Midostaurin: its odyssey from discovery to approval for treating acute myeloid leukemia and advanced systemic mastocytosis

Midostaurin (CGP41251; PKC412) is a multikinase inhibitor recently approved for 2 indications in adult patients: (1) newly diagnosed acute myeloid leukemia (AML) with fms-like tyrosine kinase 3 (FLT3) mutations and (2) advanced systemic mastocytosis (SM).^1,2  Its clinical development proceeded in 3 stages. First, based upon activity against protein kinase C (PKC), it was evaluated in clinical trials as monotherapy in chronic lymphocytic leukemia (CLL) and melanoma, and as a single agent or in combination with chemotherapy against solid tumors without preselection according to tumor genotype.^3-7  These studies provided safety and pharmacokinetic data for the compound. Secondly, supported by its vascular endothelial growth factor receptor and angiogenesis inhibition, it was evaluated in patients with diabetes-related macular degeneration.^8,9  Stages 1 and 2 demonstrated inadequate efficacy to warrant further clinical development. Following the identification of additional targets, stage 3 focused on inhibition of FLT3 in AML and KIT proto-oncogene receptor tyrosine kinase (KIT) in advanced SM.^10,11  In these latter studies, patients were selected based on a genotype associated with a clearer oncogene addiction (to FLT3 and KIT, respectively).^12,13 

Midostaurin and its active metabolites target mutant forms of FLT3 and KIT along with additional protein kinases implicated in leukemogenesis.¹⁴  The pivotal AML trial focused on patients bearing FLT3 mutations and trials in advanced SM on patients with activating KIT D816 mutations. For AML, a key decision was to focus early on first-line therapy (in combination with chemotherapy) instead of taking the traditional approach of focusing first in relapsed/refractory disease. Midostaurin demonstrated a favorable safety profile in both indications. In registrational trials, midostaurin improved overall survival (OS) in AML and showed compelling evidence of disease control as a single agent in advanced SM.^10,11,15  This review describes the developmental journey of midostaurin from its initial discovery as a PKC inhibitor to its current role as a multikinase inhibitor (Table 1).

Midostaurin emerged as a development candidate from a drug discovery program aimed to improve on the selectivity of staurosporine toward PKC,¹⁶  which was considered to be an attractive therapeutic target in oncology and several other indications.^17-19  Staurosporine, an alkaloid first isolated from Streptomyces staurosporeus,^20-22  was one of the first compounds shown to inhibit cell proliferation through protein kinase inhibition.^23-26  In 1986, staurosporine was reported to potently inhibit the enzymatic activity of PKC at low nanomolar concentrations.²⁷  The availability of adequate amounts of staurosporine, via fermentation within Ciba-Geigy, enabled medicinal chemists to embark on a program to discover novel, potent, and selective inhibitors of PKC. However, at that time, the idea of moving any inhibitor of kinase signaling from bench to clinic was met with apprehension, given the difficulty in reaching an appropriate level of target selectivity.^28,29 

Despite this skepticism surrounding the feasibility of developing kinase inhibitors as tolerated drugs, the first kinase inhibitor to reach the market was fasudil (Japan, 1995), approved as a vasodilator^30,31  and subsequently found to act by RhoA/Rho kinase inhibition.^32,33  In 2000, the natural product sirolimus, a mammalian target of rapamycin inhibitor,³⁴  was approved in the United States for preventing kidney transplant rejection.²⁰  Imatinib, the first protein kinase inhibitor approved in oncology, was indicated for the treatment of chronic myeloid leukemia (CML) and later gastrointestinal stromal tumors.³⁵ 

Midostaurin was first synthesized by Giorgio Caravatti in 1986.³⁶  Studies to investigate its potential as a PKC inhibitor revealed that it inhibited cell proliferation by interfering with cell-cycle activity.^37,38  It also inhibited solid tumor growth in murine xenograft models.³⁸  Furthermore, midostaurin demonstrated antiproliferative activity in a range of solid tumor lines, including lung, colon, breast, melanoma, and glioblastoma.¹⁷  A key aspect in the clinical development of midostaurin, a highly insoluble drug, was the identification of a microemulsion formulation that allowed for rapid absorption and high bioavailability.³⁹  Following oral administration, midostaurin is metabolized primarily by the cytochrome P450 3A4 pathway to produce 3 major active metabolites.^9,40  Over time, it has become clear that, like midostaurin, these metabolites target not only PKC but also many other serine-threonine and tyrosine kinases (Figure 1).¹⁴ 

