chronic liver disease, cardiovascular disease, diabetes, hypertension, chronic respiratory disease, non-cardiac vascular disease, haemostatic disorder, rheumatoid arthritis and other collagen vascular diseases, chronic gastrointestinal disease, immunosuppressant therapy, granulocyte colony-stimulating factor and chronic corticosteroid use.
Characteristics that were measured during the index hospitalization included the variables listed above plus: age; sex; race; income status; marital status; month and year of index admission; admission type; ICU stay on the day of antifungal initiation; the number of different medications received; oral antifungal therapy; receipt of potentially nephrotoxic16,17 or hepatotoxic18,19 medications before or during antifungal therapy; the number and type of antimicrobials received at index hospitalization; pre-existing renal disease; pre-existing liver disease; fungal disease; and individual medications received before PAF initiation.
Balancing of patient characteristics with PS matching
In order to reduce the risk of confounding, PS matching was used to create micafungin and comparator PAF cohorts that had similar prevalence of all covariates.
Separate PSs for micafungin versus other PAF treatment were estimated within two strata formed by first-use and second-use cohorts times three calendar eras (2005–07, 2008–09 and 2010–12). PS matching created groups with balanced covariate prevalence within each of the six cohort subgroups defined by these stratification factors and allowed for PS-adjusted subgroup analyses without further covariate adjustment.20
Each of the six stratum-specific PSs consisted of the fitted values from a logistic regression of micafungin versus comparator as the dependent variable, taken as a function of observed patient characteristics. Variable selection included the patient characteristics defined in the study protocol plus the product interaction terms between four age categories and 14 major indicating conditions. If the P value for removal of any of the interaction terms was >0.1, the interaction term was removed.
The formulary PAF and locally preferred PAF in treatment guidelines varied by hospital and, as a result, institution-determined characteristics were highly correlated with treatment choice. Adjustment for ‘instrumental variables’, such as hospital (in this case), can amplify confounding instead of reducing it.21,22 Therefore, model fitting for the six stratum-specific PS omitted variables and variable levels that were idiosyncratic to the hospitals. The PS model c-statistics ranged between 0.65 and 0.81 across the six strata.
The algorithm for matching each micafungin recipient sought up to three comparators with a maximum matching calliper of 0.01 on the natural PS scale. There were fewer than three available matching comparators for some of the micafungin patients, particularly in the smaller strata. The resulting matching ratio across all strata combined was 1 : 2.6.
Statistical analysis
Characteristics of all the identified, eligible and matched patients were tabulated by categories of antifungal use, separately for first-use and second-use treatment. Standardized differences were used to compare characteristics within treatment groups.23
Outcome risks were summarized from the frequency of occurrence of each of the outcomes and included 30 day risks and HRs with the corresponding 95% CI. A matched set-stratified Cox proportional hazards regression analysis provided the HR estimates.24 As effect estimates were not substantially modified by first-use versus second-use treatment status, combined HR estimates were calculated. CIs were based on robust standard errors from generalized estimating equations.25
The analyses were repeated in pre-defined patient subgroups defined by: age; sex; 14 different underlying diagnoses; days of antifungal therapy (1–5 days, 6–29 days and ≥30 days exposed); and receipt of hepatotoxic and nephrotoxic medications before or during PAF use.
Medical records abstraction to confirm successful matching
In order to evaluate whether the compared cohorts were well balanced on variables not present in the structured EMR data, we reviewed the discharge summaries of 1000 randomly selected micafungin and 1000 matched comparator PAF users from the larger hepatocellular carcinoma study from which the present data are drawn. We used a pilot-tested standard abstraction protocol with a computer-based data entry system for equal data quality across all hospitals.
