Idelalisib given front-line for treatment of chronic lymphocytic leukemia causes frequent immune-mediated hepatotoxicity

A major recent advancement in therapy for chronic lymphocytic leukemia (CLL) has been the development of small-molecule inhibitors of the p110 δ isoform (p110δ) of phosphatidyl-inositol-3-kinase (PI3K). These inhibitors block signaling from a variety of prosurvival B-cell surface receptors that converge on the PI3K pathway. Idelalisib is a highly selective oral inhibitor of p110δ that, in combination with anti-CD20 monoclonal antibodies, has shown efficacy for the treatment of relapsed/refractory CLL. Even in heavily pretreated patient populations, responses to these drug combinations are both frequent and durable, with overall response rates of 70% to 80% and median progression free survival in the range of 18 to 19 months.^1-3  This success in relapsed, refractory disease has prompted evaluation of idelalisib in the upfront setting, with particular interest for older patients who may have multiple other medical comorbidities. Given this, tolerability is crucial. Major toxicities of p110δ inhibition with idelalisib are enterocolitis, transaminitis, and pneumonitis. In the relapsed/refractory setting, these toxicities have been manageable, with grade ≥3 diarrhea/colitis seen in ∼14% of patients, grade ≥3 transaminitis in 14%, and any-grade pneumonitis in 3%.⁴  However, the rates of such toxicities when idelalisib is used as upfront therapy for previously untreated disease have not been well established. A phase 2 study has examined front-line idelalisib in patients age 65 or older, with rates of grade ≥3 diarrhea/colitis up to 42% and grade ≥3 transaminitis up to 23%.^5,6  These data suggest that toxicity rates may be higher in the front-line setting.

A better understanding of the pathogenesis of these adverse events may facilitate treatment or avoidance of them. An autoimmune mechanism for idelalisib-associated enterocolitis has been hypothesized. Mice with genetic inactivation of p110δ develop an autoimmune colitis.⁷  Histopathologic data from patients with idelalisib-associated colitis show an intraepithelial lymphocytosis.^8,9  Some patients with late-onset idelalisib-associated diarrhea that does not respond to antidiarrheal or empiric antimicrobial therapy are anecdotally responsive to steroids.¹⁰  It is unknown if idelalisib-related transaminitis occurs via a similar mechanism.

Given the known efficacy of idelalisib in relapsed/refractory CLL, but also given concern about its toxicities, we report here an initial safety analysis of a phase 2 clinical trial of front-line idelalisib used in combination with the anti-CD20 monoclonal antibody ofatumumab, with the finding that an immune-mediated transaminitis in this setting is frequent and often severe.

The study was initiated on June 16, 2014 (study 13-309; www.clinicaltrials.gov identifier NCT02135133). Of the 24 patients enrolled in the trial, 23 had CLL, and 1 patient was later shown to have lymphoplasmacytic lymphoma and removed from study. The target enrollment was 50 subjects between the Dana-Farber Cancer Institute and Massachusetts General Hospital Cancer Center, but all upfront combination studies of idelalisib were discontinued by the sponsor in March 2016 due to safety concerns. This analysis is based on data up to November 2015. Institutional review boards at each study site approved the protocol. All authors had full access to study data and were involved in data interpretation, manuscript preparation, revision, and final approval.

Eligible patients were older than 18 years and had previously untreated CLL with an indication for treatment according to the 2008 International Workshop on CLL guidelines.¹¹  Subjects were required to have measurable disease (lymphocytosis >5000 cells/µL or palpable/computed tomography–measurable lymphadenopathy > 1.5 cm) and an Eastern Cooperative Oncology Group performance status of 0 to 2. Exclusion criteria included the following: prior systemic therapy for CLL, serum creatinine ≥2.0 times the upper limit of normal (ULN), total bilirubin ≥1.5 ULN, alanine aminotransferase ≥2.5 ULN, alkaline phosphatase ≥2.5 ULN, positive test result for active hepatitis B or C, active chronic infections requiring treatment, and clinically significant cardiovascular disease. All patients provided written informed consent.

A schematic of the trial design is provided in supplemental Figure 1 (available on the Blood Web site). For the first 56 days, subjects received 150 mg idelalisib twice daily, followed by 6 months of idelalisib plus ofatumumab combination therapy, then followed by idelalisib monotherapy continued indefinitely until progression or toxicity. The initial dose of ofatumumab was 300 mg; all subsequent doses were 1000 mg. Infection prophylaxis was initially optional, but the protocol was subsequently amended to require prophylaxis against Pneumocystis jirovecii pneumonia and herpes simplex virus/varicella-zoster virus after 2 cases of P jirovecii pneumonia occurred.