First-in-human studies commenced in 1994.⁷  Low-grade gastrointestinal and hematologic toxicities were frequent but manageable. A maximum tolerated dose was not formally identified, because dose escalation was limited by the number of capsules to be ingested daily. A dosage of 150 mg/day was determined to be adequate for further phase 2 testing as a single agent. Subsequent studies of midostaurin in combination with 5-fluorouracil in patients with solid tumors,⁵  in combination with gemcitabine and cisplatin in non–small cell lung cancer,⁴  and as monotherapy in metastatic melanoma³  and low-grade lymphoproliferative disorders⁶  showed that the preclinical effects of midostaurin on these tumor types did not correlate with clinical outcomes (a modest decrease in lymphocyte counts was observed in CLL). However, the pharmacokinetic properties of midostaurin were characterized, showing that biologically relevant concentrations of the drug could be achieved at tolerated doses.

Based on its activity against vascular endothelial growth factor receptor kinase and vascular endothelial growth factor–mediated effects in mouse models,¹⁷  midostaurin was studied as an angiogenesis inhibitor for the treatment of diabetic retinopathy between 1999 and 2002.^8,9  Gastrointestinal toxicity was challenging for these patients, and the degree of efficacy, as assessed by macular edema and visual acuity, was insufficient at tolerated doses to support further development in diabetes.⁸ 

In CML, the BCR-ABL1 tyrosine kinase inhibitor imatinib revolutionized the treatment of cancer by introducing the new paradigm of targeting the underlying genetic abnormality driving a malignancy.⁴¹  In the mid-1990s, translational scientists reported that ∼30% of patients with AML had an activating mutation in the transmembrane tyrosine kinase FLT3.^42,43  The more common internal tandem duplication (ITD) mutation was found to be particularly ominous, leading to rapid relapse and short survival with standard chemotherapy.⁴⁴ 

Researchers at Novartis and Dana-Farber Cancer Institute had been closely interacting under an academic/industrial collaboration established in 1990. Within this established collaborative environment, Paul Manley selected a panel of receptor tyrosine kinase inhibitors as potential FLT3 inhibitors and sent these to James Griffin’s group at Dana-Farber Cancer Institute in 2000 to evaluate their effects on FLT3-dependent cells. Ellen Weisberg tested these compounds in the Griffin laboratory using murine Ba/F3 cells rendered growth-factor independent via transfection with constructs that encoded either ITD or tyrosine kinase domain (TKD) point mutations in the FLT3 kinase. Midostaurin had potent antiproliferative activity by downregulating recombinant FLT3-catalyzed transphosphorylation in vitro and inhibiting the proliferation of the Ba/F3-FLT3-ITD cells at concentrations <10 nM, without affecting the viability of the parental cells at concentrations ≤100 nM.⁴⁵  Reduced cell growth was determined to be due to induction of apoptosis and cell-cycle arrest. Subsequently, the metabolites of midostaurin were also found to have activity against FLT3 at concentrations substantially below those attained at steady state following oral administration of therapeutic doses in patients.⁴⁵ 

Around the same time, Gary Gilliland’s laboratory developed a murine myeloproliferative neoplasm model by transfecting murine hematopoietic stem cells with mutant FLT3 constructs and then transplanting these cells into sublethally irradiated mice. When mice underwent transplant with bone marrow transduced with FLT3-ITD, all the animals dosed with an orally bioavailable formulation of midostaurin survived at 90 days compared with ∼20% of control animals (Figure 2).⁴⁶  Of note, this model did not produce morphologic AML but rather a highly proliferative and aggressive neoplasm characterized by an overproduction of mature cells, thereby verifying the ability of mutated FLT3 to cause a proliferative thrust but also suggesting that it may not be central to the development of leukemia.