The review of discharge summaries focused on the possible existence of factors not included in the structured EMR information, including: pre-existing renal or hepatic diseases; severity of underlying disease at admission; severity of comorbidities; smoking status; alcohol abuse; height; weight; BMI; service of admission (general internal medicine, surgery or oncology); ICU; admission as transfer from other institution; pre-existing disease with adverse implications for the liver including viral hepatitis, alcoholic liver disease, non-alcoholic fatty liver disease, non-alcoholic steatohepatitis, haemochromatosis and Wilson's disease, known aflatoxin exposure, α-1-antitrypsin deficiency, intrahepatic cholestasis, bile acid synthesis disorders, tyrosinaemia type I, defects in carbohydrate metabolism, porphyrias, cystic fibrosis, Alagille's syndrome, linked sideroblastic anaemia, Fanconi anaemia, hereditary fructose intolerance and hereditary haemorrhagic telangiectasia; pre-existing drug use with adverse implications for the liver (e.g. anabolic steroids); and selected discharge medications, e.g. oral antifungal therapy and potentially nephrotoxic or hepatotoxic medications.
Results
There were 17 690 individual patients eligible for this study (Table 1). Matching users of comparator PAFs were available for 84% (2970 of 3521) of the micafungin recipients. A total of 9696 patients contributed to the analysis (Table 2).
Population summary before and after propensity score (PS) matchinga
| Patients contributing to PS matching | Post-PS matching | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| micafungin | comparator | total | 1 : 3 | 1 : 2 | 1 : 1 | micafungin | comparator | total | proportion micafungin matched (%) | |
| 30 day 1st use | 2964 | 14 721 | 17 685 | 6280 | 1161 | 1304 | 2609 | 6136 | 8745 | 88 |
| 30 day 2nd use | 560 | 677 | 1237 | 340 | 177 | 434 | 361 | 590 | 951 | 64 |
| Total | 3521 | 15 398 | 18 922 | 2970 | 6726 | 9696 | 84 | |||
| Patients contributing to PS matching | Post-PS matching | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| micafungin | comparator | total | 1 : 3 | 1 : 2 | 1 : 1 | micafungin | comparator | total | proportion micafungin matched (%) | |
| 30 day 1st use | 2964 | 14 721 | 17 685 | 6280 | 1161 | 1304 | 2609 | 6136 | 8745 | 88 |
| 30 day 2nd use | 560 | 677 | 1237 | 340 | 177 | 434 | 361 | 590 | 951 | 64 |
| Total | 3521 | 15 398 | 18 922 | 2970 | 6726 | 9696 | 84 | |||
aFive additional patients were excluded from the 17 690 patients shown in Table 1, leaving 17 685 first-use patients.
Population summary before and after propensity score (PS) matchinga
| Patients contributing to PS matching | Post-PS matching | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| micafungin | comparator | total | 1 : 3 | 1 : 2 | 1 : 1 | micafungin | comparator | total | proportion micafungin matched (%) | |
| 30 day 1st use | 2964 | 14 721 | 17 685 | 6280 | 1161 | 1304 | 2609 | 6136 | 8745 | 88 |
| 30 day 2nd use | 560 | 677 | 1237 | 340 | 177 | 434 | 361 | 590 | 951 | 64 |
| Total | 3521 | 15 398 | 18 922 | 2970 | 6726 | 9696 | 84 | |||
| Patients contributing to PS matching | Post-PS matching | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| micafungin | comparator | total | 1 : 3 | 1 : 2 | 1 : 1 | micafungin | comparator | total | proportion micafungin matched (%) | |
| 30 day 1st use | 2964 | 14 721 | 17 685 | 6280 | 1161 | 1304 | 2609 | 6136 | 8745 | 88 |
| 30 day 2nd use | 560 | 677 | 1237 | 340 | 177 | 434 | 361 | 590 | 951 | 64 |
| Total | 3521 | 15 398 | 18 922 | 2970 | 6726 | 9696 | 84 | |||
aFive additional patients were excluded from the 17 690 patients shown in Table 1, leaving 17 685 first-use patients.
The comparisons of patients receiving micafungin and those receiving other PAFs indicate substantial similarities with respect to baseline clinical characteristics after PS matching. Table S1 (available as Supplementary data at JAC Online) shows characteristics of the PS-matched cohorts, stratified by exposure group, micafungin (2970) versus comparator PAFs (6726). PAF use in the PS-matched comparison group consisted of 55% fluconazole, 16% caspofungin, 14% voriconazole, 12% amphotericin B and 2% two or more comparator PAFs on the same day.
