Outcomes of antifungal prophylaxis for newly diagnosed AML patients treated with a hypomethylating agent and venetoclax

Abstract

Antifungal prophylaxis (AFP) is recommended for acute myeloid leukemia (AML) patients receiving the combination of venetoclax (VEN) and a hypomethylating agent (HMA), but the benefit of this practice is unclear. We identified 131 patients with newly-diagnosed AML who received frontline VEN/HMA and evaluated the use of AFP and its association with invasive fungal infections (IFIs) and AML outcomes. 17% of our patients received AFP at any time. Overall incidence of any IFI (“possible,” “probable,” or “proven” infection, as defined by the European Mycoses Study Group) was 13%, and the incidence did not differ based on AFP use (p = 0.74). Median overall survival did not differ based on AFP use or lack thereof (8.1 vs 12.5 months, respectively; p=0.14). Our findings suggest that, at an institution where the incidence of fungal infections is low, there does not appear to be a role for AFP in newly-diagnosed AML patients receiving VEN/HMA.

Introduction

The oral BCL-2 inhibitor venetoclax (VEN) in combination with a hypomethylating agent (HMA) is an active, frontline therapy for patients with acute myeloid leukemia (AML) who are ineligible for more intensive, anthracycline-based induction chemotherapy due to older age (≥75 years) or medical comorbidities. For these patients, the combination of VEN and the HMA azacitidine is associated with higher complete remission (CR) rates and longer overall survival (OS) compared with azacitidine alone (CR: 36.7% vs. 17.9%, respectively; median OS: 14.7 vs 9.6 months),[1] and represents a new treatment option for a large number of AML patients.

Consensus guidelines recommend the use of antifungal prophylaxis (AFP) for AML patients who receive intensive induction chemotherapy such as anthracycline and cytarabine.[2] This recommendation has been extrapolated to VEN/HMA-treated AML patients[3]—in the randomized study of VEN/HMA, over 50% of study participants received AFP in the VEN/HMA arm.[1] However, there is no definitive evidence to support the use of AFP for VEN/HMA-treated patients, and this clinical practice requires evaluation for two main reasons: first, the incidence of invasive fungal infections (IFIs) appears low in AML patients receiving VEN/HMA therapy;[4] and second, azoles, which are common AFP agents of choice, are CYP3A4 inhibitors that require VEN dose-reductions,[5] and the impact of dose-reduction on treatment efficacy and AML outcomes has received limited study.[6]

Here, we report the pattern of AFP use and its association with IFIs and AML outcomes in patients receiving frontline VEN/HMA at our institution.

Methods

Study design and patients

This study was approved by the Institutional Review Board at the Dana-Farber Cancer Institute (DFCI). Consecutive patients aged ≥18 years who received frontline VEN/HMA for newly-diagnosed AML between 2016–2021 were identified in the DFCI Hematologic Malignancy Data Repository[7] for retrospective chart review. This study did not include relapsed/refractory AML patients. Patients were risk stratified according to European LeukemiaNet (ELN) criteria.[8] Detection of co-mutations was performed using the DFCI Rapid Heme Panel that utilizes amplicon-based next generation sequencing and contains a panel of 88 genes.[9] “Secondary-like” mutation patterns were defined by the presence of a mutation in SRSF2, SF3B1, U2AF1, ZRSR2, ASXL1, EZH2, BCOR, or STAG2.[10] For this study, de novo AML was defined to be mutually exclusive of secondary, therapy-related, and “secondary-like” AML. Patient fitness, or lack thereof, was defined according to consensus criteria.[11]

AFP was administered at the provider’s discretion. Use of AFP at any time during VEN/HMA therapy was recorded per patient. AFP was distinguished from therapeutic and empiric antifungal treatment and was defined as the use of an antifungal agent before any diagnosed or suspected fungal infection. IFIs occurring during VEN/HMA therapy were adjudicated as “possible,” “probable,” or “proven” infections according to criteria established by the European Organization for Research and Treatment of Cancer/Mycoses Study Group Education and Research Consortium.[12] Adjudication was performed independently by two authors (E.C.C and C.E.H.) with a third (N.I.) serving as tie-breaker. Neutropenia was considered a host factor for IFIs if the absolute neutrophil count (ANC) was <0.5 × 10⁹/L for ≥10 consecutive days,[12,13] and the number of such instances was recorded per patient. Therapeutic drug monitoring was performed for patients receiving voriconazole, but not for posaconazole unless there were reasons to suspect malabsorption or medication non-adherence. For voriconazole, clinical practice at our institution targeted a minimum plasma level of 1–1.5 μg/mL.