Preclinical studies such as these supported proceeding with a series of clinical trials of FLT3 inhibitors, although target selectivity and high protein binding were recognized as issues.^7,47  Midostaurin had an advantage over other investigational agents because a safe and tolerable dose had already been established in patients with solid tumors and diabetes.⁷  Midostaurin and lestaurtinib were each tested in proof-of-concept trials that enrolled patients with advanced FLT3 mutant AML.^48,49  Both trials demonstrated significant lowering of peripheral blood blast counts; however, complete remissions (defined as <5% blasts in the marrow together with adequate hematopoietic recovery) were rare, and this raised questions about the extent of single-agent efficacy. There were uncertainties surrounding whether adequate levels of sustained FLT3 inhibition were present and whether the relative resistance of marrow blasts resulted from their protection by the hematopoiesis-inducing microenvironment. Moreover, it seemed that FLT3 inhibition on its own would not be as effective in AML, as BCR-ABL1 inhibition was in chronic-phase CML, a less genetically diverse disease. Despite this, investigators proceeded with additional trials in AML. A dose-optimization study showed that twice-daily dosing of midostaurin was feasible and likely as effective as the 3-times-daily dosing used in its proof-of-concept trial.⁴⁵  Interestingly, a few patients with wild-type FLT3 (FLT3-WT) AML experienced blast reductions, although responses were more common in those with FLT3-mutated AML.

A key strategic decision in the development of midostaurin in AML was to go straight from the first proof-of-concept single-agent trial in relapsed/refractory AML⁴⁸  to development as first-line treatment in combination with chemotherapy.^10,50  This approach skipped the traditional step of further trials in relapsed/refractory disease. In contrast to midostaurin, lestaurtinib was simultaneously moved into a phase 3 trial in relapsed/refractory FLT3-ITD–positive AML, in which salvage chemotherapy was given with or without lestaurtinib.⁵¹  Response rates and OS were not improved with addition of lestaurtinib to chemotherapy. However, a novel, ex vivo plasma inhibitory assay was developed by Mark Levis, and it demonstrated that a large number of these patients had inadequate FLT3 inhibitory levels in plasma.⁴⁰  This result informed the hypothesis that the optimal setting to test a relatively nonspecific inhibitor would be in newly diagnosed patients AML to avoid the evolved resistance mechanisms which are the greatest challenge in relapsed disease.⁵² 

An opinion developed that the biological activity noted in the single-agent trials of midostaurin in AML could be improved by combining the drug with additional agents such as standard cytotoxic chemotherapy. Subsequently, a phase 1b trial in patients with newly diagnosed FLT3-unselected AML who received standard 7 + 3 (cytarabine plus daunorubicin) induction and high-dose cytarabine consolidation chemotherapy combined with various doses and schedules showed that continuous dosing with midostaurin at either 50 or 100 mg twice daily was not tolerated due to gastrointestinal toxicity.⁵³  Fortunately, schedules of midostaurin at 50 mg twice daily were feasible either on days 1 to 7 and 15 to 22 of the 28-day cycle or on days 8 to 22. Sequential dosing was chosen for further study, because, as noted by Donald Small, concomitant administration of midostaurin and chemotherapy might increase toxicity and perhaps be antagonistic from a cell-cycle standpoint.⁵⁴  A key finding from this trial was that patients with mutant FLT3 experienced comparable outcomes to those in patients with FLT3-WT AML. Although encouraging, this result was based on only 13 patients with FLT3-mutated AML, 4 of whom harbored FLT3-TKD mutations, which have a somewhat better prognosis than FLT3-ITD–positive AML patients.

Investigators then proposed to compare standard chemotherapy to standard chemotherapy plus midostaurin in patients with newly diagnosed AML at the dose and schedule shown to be safe in the phase 1b trial. Debate ensued about whether this trial should stratify patients by FLT3 mutation status or disregard the presence of a FLT3 mutation. A trial that did not stratify would obviate the need for rapid genetic testing at diagnosis, but it would have to be powered to detect a benefit in only the 30% of patients with a FLT3 mutation. The designers decided to develop rapid genetic testing at diagnosis and enroll only patients with a FLT3 mutation. However, limiting enrollment onto the C10603/RATIFY trial to FLT3 mutant AML would allow an assessment of the relative importance of FLT3 inhibition vs off-target effects by comparing the magnitude of benefit from midostaurin in ITD high vs low allelic ratio patients, with the latter being likely less dependent on FLT3 signaling. Also, potentially informative as to the importance of FLT3 inhibition for clinical efficacy, a limited number of correlative studies are currently being performed which assess the correlation of FLT3 autophosphorylation status at entry and outcome. To test the hypothesis that the multikinase inhibitor could potentially be active beyond FLT3 inhibition, a phase 3 trial in patients with AML without a FLT3 mutation is currently being planned.