After identification of hepatotoxicity as a frequent adverse event, the trial protocol was modified in September 2014 to enhance early recognition of transaminitis and standardize treatment. Liver function tests were monitored twice weekly from week 3 to 16. The development of grade 1 transaminitis was treated with 40 mg prednisone daily. Idelalisib was held for the development of grade 2 transaminitis or worsening of grade 1 transaminitis while on steroids. Grade ≥3 transaminitis was treated with 1 mg/kg prednisone in addition to discontinuation of idelalisib. For patients with grade ≥3 transaminitis without immediate response to steroids, mycophenolate mofetil was considered.

Blood samples were obtained from enrolled subjects and processed by the CLL Research Consortium Tissue Core at the UC San Diego Moores Cancer Center. Ficoll-Hypaque density-gradient centrifugation was used to obtain mononuclear cells. ZAP-70 status and immunoglobulin heavy chain variable (IGHV) mutation status were assessed by the Consortium Tissue Core as per established criteria (see Rassenti et al^12,13  for ZAP-70 and Ghia et al,¹⁴  Widhopf et al,¹⁵  Giudicelli et al,¹⁶  and Lefranc et al¹⁷  for IGHV). Sequences with <98% homology to the corresponding germline IGHV gene were considered mutated.

Peripheral blood mononuclear cells were isolated from 16 subjects at baseline, 15 subjects at day 28 (±14 days, depending on when toxicity developed), and 5 subjects at day 130 (±21 days, depending on when toxicity developed). At these time points, the patients were experiencing toxicity, but the drug had not yet been held or steroids initiated. Mass cytometric (CyTOF) analysis was performed with a panel of monoclonal antibodies targeting 26 surface-membrane and 9 intracellular markers. The Wilcoxon matched-pairs signed rank test was used to compare percentages of T-cell subsets obtained from CyTOF analysis as well as cytokine concentrations; any samples without a matched baseline time point were not included in significance calculations, and samples with values from day 28 and day 130 were used twice.

Cytokine analysis was performed on serum collected from subjects at the indicated time using the Magnetic Luminex Performance Assay (catalog number FCSTM03-13, R&D Biosystems). Each sample was analyzed in duplicate. Concentrations reported are the average of all values. Mann-Whitney U test was used for statistical comparison.

All patients who received any amount of study treatment were included in the analysis. The median time on therapy was 7.7 months (range, 0.7-16.1 months), and median follow-up time was 14.7 months (range, 1.2-16.8 months). All reported P values are 2 sided, and no adjustments have been made for multiple comparisons.

At the time of data cutoff, 24 patients had enrolled. Baseline subject characteristics are provided in Table 1. The 24 subjects enrolled had a median age of 67 years (range, 58 to 85 years) and included 6 women and 18 men. Seventeen subjects (71%) had high-risk Rai stage 3-4 disease, and 29% had bulky lymphadenopathy defined by the presence of at least 1 lymph node ≥5 cm. From the 21 patients with bone marrow biopsy specimens at enrollment, lymphocytes comprised a median of 80% (range, 35% to 95%) of the intertrabecular space. Two patients (8%) had del 11q, and an additional four patients (17%) had either del 17p, a TP53 mutation, or both.

Multiple subjects developed severe hepatotoxicity. In a representative index case, the patient suddenly developed a grade 3 alanine aminotransferase (ALT) and asparate aminotransferase (AST) elevation on day 28 of idelalisib monotherapy (Figure 1A). The drug was stopped. Despite the drug being held, the transaminitis worsened, reaching a maximum AST of 1251 U/L and ALT of 2237 U/L on day 35. On day 34, the subject underwent a liver biopsy, and on day 35, steroids were initiated. The liver function tests normalized after 3 weeks of steroid treatment.