HMA type and dose, as well as modifications to VEN dose, duration, and cycle length, were recorded. Objective responses were determined per ELN criteria[8] and recorded at time of best response. When available, measurable residual disease (MRD) status as determined by multiparametric flow cytometry (HematoLogics, Seattle, WA) was documented. MRD negativity was defined as disease quantification <0.1% per ELN guidelines.[14] Samples without detectable residual disease but with a lower-limit of detection ≥0.1% (such as due to limitations of sample quality) were deemed indeterminate.

Statistical analysis

Descriptive statistics on patient characteristics and outcomes were compared between groups via nonparametric testing including Wilcoxon rank-sum, chi-squared, and the Fisher Exact tests. OS was estimated using the method of Kaplan and Meier, and log-rank tests were used to compare survival outcomes between groups. Multivariable Cox proportional-hazards models were used to obtain hazard ratios between groups. A competing risk model for time to first IFI was constructed incorporating death as a competing risk. The Akaike information criterion (AIC) score was used to compare different time-to-IFI models constructed using different covariates. A lower AIC score reflects a better-fitting model.

Results

Patient population

A total of 131 patients with newly-diagnosed AML treated with frontline VEN/HMA were identified (Table 1). A minority of these patients received AFP at any time (17%). For the overall cohort, the median age was 72 years. The majority of patients were less than 75 years old (66%), male (65%), and white (91%). Forty-three percent of patients lacked fitness to receive intensive chemotherapy according to consensus-based criteria.[11] Most patients had ELN intermediate- or adverse-risk disease (92%) and secondary-, therapy-related, or “secondary-like” AML (71%).[10] Frequently mutated genes included TP53 (37%), ASXL1 (22%), and RUNX1 (19%). These baseline characteristics did not differ between patients who did and did not receive AFP.

Table 1:

Patient and disease characteristics.

	No AFP use N = 109		Yes AFP use N = 22		Total N = 131		P-value
DEMOGRAPHICS
Age at diagnosis, median (range)	72	(22, 89)	71	(25, 83)	72	(22, 89)	0.29
Age ≥ 75 years, n (%)	37	(34%)	7	(32%)	44	(34%)	0.85
Male sex, n (%)	69	(63%)	16	(73%)	85	(65%)	0.40
Race, n (%)							0.60
White	99	(91%)	20	(91%)	119	(91%)
Black	3	(2.8%)	0	(0%)	3	(2.3%)
Asian	4	(3.7%)	2	(9.1%)	6	(4.6%)
Other	3	(2.8%)	0	(0%)	3	(2.3%)
Fitness at diagnosis,[11] n (%)							0.26
Fit	60	(55%)	15	(68%)	75	(57%)
Unfit	49	(45%)	7	(32%)	56	(43%)
DISEASE CHARACTERISTICS
ANC (109/L) at diagnosis, median (range)	0.99	(0.0, 73.0)	0.82	(0.1, 18.7)	0.98	(0.0, 73.0)	0.65
# patients with at least one instance of ANC <0.5 × 109/L for ≥10 days, n (%)	94	(86%)	19	(86%)	113	(86%)	>0.99
# instances of ANC <0.5 × 109/L for ≥10 days per patient, median (range)	1	(1–7)	1	(1–4)	1	(1–7)	0.64
Cytogenetic risk n (%)							0.39
Favorable	9	(8.6%)	1	(4.5%)	10	(7.6%)
Intermediate	18	(17%)	7	(32%)	25	(20%)
Adverse	78	(74%)	14	(64%)	92	(72%)
AML ontogeny, n (%)
De novo	33	(30%)	5	(23%)	38	(29%)	0.48
Secondary	52	(48%)	13	(59%)	65	(50%)	0.33
Therapy-related	5	(4.6%)	3	(14%)	8	(6.1%)	0.13
Secondary-like[10]	48	(44%)	5	(23%)	53	(40%)	0.063
Therapy for preceding HM, n (%)
Prior HMA	14	(13%)	9	(41%)	23	(18%)	0.004
Prior allo- or auto-HCT	14	(13%)	3	(14%)	17	(13%)	>0.99
Prior HMA or HCT	25	(23%)	10	(45%)	35	(27%)	0.029
Molecular mutations, n (%)
TP53	38	(35%)	10	(45%)	48	(37%)	0.35
ASXL1	25	(23%)	3	(14%)	28	(21%)	0.41
RUNX1	23	(21%)	2	(9.1%)	25	(19%)	0.25
NPM1	15	(14%)	0	(0%)	15	(11%)	0.074
FLT3-ITD	8	(7.3%)	1	(4.5%)	9	(6.9%)	>0.99
TREATMENT CHARACTERISTICS
HMA type and duration, n (%)							>0.99
Azacitidine 7-day	36	(33%)	7	(32%)	43	(33%)
Decitabine 5-day	54	(50%)	11	(50%)	65	(50%)
Decitabine 10-day	19	(17%)	4	(18%)	23	(18%)
VEN <400mg/day at any time,* n (%)	25	(23%)	21	(95%)	46	(35%)	<0.001
Average days of cycle delay during first 4 cycles, median (range)	6	(0, 76)	7	(0, 37)	7	(0, 76)	0.61
VEN <21 days/cycle at any time, n (%)	48	(44%)	12	(55%)	60	(46%)	0.41
Antibacterial ppx at any time, n (%)	59	(54%)	18	(82%)	77	(59%)	0.016
Antiviral ppx at any time, n (%)	85	(78%)	20	(91%)	105	(80%)	0.24
Antifungal ppx at any time, n (%)
Fluconazole	0	(0%)	15	(68%)	15	(11%)	<0.001
Voriconazole	0	(0%)	2	(9.1%)	2	(1.5%)	0.027
Posaconazole	0	(0%)	2	(9.1%)	2	(1.5%)	0.027
Isavuconazole	0	(0%)	3	(14%)	3	(2.3%)	0.004
Micafungin	0	(0%)	0	(0%)	0	(0%)	N/A