In the PS-matched study cohorts, 13% of micafungin patients and 12% of comparator PAF patients experienced treatment-emergent acute liver injury (Table 3). A total of 205 micafungin patients (7%) and 698 comparator PAF patients (10%) who did not have ALT or AST values recorded during treatment or follow-up dropped out of the analysis. With adjustment for censoring due to death before the end of the planned 30 day follow-up, the HR comparing micafungin with PAF patients was 0.99 with a 95% CI of 0.86–1.14.
30 day occurrence of acute liver injury among all parenteral antifungal users
| Group | Patients (N) | Patient-days | No. of events | Crude risk | 95% CI | 30 day ratea | 95% CI | HRb | 95% CI |
|---|---|---|---|---|---|---|---|---|---|
| Micafungin | 2765 | 95 627 | 371 | 0.13 | 0.12–0.14 | 0.12 | 0.11–0.13 | 0.99 | 0.86–1.14 |
| Missingc | 205 | ||||||||
| Comparator | 6028 | 196 191 | 743 | 0.12 | 0.11–0.13 | 0.11 | 0.10–0.12 | 1.00 | |
| Missingc | 698 | ||||||||
| Combined | 8793 | 291 818 | 1114 | 0.13 | 0.12–0.14 | 0.11 | 0.10–0.12 |
| Group | Patients (N) | Patient-days | No. of events | Crude risk | 95% CI | 30 day ratea | 95% CI | HRb | 95% CI |
|---|---|---|---|---|---|---|---|---|---|
| Micafungin | 2765 | 95 627 | 371 | 0.13 | 0.12–0.14 | 0.12 | 0.11–0.13 | 0.99 | 0.86–1.14 |
| Missingc | 205 | ||||||||
| Comparator | 6028 | 196 191 | 743 | 0.12 | 0.11–0.13 | 0.11 | 0.10–0.12 | 1.00 | |
| Missingc | 698 | ||||||||
| Combined | 8793 | 291 818 | 1114 | 0.13 | 0.12–0.14 | 0.11 | 0.10–0.12 |
aThe 30 day rate is the number of events divided by the number of patient-days of observation, multiplied by 30. It differs from the crude risk in that it takes account of censoring due to death.
bThis HR is adjusted for the factors entering into the PS. The variable-ratio matching was incorporated into the analysis by conditioning on the match sets in a stratified proportional hazards regression analysis, which also accounts for censoring due to death. Standard errors were computed robustly because some second-use patients may also have contributed as first-use patients.
cMissing means that these patients had no follow-up laboratory test result recorded in the hospital system.
30 day occurrence of acute liver injury among all parenteral antifungal users
| Group | Patients (N) | Patient-days | No. of events | Crude risk | 95% CI | 30 day ratea | 95% CI | HRb | 95% CI |
|---|---|---|---|---|---|---|---|---|---|
| Micafungin | 2765 | 95 627 | 371 | 0.13 | 0.12–0.14 | 0.12 | 0.11–0.13 | 0.99 | 0.86–1.14 |
| Missingc | 205 | ||||||||
| Comparator | 6028 | 196 191 | 743 | 0.12 | 0.11–0.13 | 0.11 | 0.10–0.12 | 1.00 | |
| Missingc | 698 | ||||||||
| Combined | 8793 | 291 818 | 1114 | 0.13 | 0.12–0.14 | 0.11 | 0.10–0.12 |
| Group | Patients (N) | Patient-days | No. of events | Crude risk | 95% CI | 30 day ratea | 95% CI | HRb | 95% CI |
|---|---|---|---|---|---|---|---|---|---|
| Micafungin | 2765 | 95 627 | 371 | 0.13 | 0.12–0.14 | 0.12 | 0.11–0.13 | 0.99 | 0.86–1.14 |
| Missingc | 205 | ||||||||
| Comparator | 6028 | 196 191 | 743 | 0.12 | 0.11–0.13 | 0.11 | 0.10–0.12 | 1.00 | |
| Missingc | 698 | ||||||||
| Combined | 8793 | 291 818 | 1114 | 0.13 | 0.12–0.14 | 0.11 | 0.10–0.12 |
aThe 30 day rate is the number of events divided by the number of patient-days of observation, multiplied by 30. It differs from the crude risk in that it takes account of censoring due to death.