Cumulatively, 19 subjects (79%) experienced at least grade 1 ALT or AST elevation during the study, with 13 subjects (54%) experiencing grade ≥3 transaminitis. The median time to initial development of any grade ≥2 transaminitis was 27 days (range, 14-133 days). This was before the initial administration of ofatumumab on day 56, exonerating ofatumumab as the cause. At week 5, the time of maximum incidence, 11 subjects (46%) had an elevated ALT, with 5 subjects (21%) experiencing grade 3 ALT elevations and 3 subjects (13%) experiencing grade 4 ALT elevations (Figure 1B). The fraction of subjects with ALT abnormalities decreased over time, but this reflected active intervention rather than self-resolution of the process. The pattern of liver injury was hepatocellular, with more severe elevations in the transaminases compared with alkaline phosphatase and bilirubin (supplemental Figure 2). Patients could be divided into 4 groups based on the adverse effects they experienced (Figure 1C). In addition to the 12 subjects who developed early grade ≥2 transaminitis around day 28 of idelalisib therapy, a second group of 4 subjects developed a delayed grade ≥1 hepatotoxicity around day 130. A third group of subjects did not experience hepatotoxicity but discontinued the study due to other adverse events (colitis and rash). A fourth group of subjects tolerated the therapy well, with 4 subjects remaining on therapy for >1 year without experiencing transaminitis, pneumonitis, or colitis.

The incidence of any grade adverse event, regardless of whether the event was considered related to the study treatment, was 100%. The incidence of any adverse event of grade 3 was 50% and any adverse event of grade 4 was 38% (Table 2). Transaminase elevations were the most frequent toxicity. Diarrhea and colitis were also frequent, with 46% of patients experiencing diarrhea/colitis (17% grade ≥3). The median time to onset of grade ≥2 diarrhea/colitis was 228 days (range, 100-392 days). Pneumonitis was seen in 13% of patients (8% grade ≥3), with a median time to onset of 117 days (range, 40-223 days). Hematologic toxicities included anemia (8%, with 4% grade ≥3), neutropenia (46%, with 29% grade ≥3), and thrombocytopenia (8%, with 0% grade ≥3). Opportunistic infections included 2 cases of P jirovecii pneumonia, 1 case of Aspergillus pneumonia, 1 brain abscess due to an unknown organism (thought likely to be P jirovecii) that resolved on trimethoprim/sulfamethoxazole, 1 case of cytomegalovirus colitis, and 1 case of herpes simplex virus esophagitis. There were no deaths on study or during the follow-up period.

A younger age was a significant risk factor for the development of early hepatotoxicity (Figure 2A). The median age of subjects who did not have any hepatotoxicity was 72 years, whereas the median age of subjects who did have early hepatotoxicity was 61 years (P = .02). All subjects under age 65 required systemic steroids for immune-mediated toxicities at some point during the trial. A total of 75% of patients with mutated IGHV experienced early hepatotoxicity, whereas only 25% of patients with unmutated IGHV experienced early hepatotoxicity (P = .039). The combination of age and IGHV status was highly predictive of early hepatotoxicity, with an area under the receiver-operating characteristic curve of 0.9762 (Figure 2B). Measurements of disease burden, including β₂-microglobulin, LDH, absolute lymphocyte count, Rai stage at baseline, intertrabecular bone marrow involvement at screening, and computed tomography–assessed lymph node size burden were not predictive of early hepatotoxicity (supplemental Figure 3).

No viral etiology for the transaminitis could be identified, and no other drug culprits were suspected. All subjects tested negative for hepatitis B surface antigen and hepatitis C antibody at enrollment. Hepatitis B surface antibody alone was positive in 5 subjects, consistent with immunization. Hepatitis B surface antibody and core antibody were positive at enrollment in 2 subjects who developed transaminitis, but one of these subjects was on monthly intravenous immunoglobulin infusions (a possible source of false-positive serologies), and in both cases, the viral load remained undetectable. Upon developing transaminitis, 4 subjects underwent repeat testing for hepatitis A, B, and C, as well as EBV and cytomegalovirus, which in all cases was negative.

In order to determine the cause of hepatotoxicity, 2 patients who developed severe transaminitis underwent liver biopsy. These specimens were compared with specimens taken from a normal liver and from the liver of an otherwise healthy patient with CLL. No significant mononuclear infiltrate was seen in healthy liver, whereas clusters of CD20⁺ periportal lymphocytes and scattered intraparenchymal CD8⁺, perforin-negative cytotoxic T cells were seen in the patient with CLL, consistent with previous reports.¹⁸  However, biopsy specimens from the CLL patients experiencing idelalisib-related transaminitis showed an increased infiltrate of CD8⁺ cytotoxic T cells that also stained positive for perforin, indicating an activated state (Figure 3A). Similar biopsy findings were noted in sections of duodenum taken from a subject on trial experiencing idelalisib-related enteritis, with CD8⁺ cytotoxic lymphocytes infiltrating the lamina propria (Figure 3B).