^*Pertains only to the steady-state dose of venetoclax (i.e. excludes the initial dose ramp-up of venetoclax during start of Cycle 1, which is standard of care).

Abbreviations: AFP = antifungal prophylaxis, ANC = absolute neutrophil count, AML = acute myeloid leukemia, HM = hematologic malignancy, HMA = hypomethylating agent, HCT = hematopoietic cell transplant, allo = allogeneic, auto = autologous, VEN = venetoclax, ppx = prophylaxis

A higher proportion of patients who received any AFP had prior treatment with a hypomethylating agent (HMA) and/or hematopoietic cell transplantation (HCT) for an antecedent hematologic malignancy compared to patients who received no AFP (45% vs. 23%, respectively; p=0.029, Table 1). A higher proportion of patients in the AFP cohort also received antibacterial prophylaxis at any time (82% vs. 54%, p=0.016). AFP usage was not associated with more frequent shortening of VEN dose-duration to <21 days per cycle (p=0.41) or increase in the median days of delay between cycles (p=0.61). However, patients receiving AFP more frequently underwent VEN dose-reduction (91% vs. 19%, p<0.001); this was expected since all patients receiving AFP were treated with azoles. No patient received an echinocandin as prophylaxis. Azoles used include mold active agents such as voriconazole (9.1%), posaconazole (9.1%), and isavuconazole (14%). The most frequently used agent was fluconazole (68%), which lacks substantial mold activity. For patients receiving voriconazole (n=2), adequate plasma drug levels were detected (1.4 and 1.5 μg/mL).

AML outcomes

AML best response was recorded after a median of two cycles of VEN/HMA, and this interval did not differ based on AFP use (p=0.20). The distribution of best responses was different according to AFP use (p=0.046, Table 2), but when specifically considering the composite rate of complete remission (CR) and CR with incomplete hematologic recovery (CRi), the difference did not reach statistical significance (32.1% for patients receiving AFP vs. 53% for patients who did not, p=0.080). Patients receiving AFP had lower rates of MRD-negative CR/CRi (p=0.023, Table 3), but comparisons are limited by the small number of patients undergoing MRD assessment in the AFP cohort (n=7).

Table 2:

Best AML response.