bThis HR is adjusted for the factors entering into the PS. The variable-ratio matching was incorporated into the analysis by conditioning on the match sets in a stratified proportional hazards regression analysis, which also accounts for censoring due to death. Standard errors were computed robustly because some second-use patients may also have contributed as first-use patients.
cMissing means that these patients had no follow-up laboratory test result recorded in the hospital system.
Among micafungin patients, 63% experienced treatment-emergent acute renal injury, as did 65% of comparator PAF patients (Table 4). Patients who did not have an eGFR value recorded during treatment or follow-up dropped out of the analysis. There were 44 such micafungin patients (1%) and 106 comparator patients (1%). The proportional hazards model comparing micafungin versus comparator patients yielded an HR of 0.93 (95% CI 0.87–0.99).
30 day occurrence of acute renal injury among all parenteral antifungal users
| Group | Patients (N) | Patient-days | No. of events | Crude risk | 95% CI | 30 day ratea | 95% CI | HRb | 95% CI |
|---|---|---|---|---|---|---|---|---|---|
| Micafungin | 2926 | 56 011 | 1832 | 0.63 | 0.61–0.65 | 0.98 | 0.94–1.02 | 0.93 | 0.87–0.99 |
| Missingc | 44 | ||||||||
| Comparator | 6620 | 118 053 | 4281 | 0.65 | 0.64–0.66 | 1.09 | 1.06–1.12 | 1.00 | |
| Missingc | 106 | ||||||||
| Combined | 9546 | 174 064 | 6113 | 0.64 | 0.63–0.65 | 1.05 | 1.02–1.08 |
| Group | Patients (N) | Patient-days | No. of events | Crude risk | 95% CI | 30 day ratea | 95% CI | HRb | 95% CI |
|---|---|---|---|---|---|---|---|---|---|
| Micafungin | 2926 | 56 011 | 1832 | 0.63 | 0.61–0.65 | 0.98 | 0.94–1.02 | 0.93 | 0.87–0.99 |
| Missingc | 44 | ||||||||
| Comparator | 6620 | 118 053 | 4281 | 0.65 | 0.64–0.66 | 1.09 | 1.06–1.12 | 1.00 | |
| Missingc | 106 | ||||||||
| Combined | 9546 | 174 064 | 6113 | 0.64 | 0.63–0.65 | 1.05 | 1.02–1.08 |
aThe 30 day rate is the number of events divided by the number of patient-days of observation, multiplied by 30. It differs from the crude risk in that it takes account of censoring due to death.
bThis HR is adjusted for the factors entering into the PS. The variable-ratio matching was incorporated into the analysis by conditioning on the match sets in a stratified proportional hazards regression analysis, which also accounts for censoring due to death. Standard errors were computed robustly because some second-use patients may also have contributed as first-use patients.
cMissing means that these patients had no follow-up laboratory test result recorded in the hospital system.