There were tremendous difficulties in performing a trial enrolling only the 30% of patients with AML harboring FLT3 mutations. Fortunately, the global leukemia community united in this effort. This undertaking required access to patients from multiple cooperative groups worldwide (Figure 3),⁵⁵  the drug knowledge base, and substantial resources of a large pharmaceutical company, as well as a partnership with the Cancer Therapy and Evaluation Program of the National Cancer Institute. The Cancer and Leukemia Group B (CALGB), now the Alliance for Clinical Trials in Oncology, and Novartis reached an agreement whereby CALGB and Cancer Therapy and Evaluation Program sponsored the trial in the United States and Canada, and Novartis sponsored the trial in other countries. Moreover, to accomplish the necessary FLT3 testing, investigators from 9 academic laboratories in 5 countries harmonized a polymerase chain reaction–based FLT3 mutation testing method with rapid turnaround time (≤48 hours).¹⁰  Overall, 3277 patients with AML were screened between May 2008 and October 2011 to identify the 717 eligible patients with FLT3 mutations enrolled in the CALGB 10603/RATIFY trial, which involved 225 sites in 17 countries.⁵⁵  The crucial cooperation of expert leukemia investigators around the world cannot be overstated.

Even with these enormous logistical hurdles overcome, a trial with realistically achievable goals and end points still had to be designed. The investigative team understood that it was important to perform a prospective, randomized, placebo-controlled trial that included a maintenance phase to maximize the efficacy of midostaurin by keeping the mutant clone suppressed. Although event-free survival (EFS) was originally considered as a potential endpoint, investigators felt that OS would yield a real-world confirmation of the clinical benefit of midostaurin in this patient population, given the likelihood that allogeneic transplant would occur in a sizable minority of patients.

Concerns were expressed throughout the regulatory process that stem cell transplant might obscure any beneficial effect of midostaurin. However, the investigators insisted that the addition of midostaurin to initial chemotherapy could lower the residual disease burden prior to transplant and thus improve long-term outcomes. Moreover, there were concerns that if the primary end point required censoring at the time of transplant, too large a sample size would be required. Similarly, the question of maintenance therapy was also raised, pointing to the options of introducing a second randomization, a 2-by-2 design, or a separate trial. However, such designs would have made an already large study untenable. It was finally agreed to proceed with a primary end point based on OS uncensored for transplant, with final analysis when 75% of events (deaths) had occurred.

Study conduct went surprisingly well. The commitment displayed by European investigators manifested by key scientific contributions and prodigious accrual was essential. Because of the very rapid turnaround of the FLT3 genotyping, it was feasible to rapidly screen and enroll patients with FLT3-mutated AML within 2 to 3 days of presentation. The study was stratified based on the type of FLT3 mutation: FLT3-TKD (relatively more favorable outcome), FLT3-ITD with a high allelic ratio (poor prognosis), and FLT3-ITD with low allelic ratio.⁵⁶  The allelic ratio was defined as the mutant to wild-type fraction and ranged from 0.05 to >1. A planned interim analysis was carried out when 50% of events occurred.¹⁰  After ∼500 patients had enrolled, it became clear that initial study assumptions underestimated the number of patients who would be transplanted in first remission and the number of patients with TKD mutation, both of which would yield less-frequent events. Therefore, an amendment expanded the trial from 500 to 714 patients.

Several years after the trial completed enrollment, it was evident that the expected number of events was not going to occur within a reasonable time frame. The number of deaths plateaued, forecasting the necessary 509 OS events to occur in 2025.¹⁰  Therefore, the statistical plan was amended to allow the study to be analyzed when a sufficient number of EFS events (failure to achieve remission, relapse, or death from any cause) occurred.