All cases of transaminitis resolved by holding the drug, initiating immunosuppression (1 subject required mycophenolate mofetil in addition to steroids), or both. Many subjects were initiated on steroids around day 28, and a second group of subjects required steroids around day 130 (supplemental Figure 4). The median time from initiation of steroids to reduction of transaminitis to grade ≤1 was 8 days (range, 3-21 days). Patients who developed hepatotoxicity but were able to remain on study drug at the time of data cutoff had a short time from onset of transaminitis to initiation of steroids (median, 0 days; range, 0-4 days) and had long steroid tapers (median, 129 days; range, 31-235 days). In comparison, subjects who developed hepatotoxicity and eventually had to discontinue drug had a longer time to steroid initiation (median, 7 days; range, 0-17 days) and shorter steroid tapers (median, 70 days; range, 7-198 days). By the end of the trial, 16 subjects (67%) had received steroids for transaminitis and 19 subjects (79%) had received steroids for presumed autoimmune toxicities.

In 12 instances, subjects who developed grade ≥2 transaminitis were rechallenged with idelalisib after it was held for a median of 21 days (range, 10-43 days). Transaminases had normalized for at least 1 week prior to the reintroduction of idelalisib. Six subjects developed recurrent transaminitis. In all cases of recurrent transaminitis, the elevated liver enzymes developed within 1 to 4 days of re-exposure to idelalisib (for example, Figure 3C). A total of 5 subjects were rechallenged while off steroids, and 4 developed recurrent transaminitis (grade 2, 1; grade 3, 2; and grade 4, 1). Seven subjects were rechallenged while on steroids, and 2 developed recurrent transaminitis (grade 3, 1; grade 2, 1). This protective effect of steroids was also noted when all episodes of recurrent transaminitis, pneumonitis, or colitis were examined in aggregate. In total, after an initial episode of transaminitis, pneumonitis, or colitis, resumption of idelalisib led to 22 subsequent, separate additional grade ≥2 toxicities (including only transaminitis, pneumonitis, or colitis). Of these, 17 occurred while subjects were off steroids (grade 2, 9; grade 3, 7; grade 4, 1), and 5 occurred while subjects were on steroids (grade 2, 3; grade 3, 2; and grade 4, 0).

We measured concentrations of a panel of cytokines in serum obtained from patients at baseline and at day 28, the time of maximum incidence of hepatotoxicity. At day 28, patients experiencing hepatotoxicity, compared with those without hepatotoxicity, had significantly higher levels of the proinflammatory cytokines CCL-3 (median, 1983 pg/mL vs 1429 pg/mL, P = .01; Figure 4A) and CCL-4 (median, 2351 pg/mL vs 303 pg/mL, P = .002; Figure 4B), consistent with an underlying inflammatory process as the cause of the transaminitis. Subjects with and without hepatotoxicity showed no significant difference in the other tested cytokines (CCL-2, CXCL-5, and vascular endothelial growth factor) at baseline or at day 28 (supplemental Figure 5).

Inactivation of p110δ in mice decreases the number and function of regulatory T cells (Tregs).¹⁹  Through the use of mass cytometry, we investigated the effects of idelalisib on T cells in patients exposed to idelalisib. Samples of peripheral blood mononuclear cells had been collected from patients at baseline, day 28, and day 130 (at the time of late hepatotoxicity). At day 28, the fraction of CD3⁺ T cells that were CD4⁺ helper cells trended higher, whereas CD8⁺ cytotoxic cells trended lower (supplemental Figure 6A-B). Decreases in the Treg population were frequent. At baseline, the median percentage of total CD4⁺ T cells that were CD3⁺CD4⁺ FoxP3⁺CD25^hi Treg cells was 5.8%. This decreased to a median of 4.0% after 28 days on therapy and 2.8% after 130 days on therapy, although these decreases were not statistically significant (Figure 4C). Out of 19 matched pairs, 13 patients (68%) experienced a decrease in Treg percentage while on therapy, with a median relative loss of 42% of the Treg fraction. When patients were stratified by toxicity at the time of sample collection (including those with transaminitis or colitis), those without toxicity had a median Treg percentage of 5.7%, whereas those experiencing toxicity had a significantly lower median Treg percentage of 2.3% (p=.03, Figure 4D). In patients experiencing toxicity on idelalisib, there was a trend toward a lower Treg:CD4⁺ ratio (supplemental Figure 6C).