30 day occurrence of acute renal injury among all parenteral antifungal users
| Group | Patients (N) | Patient-days | No. of events | Crude risk | 95% CI | 30 day ratea | 95% CI | HRb | 95% CI |
|---|---|---|---|---|---|---|---|---|---|
| Micafungin | 2926 | 56 011 | 1832 | 0.63 | 0.61–0.65 | 0.98 | 0.94–1.02 | 0.93 | 0.87–0.99 |
| Missingc | 44 | ||||||||
| Comparator | 6620 | 118 053 | 4281 | 0.65 | 0.64–0.66 | 1.09 | 1.06–1.12 | 1.00 | |
| Missingc | 106 | ||||||||
| Combined | 9546 | 174 064 | 6113 | 0.64 | 0.63–0.65 | 1.05 | 1.02–1.08 |
| Group | Patients (N) | Patient-days | No. of events | Crude risk | 95% CI | 30 day ratea | 95% CI | HRb | 95% CI |
|---|---|---|---|---|---|---|---|---|---|
| Micafungin | 2926 | 56 011 | 1832 | 0.63 | 0.61–0.65 | 0.98 | 0.94–1.02 | 0.93 | 0.87–0.99 |
| Missingc | 44 | ||||||||
| Comparator | 6620 | 118 053 | 4281 | 0.65 | 0.64–0.66 | 1.09 | 1.06–1.12 | 1.00 | |
| Missingc | 106 | ||||||||
| Combined | 9546 | 174 064 | 6113 | 0.64 | 0.63–0.65 | 1.05 | 1.02–1.08 |
aThe 30 day rate is the number of events divided by the number of patient-days of observation, multiplied by 30. It differs from the crude risk in that it takes account of censoring due to death.
bThis HR is adjusted for the factors entering into the PS. The variable-ratio matching was incorporated into the analysis by conditioning on the match sets in a stratified proportional hazards regression analysis, which also accounts for censoring due to death. Standard errors were computed robustly because some second-use patients may also have contributed as first-use patients.
cMissing means that these patients had no follow-up laboratory test result recorded in the hospital system.
For all subgroups and both outcomes, the 95% CI of the HR included the null value of 1.0, except for renal injury among persons with 1–5 days of PAF, in which the observed 30 day risk was lower in micafungin recipients than in the recipients of the PAF comparator agents (HR = 0.84; 95% CI 0.74–0.96). The small sizes of some of the subgroups meant that the corresponding estimates had very wide CIs.
In-hospital mortality for micafungin patients across all hospitals after PS matching was 504/2970 (17.0%) and for comparator PAFs was 1080/6726 (16.1%) (relative risk 1.06; 95% CI 0.96–1.16).
The medical record analysis of 998 matched pairs (out of 1000 pairs sought) showed no meaningful difference in self-reported tobacco and alcohol use or BMI (Supplementary Data). The proportions of patients staying in an ICU were 41% and 40% for micafungin and comparator PAFs, respectively. There was a difference in the prescribing of potentially hepatotoxic medications for use after discharge, which was less frequent among micafungin patients (61% versus 72%). Post-discharge prescription of oral antifungal agents was also less frequent among micafungin patients than among comparators.
In order to evaluate the potential confounding effect of the 11% difference in post-discharge use of potentially hepatotoxic medications, an algebraic sensitivity analysis of residual confounding was conducted.26 Even if the relative risk of acute hepatic injury from the post-discharge use of potentially hepatotoxic medication use was 4-fold elevated, an association between micafungin and acute hepatic injury would be underestimated by only 10% for an imbalance of the magnitude actually observed (Supplementary Data). Scenarios that invoke less extreme associations between risk factors and hepatic injury resulted in much smaller biases.
Discussion
In a multicentre cohort study of 9696 patients with a range of underlying conditions in seven tertiary care hospitals, multivariate PS-matched analyses showed no difference in the occurrence of treatment-emergent liver or renal injury between micafungin and comparison PAF use. The data suggest that micafungin may be associated with a 7% lower risk of renal injury overall.
The lack of an association between the use of parenteral micafungin and short-term hepatic or renal injury within 30 days in the largest study on this topic in routine clinical care is reassuring for clinicians and is in accordance with earlier randomized and observational studies.27–31 Generalization from this study is limited to very sick patients, as it was conducted in patients hospitalized in tertiary care centres and the cohort experienced an in-hospital mortality risk of 17%. Many of the parenteral agents studied here are in fact preferentially used in very sick patients, so the study population may be informative about persons in whom these drugs are actually used.