When the study was unblinded, the midostaurin arm showed an improved OS in the analyses both censored and uncensored for transplant (Figure 4). Results were consistent across FLT3 subgroups (Figure 5). The data from RATIFY led to approval for midostaurin in adults with newly diagnosed FLT3-mutated AML in the United States, European Union, and other countries (Table 2). In addition, for the purposes of US Food and Drug Administration approval of midostaurin, a companion FLT3 diagnostic test was developed through a partnership between Novartis and Invivoscribe.

Following an initial case report in 1 patient with MCL,⁵⁷  studies in patients with advanced SM were performed. In this disease, midostaurin targets mutant forms of the KIT receptor tyrosine kinase, detected in >80% of patients.^11,56  Two single-arm, phase 2 studies, representing the largest prospective clinical trials program conducted in this rare disorder, demonstrated the efficacy of midostaurin 100 mg twice daily in advanced SM. Midostaurin elicited high rates of response in the A2213 study and the D2201 study (60% and 69%, respectively); responses were durable and associated with reduced disease burden and improved quality of life.^11,15  These studies led to the approval of midostaurin in advanced SM at the same time as the approval in AML.^1,2 

The successful development of midostaurin provides several lessons for clinical trialists, in general and specifically in AML. Although they were not successful in terms of clinical outcomes in solid tumors, CLL, or diabetic retinopathy, the first 2 stages were crucial for having identified tolerable dosing, characterizing pharmacokinetics, and establishing the safety profile of midostaurin. This “early” work enabled investigators to efficiently design trials once AML was selected as a potential indication.

This journey shows the feasibility of conducting a large, international phase 3 trial in a subset of patients molecularly defined using a well-validated test before chemotherapy is initiated. Extensive coordination between clinical trial cooperative groups, the pharmaceutical industry, and governmental agencies (the Cancer Therapy Evaluation Program of the National Cancer Institute in particular) united in a common goal enabled the successful completion of the RATIFY study. Patience was required for industry and cooperative groups to understand each other well and align expectations to foster an open and transparent collaboration. For example, both entities had to be open to using different processes than normal to (1) preserve the independence of academic oversight while being cognizant of industrial requirements, and (2) involve a wide range of regulatory, technical, and scientific disciplines. The logistical support required was considerable but manageable. Additionally, the journey underscores the importance of quickly moving investigation of new therapies to the first-line setting whenever appropriate, considering the safety and efficacy profile of the drug, along with knowledge concerning the molecular target addressed by the new therapy.

For example, with today’s knowledge that the clonal heterogeneity of AML is much higher at diagnosis than at relapse,⁵⁸  the RATIFY data support the hypothesis that a compound targeting multiple kinases relevant to AML would be more appropriate for newly diagnosed disease to capture this clonal heterogeneity. The improved survival observed in RATIFY is in line with that hypothesis as well as with prior in vitro work suggesting that less specific FLT3 inhibitors (eg, midostaurin) have better antiproliferative effect than more specific inhibitors in patient samples taken at diagnosis vs samples taken at relapse.^57,59 

In retrospect, several aspects of this development could have been improved. Since the initial study design a decade ago, much has been learned about AML disease biology, and many important technological innovations have been introduced. For example, although samples have been collected by some participating groups and will be analyzed in the future, measurable residual disease (MRD) and comprehensive genetic profiling were not formal end points. This information could be very useful in better understanding the striking effect of midostaurin in patients who received a transplant in first complete remission immediately after induction and consolidation chemotherapy, which could be due to a lower disease burden prior to transplant. This technology could also be useful to characterize the effect of maintenance therapy. The collection of MRD data are now strongly recommended for clinical trials.⁶⁰  In addition, samples collected during the trial will be used to construct a comprehensive genetic profile for each patient at diagnosis to determine which genetic subpopulations of patients are more likely to benefit from the addition of midostaurin to chemotherapy. For example, data investigating the impact of NPM1 comutations on the outcomes of these patients will be reported shortly. However, including biomarkers and complex genetics in the initial study design for all patients would have added even more complexity to a trial that was already large and complex. The RATIFY trial also illustrates the challenges of selecting appropriate end points. OS clearly remains the gold standard, but it is increasingly difficult to interpret in AML owing to the increase in the use of transplant in first complete remission and the availability of more effective salvage therapies. In addition, OS requires extensive follow-up; in RATIFY, the time from the start of enrollment to the primary analysis took 7 years. Although harboring their own methodological challenges, alternative end points less influenced by postrelapse therapy (eg, EFS) are increasingly used in contemporary phase 3 trials.