Idelalisib is a p110δ inhibitor that has shown impressive efficacy for the treatment of relapsed, refractory CLL. Here, we found that idelalisib used as upfront therapy for CLL caused an early, severe (grade ≥3) hepatotoxicity in 54% of patients. Subjects who are younger and subjects with IGHV mutated disease were more likely to experience this early hepatotoxicity. Steroids were effective at treating the transaminitis once it developed, and the incidence of recurrent hepatotoxicity was lower if subjects were receiving steroids at the time of reintroduction of the study drug. In keeping with preclinical data, idelalisib caused a reduction in the number of Tregs in the peripheral blood, particularly in patients experiencing toxicities.

Multiple lines of evidence point to an autoimmune mechanism as the cause of the hepatotoxicity seen. Perhaps surprisingly, younger age was identified as a risk factor for early hepatitis, consistent with the view that younger subjects may have a more robust immune system than older subjects. An activated T-cell infiltrate is seen on liver biopsy specimens of patients who experienced idelalisib-related hepatitis, but not in otherwise healthy patients with CLL. Immunosuppression with steroids (and, in a single case, mycophenolate mofetil) was effective at treating the transaminitis and decreasing the incidence of subsequent transaminitis upon re-exposure to the drug. The observation that the transaminitis takes weeks to develop but then rapidly recurs upon re-exposure to idelalisib is suggestive of liver injury secondary to an adaptive immune response.²⁰  Cytokine analysis demonstrated increased levels of CCL-3 and CCL-4, which are known to play proinflammatory roles in murine²¹  and human²²  immune-mediated hepatitis. CyTOF analysis of the peripheral blood did show a decrease in Tregs, which would correlate with an inflammatory state.

Preclinical evidence also supports the conclusion that on-target inhibition of p110δ by idelalisib can cause autoimmune toxicities and furthermore that the mechanism underlying this could be due to effects on Tregs. Mice with genetic inactivation of p110δ develop an autoimmune colitis characterized by numerous intraepithelial lymphocytes.^7,23  Systemic p110δ inactivation has broad effects on the immune system, decreasing B-cell function and number as well as decreasing the ability of naive T cells to differentiate into both Th1 and Th2 subtypes.^7,24,25  However, on an organismal level, this impaired effector cell function is often counterbalanced by inhibition of Tregs, as p110δ is also critical for the survival and function of Tregs.¹⁹  For example, despite the fact that both CD8⁺ and CD4⁺ T cells from p110δ-deficient mice demonstrate reduced production of cytotoxic mediators and interferon-γ, respectively, mice with genetic inactivation of p110δ are resistant to tumorigenesis. Adoptive transfer experiments showed that p110δ inactivation in Tregs is both necessary and sufficient to confer this resistance to tumor growth.²⁶  In agreement with this, germline genetic mutations that disrupt Treg function in mice and humans, including FOXP3, lead to autoimmune syndromes with hepatitis, enteritis, and pneumonitis.^27,28 

The rates of grade ≥3 transaminitis reported here are higher than rates previously reported for this drug, as summarized in Table 3. In the phase 1 trial, subjects were the most heavily pretreated, with a median of 5 prior therapies, and the rate of grade ≥3 hepatotoxicity was the lowest, at 2%.²⁹  When toxicity data were aggregated from 8 clinical trials examining idelalisib used as later-line treatment of indolent B-cell malignancies, the rate of grade ≥3 transaminitis was 14%.⁴  In addition to the trial presented here, one other published trial examined idelalisib in the upfront setting, given to patients age 65 or older. In the cohort treated with combination rituximab therapy, 67% of subjects experienced a transaminase elevation of any grade (23% grade ≥3)⁶ ; in the cohort treated with idelalisib alone, 24% experienced a transaminase elevation (22% grade ≥3).⁵  Thus, as the median age and number of prior therapies increased, the frequency of immune-mediated adverse events decreased. This may reflect the long-lasting effects that some CLL therapies, and the disease itself, can have on the immune system, as well as an accruing immune senescence with age. For example, for at least 2 years after fludarabine therapy, absolute CD4⁺ and CD8⁺ T-cell counts remain less than half pretreatment levels.³⁰  In keeping with these observations, when idelalisib was given to healthy young volunteers, toxicity was frequent, with 5 subjects out of 24 (21%) experiencing grade 3 transaminitis after only 7 to 9 days on 150 mg idelalisib twice daily.³¹  The association between IGHV mutation status and early hepatotoxicity may reflect a unique interaction between mutated IGHV neoplastic cells and T-cell subsets.³²  For example, 2 studies have previously reported that patients with mutated IGHV CLL have reduced numbers of Tregs.^33,34  We did find a lower baseline median percentage of Tregs in our patients with IGHV mutated disease (data not shown), although the difference was not statistically significant.