	No AFP use N = 109		Yes AFP use N = 22		Total N = 131		P-value
Best response[8]							0.046
CR	42	(39%)	2	(9.1%)	44	(34%)
CRi	15	(14%)	5	(23%)	20	(15%)
MLFS	15	(14%)	4	(18%)	19	(15%)
Persistent disease	30	(28%)	8	(36%)	38	(29%)
Not evaluable*	7	(6.4%)	3	(14%)	10	(7.6%)
Proceeded to HCT after VEN/HMA, n (%)	23	(21%)	3	(14%)	26	(20%)	0.56
Relapsed disease after VEN/HMA, n (%)	36	(33%)	6	(27%)	42	(32%)	0.60

^*Due to lack of bone marrow evaluation following VEN/HMA initiation.

Abbreviations: AFP = antifungal prophylaxis, CR = complete remission, CRi = CR with incomplete hematologic recovery, MLFS = morphologic leukemia free state, HCT = hematopoietic cell transplantation, VEN = venetoclax, HMA = hypomethylating agent

Table 3:

Flow cytometry MRD status for patients achieving CR/CRi at best response.

	Achieved CR/CRi at best response
	No AFP use N = 57		Yes AFP use N = 7		Total N = 64		P-value
MRD status at best response							0.023
CR/CRi, MRD-	21	(37%)	0	(0%)	21	(33%)
CR/CRi, MRD+	23	(40%)	2	(29%)	25	(39%)
CR/CRi, MRD indeterminate*	2	(3.5%)	1	(14%)	3	(4.7%)
CR/CRi, MRD not available	11	(19%)	4	(57%)	15	(23%)

^*For these samples, no MRD was detected but lower-limits of detection were ≥0.1% due to sample quality.

AFP = antifungal prophylaxis, CR = complete remission, CRi = CR with incomplete hematologic recovery, MRD = measurable residual disease

The median OS for the entire cohort was 11.4 months (95% CI 9.1–14.6) with a 1-year OS of 48%. The most frequent causes of death were disease progression (53%), infection (18%), and hemorrhage (5.3%). Overall survival did not differ based on AFP use (12.5 months for patients receiving no AFP versus 8.1 months for those who did; p=0.14, Figure 1), though this comparison may have been limited by the sample size. There was no difference in the proportion of patients proceeding to HCT or experiencing relapsed disease following VEN/HMA therapy according to AFP use (Table 2). Median survival was worse for patients who did not achieve CR/CRi versus patients who did (8.3 months vs. 15.4 month, respectively; p=0.0002). In a multivariable model that included patient age, AML ontogeny, prior therapy for the antecedent hematologic malignancy, number of instances of severe neutropenia lasting ≥10 days, TP53 mutation, and use of AFP, only older age and presence of mutated TP53 was associated with worsened OS (Table 4). TP53-mutation was the most poorly prognostic covariate with a HR of 2.83 (p<0.001).

Figure 1:

Overall survival.

Abbreviations: AFP = antifungal prophylaxis

Table 4:

Multivariable analysis of survival and time to IFI with death as a competing risk.

	HR	95% CI	P-value
SURVIVAL
Age at diagnosis	1.03	[1.0003–1.05]	0.047
s/t-AML or secondary-like[10] AML	1.61	[0.94–2.76]	0.086
TP53 mutation	2.83	[1.73–4.63]	<0.001
Treatment received for preceding HM	1.07	[0.63–1.81]	0.80
AFP use during first 4 cycles of VEN/HMA	1.33	[0.77–2.30]	0.31
# instances of ANC <0.5 × 109/L for ≥10 days	0.86	[0.73–1.01]	0.068
TIME TO INVASIVE FUNGAL INFECTION
Age at diagnosis	0.93	[0.89–0.98]	0.0032
# instances of ANC <0.5 × 109/L for ≥10 days	0.57	[0.27–1.20]	0.14
Lack of fitness	4.21	[1.18–15]	0.027
TP53 mutation	3.06	[1.02–9.15]	0.046

Abbreviations: s/t-AML = secondary or therapy-related AML, HCT = hematopoietic cell transplantation, HMA = hypomethyalting agent, VEN = venetoclax, AFP = antifungal prophylaxis, IFI = invasive fungal infection, ANC = absolute neutrophil count

Invasive fungal infection outcomes

The proportion of patients who developed any IFI (possible, probable, or proven infection) in the entire cohort was 13.0%, and the incidence did not differ based on AFP usage (Table 5, p=>0.99). Detailed patient characteristics are provided in Table S1. One patient (Pt #8) developed two IFIs. There was no difference in the distribution of possible, probable, or proven IFIs (p=0.74) between the two cohorts, nor when considering only probable and proven IFIs (p=>0.99). Of the proven IFIs, the identified fungal species included Candida parapsilosis, Candida krusei, Fusarium species, and Aspergillus fumigatus.