Despite limitations, the RATIFY trial represents a shining example of a successful worldwide collaboration among industry, government, and academia. This model may be a paradigm for making progress in orphan diseases, which are increasingly defined by oncogenic drivers that occur even less frequently than FLT3 or KIT. The RATIFY study led to the first targeted therapy for AML with midostaurin benefitting a sizable subset of AML patients with a historically poor prognosis. On the other hand, the 16-year period between the discovery that midostaurin had preclinical efficacy as an FLT3 inhibitor to the approval of the drug was unacceptably long. While carefully conducted sequential clinical trials undoubtedly contributed to this success, a more rapid readout end point than OS would have been ideal. Alternative innovative trial designs are imperative to cope with a rapidly increasing number of potential new therapies available for clinical trial investigation. Smaller randomized controlled trials using a surrogate end point (eg, EFS) or the achievement of an MRD-negative state may be a quicker way to determine the efficacy of multiple novel targeted agents.

The authors thank James Griffin for his critical review of this manuscript. The authors thank the leaders of the participating cooperative groups for their contributions to the RATIFY study: Hartmut Döhner and Arnold Ganser (AMLSG), Sergio Amadori and Francesco Lo-Coco (GIMEMA), Jorge Sierra (CETLAM), Miguel A. Sanz (PETHEMA), Theo de Witte (EORTC), Dietger Niederwieser (OSHO), Gerhard Ehninger (SAL), Frederick R. Appelbaum (SWOG), Martin S Tallman (ECOG), and Joseph Brandwein (NCIC-CTG). The authors also thank the steering committee for the advanced SM trials: Jason Gotlib (Stanford Cancer Institute, Palo Alto, CA), Andrea Reiter (University Hospital Mannheim of the University of Heidelberg, Heidelberg, Germany), Hanneke C. Kluin-Nelemans (University Medical Center Groningen, Groningen, The Netherlands), Cem Akin (University of Michigan School of Medicine, Ann Arbor, MI), and Peter Valent (Medical University of Vienna, Vienna, Austria). The authors appreciate the contributions of Ilene Galinsky, whose assistance regarding study development, education, and trial conduct were essential.

Medical editorial assistance for the parts of this manuscript written by R. Capdeville and P. Manley was provided by Pamela Tuttle, CMPP, of Articulate Science LLC and funded by Novartis Pharmaceuticals.

Contribution: R.M.S., R.A.L., and R.C. designed and performed research; collected, analyzed, and interpreted data; and drafted and approved the manuscript; and P.W.M. designed research, analyzed and interpreted data, and drafted and approved the manuscript.

Conflict-of-interest disclosure: R.M.S. is a consultant for Novartis, Amgen, Celator, Bristol-Myers Squibb, Karyopharm, Janssen, Jazz Pharmaceuticals, Orsenix, Fujifilm, Ono, Pfizer, Celgene, Astellas, Juno, Abbvie, Cornerstone, Seattle Genetics, Arog, Merck, and Sumitomo; has received clinical research support to his institution from Novartis, Celator, Karyopharm, Arog, and Daichi-Sankyo; served on the Data Safety Monitoring Board for Sunesis, Celgene, and Argenix; and received clinical research support to the cooperative group for investigator meetings and data support. R.A.L. is a consultant for Novartis and Astellas. R.C. and P.W.M. are employees of Novartis AG. R.C. has equity ownership of Novartis Pharma AG.

Correspondence: Richard M. Stone, Department of Medical Oncology, Dana-Farber Cancer Institute, Harvard Medical School, 450 Brookline Ave, Dana 2058, Boston, MA 02215-5450; e-mail: richard_stone@dfci.harvard.edu.