Transaminitis is seen when other drugs within this class are tested in the upfront setting, as would be expected if hepatotoxicity is due to on-target p110δ inhibition. TGR-1202 is a p110δ inhibitor with hepatotoxicity seen in 2% of patients treated for relapsed, refractory disease.³⁵  However, a recent trial combining TGR-1202 with obinutuzumab and chlorambucil for CLL treatment had rates of grade ≥3 transaminitis of 28%.³⁶  Eighty-three percent of the subjects in this 18-patient trial were treatment naive. Our group recently reported preliminary results of the phase 1 portion of a trial of the p110δ/γ inhibitor duvelisib in combination with fludarabine, cyclophosphamide, and rituximab for the upfront treatment of CLL. We observed a 29% rate of grade ≥3 transaminitis, occurring at time points similar to those reported here.³⁷  In this duvelisib trial, the combination of p110δ inhibition with chemoimmunotherapy or concomitant p110γ inhibition may have decreased the overall rates of autoimmune toxicity.

Limitations of the study include the small number of patients enrolled and the correspondingly small number of samples available for correlative studies. Our analysis here focuses on the characteristics of the early hepatotoxicity seen at day 28 and may not be applicable to other toxicities seen with the drug, including enterocolitis, pneumonitis, and the delayed hepatotoxicity occurring around day 130. Correlative studies regarding the mechanism of drug-mediated hepatotoxicity were analyzed retrospectively, and while these findings are important for future hypothesis generation, they cannot show causation.

While this manuscript was in preparation, Gilead closed seven randomized trials of idelalisib in B-cell malignancies (5 in treatment-naive patients) due to an excess of infectious deaths. In our study, opportunistic infections were noted prior to mandating prophylaxis. Further studies will be required to determine any relationship between these infectious toxicities and the autoimmune toxicities seen here, but both may relate to the immunologic effects of inhibiting p110δ. As additional drugs are developed within this class, subjects should be closely monitored for infection as well as transaminitis (and other autoimmune phenomena), with a low threshold for infectious prophylaxis for the former and immunosuppressants for the latter, based on the likely immune-mediated origin of this toxicity.

The authors thank all the patients and their families who participated in this trial and donated extra blood and bone marrow for analysis.

This study is an investigator-initiated trial run by J.R.B. and funded by Gilead and Novartis; both companies provided their study drug at no cost. Correlative studies were funded by the Dana-Farber Cancer Institute Medical Oncology Grant program 2015, as well as the Melton Family Fund for CLL Research and the Susan and Gary Rosenbach Fund for Lymphoma Research. This work was supported in part by a grant from the National Institutes of Health, National Cancer Institute (PO1-CA81534) of the CLL Research Consortium (L.R. and T.J.K.).

Contribution: J.R.B. designed the trial; B.L.L., S.N.K., T.R.M., E.A.M., J.R., and J.R.B. performed the research and analyzed data; J.F. and J.H. collected and analyzed data; M.S.D., D.C.F., A.S.F., C.A.J., P.A., J.S.A., J.E.A., and J.R.B. enrolled patients; L.R., S.F., and T.J.K. managed sample collection for correlative analyses; H.T.K., S.K., and T.M. performed statistical analyses; B.L.L. wrote the manuscript with input from J.R.B.; and all authors approved the final version of the manuscript.

Conflict-of-interest disclosure: J.R.B. has served as a consultant for Gilead, Infinity, Janssen, Pharmacyclics, and Roche/Genentech. M.S.D. has served as a consultant for Gilead, Infinity, Janssen, Pharmacyclics, TG Therapeutics, Abbvie, and Roche/Genentech. J.S.A has served as a consultant for Gilead, Infinity, and Pharmacyclics. The remaining authors declare no competing financial interests.

The current affiliation for T.R.M. is Instituto de Medicina Molecular, Faculdade de Medicina da Universidade de Lisboa, Lisbon, Portugal.

The current affiliation for J.F. is Sackler School of Medicine, Tel Aviv University, Tel Aviv, Israel.

Correspondence: Jennifer R. Brown, Dana-Farber Cancer Institute, Mayer Building Room 226, 450 Brookline Ave, Boston, MA 02215; e-mail: jennifer_brown@dfci.harvard.edu.