Table 5:

IFI outcomes.

	No AFP use N = 109		Yes AFP use N = 22		P-value
# patients with IFI	15	(14%)	2	(9.1%)	0.74
Per infection[12]					>0.99
Possible	12		2
Probable	1		0
Proven	3		0

Abbreviations: AFP = antifungal prophylaxis, IFI = invasive fungal infection

Neither prior therapy for an antecedent hematologic malignancy nor secondary/”secondary-like” AML was significantly associated with IFIs on univariate analysis (p=0.16 and p=0.28, respectively). In a competing risk model for time to first IFI with death as a competing risk, poor patient fitness and presence of a TP53 mutation were associated with a higher hazard of IFI (HR 4.2 and HR 3.1, respectively; p=0.027 and p=0.046, Table 4). Older age was associated with a decreased risk of IFI, but the hazard reduction was small (HR 0.93). A higher number of instances of prolonged, severe neutropenia (defined in this study as <0.5 × 10⁹/L for ≥10 consecutive days) was not associated with shortened time to an IFI.

Given the limited number of IFIs, we were restricted in the number and choice of covariates in the multivariable competing risks model of time to IFI. In particular, we could not incorporate AFP use as a covariate because of the very low number of patients who developed an IFI in the AFP cohort (n=2). Indeed, incorporating AFP use in an alternative model was associated with a higher AIC score (Model 2, Table S2), denoting the model’s poorer fit of the data. Similarly, an alternative model incorporating AML ontogeny also had a higher AIC score and poorer data fit (Model 3, Table S2).

Discussion

Antifungal prophylaxis was broadly recommended for patients during the clinical trial development of VEN/HMA, but in clinical practice the need for AFP for all VEN/HMA-treated patients remains uncertain. Current scrutiny on this issue centers on whether IFIs are indeed frequent adverse events for VEN/HMA-treated patients, whether AFP meaningfully decreases rates of IFI, and whether use of azoles impact AML outcomes due to their drug-drug interaction with VEN.[5] In this study, we show that AFP use and incidence of IFIs at our institution were low, and that AFP use did not decrease the rate of IFIs and is not associated with OS.

The use of AFP for VEN/HMA-treated patients is extrapolated from the use of AFP for leukemia patients treated with intensive anthracycline-based regimens. In patients receiving intensive chemotherapy, IFIs can be frequent and life-threatening, with reported incidences as high as 36%[15–17] and an infection-related mortality rate exceeding 50% depending on the fungal species.[18–20] Recurrent, protracted neutropenia arising from disease and/or treatment is a major risk factor for IFIs,[13,21] and this is a prominent shared toxicity for VEN/HMA-treated patients,[22] since responding patients frequently continue VEN/HMA indefinitely if they are not candidates for curative allogeneic HCT. On the other hand, VEN/HMA is associated with less mucosal disruption than intensive induction chemotherapy, which may lower the risk of IFI. The true risk of IFIs in VEN/HMA treated patients therefore cannot be determined a priori.

Previously published reports have reported low incidence of IFIs in VEN/HMA treated patients. In the phase 1b study of VEN/HMA for newly-diagnosed AML patients, Grade 3/4 fungal infections were seen in only 8% of patients.[23] In a single-center retrospective study that included 55 newly-diagnosed AML patients, just 5% of these patients experienced an IFI.[4] Notably, patients in both reports were frequently treated with AFP: 46% of patients in the phase 1b study and 68% of the newly-diagnosed cohort in the retrospective report received a prophylactic antifungal agent of some kind, and no comparison of IFI rates between patients who did versus did not receive AFP was performed. Thus, these studies do not elucidate to what extent the low incidence of IFI in VEN/HMA-treated AML patients is attributable to, and therefore warrants, concurrent AFP.