This multicentre study is based on medical information that was generated during the routine provision of care using EMR systems and other data systems. As these data are subject to miscoding and misclassification and are not standardized across hospitals, a number of quality assurance steps were closely followed. All of these steps plus plausibility checks and pilot data ensured that the data are of the highest quality possible in this setting. The type of PAF exposure was electronically coded by pharmacy dispensing systems and laboratory test results that were the basis for the study outcome measures were automatically captured in electronic form by central laboratory facilities. Thus, the exposure and outcome measurements for this study are thought to be of high quality. We chose a standard laboratory test-based definition of renal dysfunction/injury of a 20% eGFR reduction, which has been used frequently in the literature. In this very sick study population, a 20% reduction in eGFR proved to be common, experienced by >50% of patients. Disease and comorbidity assessment including their severity is a mix of recorded hospital discharge diagnoses and procedures, which are generally recorded with high quality.32
This micafungin safety study combined several antifungal agents into a single comparison group that reflects prescribing for patients similar to those who actually received micafungin in US tertiary care centres during the study period. The availability of medically similar patients for the comparator PAF group, as evidenced by the balance of all patient characteristics after PS matching, suggests that the choice between micafungin and comparator PAFs may be determined principally by protocols in the treating institutions and not by patient factors.
In interpreting clinical studies without baseline randomization, decision-makers should be concerned about whether the treatment groups had the same baseline risk of the outcome. This study is based on a multivariate PS analysis that included 62 variables plus 14 interaction terms. PS matching is known as a technique that reliably balances all patient covariates. In this study, the matching resulted in no relevant residual confounding in observed patient characteristics. Although in-hospital mortality was not a pre-planned study endpoint, there was no difference between treatment groups across all hospitals. The similarity of in-hospital mortality in the compared groups is consistent with the hypothesis that the PS matching had its intended effect of making the micafungin and PAF comparator cohorts similar with respect to underlying morbidity.
The similarity of the groups extends to baseline characteristics derived from medical chart review of 998 patient-pairs. These characteristics were not available for the PS-matching algorithm and provide a further check on the efficacy of matching beyond the observed characteristics in producing balanced cohorts. Use of ICU services, an important indicator of morbidity, was essentially identical.
There were modest group imbalances in the post-discharge prescribing of potentially hepatotoxic medications, with lower use in patients who had received micafungin. Quantitative estimates showed that the observed difference would be expected to result in no more than a 10% decrease in relative risk estimates for micafungin versus comparator PAFs, even for an extreme range of plausible scenarios of risk.
Overall, for a wide spectrum of underlying conditions, we observed no increase in liver injury by micafungin and possibly a reduced risk of renal dysfunction by micafungin compared with other PAF medications.
Funding
This study was sponsored by Astellas Pharma Europe through a research contract with WHISCON, LLC. Astellas is the manufacturer of micafungin, one of the study drugs.
Transparency declarations
S. Schneeweiss is a consultant to WHISCON, LLC and to Aetion, Inc., a software manufacturer in which he also owns equity. He is principal investigator of investigator-initiated grants to the Brigham and Women's Hospital from PCORI, NIH, Genentech, and Boehringer Ingelheim unrelated to the topic of this study. A. M. W. and L. W. are both employees of WHISCON, LLC. M. D. J. is consultant to Charles River Laboratories and Astellas Pharma US, and she received a research contract from WHISCON, LLC. F. M. M. is consultant to WHISCON, LLC and Merck & Co, and he together with C. A. V. received a research contract from WHISCON, LLC. P. L. C. and J. N. received a research contract from WHISCON, LLC. F. P. S. received a research contract from WHISCON, LLC. S. Shoham is a consultant to Theravance, Inc., Biota Pharmaceuticals, Inc., Cidara Therapeutics, Inc., and received research support from Pfizer, ViroPharma, Inc., Scynexis Inc., Chimerix, Inc., Ansun Biopharma, Inc., Gilead, Merck & Co, and Astellas Pharma US, Inc., and WHISCON, LLC. M. W. received a research contract from WHISCON, LLC. T. A. was senior biostatistician at Astellas Pharma Europe and an Astellas employee. He is recently deceased. All other authors: no potential conflicts of interest to declare.
Acknowledgements
We thank Drs John Seeger, Humberto Ferguson and Joop van Oene for their support to the study at various stages.
References
Author notes
Co-authors 2–14 are listed in alphabetical order.