In contrast, our study demonstrates a low rate of IFI in a cohort of 131 VEN/HMA-treated patients who were primarily not treated with AFP. The overall incidence of patients who developed any IFI (possible, probable, or proven infection) was 13.0%. This is in line with the 17% incidence noted by another single-institution study in which AFP use was infrequent.[24] In addition, we report no difference in the incidence of possible, probable, or proven IFIs based on AFP use (p = 0.74). When considering only probable or proven IFIs as is typically done in the infectious disease literature,[4] the incidence of fungal infection with and without AFP use was 0% and 3.7%, respectively (p=>0.99). Our findings suggest that the incidence of IFI can be low even in the absence of AFP use, and calls into question the benefit of using AFP for all VEN/HMA-treated AML patients. In whom might AFP be beneficial? Our sample size limits subgroup analysis, but our competing risk model for time to IFI suggests patients who are unfit (HR 4.2) and who have poorly-prognostic TP53-mutated disease (HR 3.1) could benefit, and this should be confirmed in future studies. In particular, the association of IFIs to TP53-mutations may be related to reports of immune dysregulation in TP53-mutated myeloid neoplasms.[25,26] Our model also noted older age to be associated with a lower hazard of developing IFI, but the hazard reduction was small (HR 0.93) and unlikely to be clinically meaningful.

Although the benefit of AFP may be questionable, we found that concurrent VEN dose-reduction at least does not appear to compromise AML outcomes in the real-world setting. Prior study on this issue has been scant.[3] A pharmacokinetic study first described that VEN dose-reduction was necessary when given concurrently with CYP3A4 inhibitors such as posaconazole,[5] but the clinical impact of this dose-reduction on AML outcomes has only been investigated as a part of the initial phase 1b study of VEN/HMA.[6] In this latter study, a small, 12-patient cohort received posaconazole prophylaxis concurrently with VEN dose-reduced from 400mg to 100mg or 50mg. These patients achieved a composite CR/CRi rate of 67%, comparable to the 60% CR/CRi rate attained by the remaining study participants who received non-azole AFP (n=45). Our study contributes substantially more observations on AML outcomes, and we show no difference in median OS between patients who did versus did not receive AFP (8.1 months vs 12.5 months; p=0.14), although this may have been limited by our sample size.

Of note, we recorded lower rates of CR/CRi in the AFP cohort (Table 2), but we hypothesize this is due to the higher frequency of antecedent hematologic malignancies requiring prior HMA or HCT in the cohort (Table 1), reflecting more resistant disease. In addition, we note that the median OS for our entire cohort (11.4 months) is less than that reported in the randomized study of VEN/HMA (14.7 months).[1] This is likely because our cohort, in contrast to that of the randomized study, included patients who received prior treatment for an antecedent hematologic malignancy, and patients for whom VEN/HMA was initiated for factors other than lack of fitness, such as TP53-mutated disease (n=31), non-de novo AML ontogeny (n=6), or prior HCT (n=6), reflecting real-world patterns of patient selection for this regimen. Finally, our multivariable survival model provides a sobering reminder that TP53-mutated disease remains a major driver of poor long-term survival, and that, at least in clinical contexts where the incidence of IFI is low, AFP has comparatively little impact on survival outcomes.

While providing important insights, as with any retrospective study, our study has limitations. One example is the use of AFP per clinician’s discretion rather than per standardized criteria, which can lead to selection bias and patient heterogeneity (Table 1). We made efforts to address this by using a well-annotated patient database and employing multivariable analysis, but uncaptured biases likely remain. Secondly, the overall low incidence of IFIs in our study cohort limited our assessment of IFI risk factors in a multivariable time-to-IFI model, such as the relative impact of AFP use (Table S2), AML ontogeny, or HMA regimen. Investigation of these and other IFI risk factors would be better served by studies at centers where IFIs are more prevalent. Similarly, our findings regarding AFP usage may not be generalizable to institutions that experience a higher incidence of IFIs.

In conclusion, AFP use and incidence of IFI were overall low at our institution. Receiving azole AFP was not associated with a fewer suspected or confirmed IFIs, or with change in OS. At an institution where the incidence of fungal infections is low, there presently does not appear to be a role for antifungal prophylaxis in newly-diagnosed AML patients receiving VEN/HMA.