The phosphoinositide 3-kinase delta (PI3Kδ) is a member of the class I lipid kinase family based on its sequence homology in kinase domains and its substrate selectivity (1, 2). First isolated from a cDNA library made from monocytes, the PI3Kδ gene is selectively expressed in leukocytes. Signaling through this kinase is via cytokine receptors (3, 4). Subsequent functional studies revealed impaired B and T cell antigen receptor signaling in mouse models which have a nonfunctional kinase allele (4). Development of pan-kinase and delta selective inhibitors marked the beginning of a new era in the treatment of B cell malignancy. Meanwhile, the role of PI3Kδ in Treg function prompted exploration of it as a tool for breaking the immune tolerance to boost antitumor immunotherapy. The science behind the translational efforts was captured in this review (5). Unfortunately, clinical use of this class of drug in management of B cell malignancies and research came to halt when all FDA approvals of this drug class were withdrawn (6). Here, we discuss the initial approval of four PI3K inhibitors. We present preclinical studies supporting use of the inhibitor in modulating the tumor immune microenvironment (TIME) to boost tumor immunity. We highlight the challenges of using such inhibitors in the clinic that led to withdrawal of this class of PI3Kδ inhibitors in hematologic malignancy. Finally, we propose strategies to mitigate the adverse effects while retaining the efficacy in future clinical trials, highlighting potential areas of further exploration. We conclude that exploration of PI3Ki will continue despite this recent setback.

PI3K inhibitors conferred high overall response rate between 57-74% and improvement in progression free survival between 11.0 to 21.5 months in single agent treatment or combination settings. The high response rate was linked to their ability to drive B cells out of their protective lymph nodes or spleen niche (section 3 below). Long term observation using overall survival data however, failed to show advantage over the control arms. The dosing schedules were associated with high incidence of adverse events.

Idelalisib is an orally bioavailable ATP-competitive kinase inhibitor given continuously that targets the PI3K p110 isoform δ (PI3Kδ) with high potency and selectivity. In an open labeled single arm phase 2 study, 125 patients with indolent lymphoma who had not responded to rituximab or alkylating agents or who had recurrence after 6 months were given 150 mg of drug twice daily until disease progression or withdrawal. Overall response rate (ORR), which was the primary end point, was 57% with 6% complete response. Median progression-free survival (PFS) was 11 months. Severe adverse events included neutropenia (27%), increased liver function tests (13%), diarrhea (13%) and pneumonia (7%). As the first-in-class drug, idelalisib’s initial approval for B cell malignancies in July 2014 included indications in relapsed/recurrent CLL, SLL and follicular lymphoma (FL) (7). The approval in FL and in SLL was an accelerated one that mandated post-marketing trials which suffered from slow accrual. The regular approval was then noted to have a death rate due to drug toxicity more than double that of the control arm in a randomized trial in 2016 and FDA mandated a box warning on the label (8). The sponsor voluntarily withdrew idelalisib from the market in 2022 due to inability to complete the confirmation study.

Copanlisib, which is considered a pan-PI3K inhibitor, received accelerated approval in 2017 for relapsed FL. The approval was based on CHRONOS-3, a multicenter, randomized, double blinded, placebo-controlled phase 3 trial of 652 patients with CD20 positive indolent lymphoma which recurred 6–12 months before the last dose of rituximab. Patients were randomized into receiving either copanlisib and rituximab or rituximab and placebo at a 3:1 ratio and were observed for PFS. The drug was given intravenously on days 1, 8, and 15 every 4 weeks. The trial showed a PFS advantage of 21.5 months with copanlisib plus rituximab over rituximab plus placebo (13.8 months). The most common serious adverse events (SAEs) were hyperglycemia (56% vs 8%) and hypertension (40% vs 9%). The drug was well tolerated overall with only one death from pneumonitis (9). Interim analysis, however, showed no overall survival (OS) advantage. In 2021, the agreement between the FDA and the drug maker was voluntarily withdrawn based on CHRONOS-3 results pending new study results from the ongoing CHRONOS-4. Unfortunately, CHRONOS-4 was not a positive study in relapsed indolent lymphoma showing no PFS or OS benefit in patients receiving rituximab and bendamustine plus copanlisib compared to rituximab and bendamustine plus placebo (32.9 m vs 33.3 m; HR: 1.13; 95% CI: 0.80–1.60; P = 0.71) (10). Indeed, the curve of OS showed potential harm in patients receiving copanlisib compared with placebo. Copanlisib was voluntarily withdrawn from the market for having no efficacy.

Duvelisib is a dual inhibitor of PI3Kδ, γ developed for treatment of hematologic malignancies. In 2018, the FDA approved durvelisib in relapsed/refractory (R/R) CLL/SLL after two or more lines of treatment based on the DUO trial (11). The DUO trial is a global phase 3 trial comparing the efficacy of monotherapy for relapsed/refractory CLL/SLL with duvelisib (25 mg twice a day orally) or ofatuzumab. Patients were randomized in a 1:1 ratio into either treatment. The primary end point, PFS, was 13.3 months for duvelisib versus 9.9 months for ofatumumab (HR: 0.52; 95% CI: 0.39–0.69; P<0.0001). The most common adverse events were diarrhea, neutropenia, pyrexia, anemia and cough. The ORR was significantly higher in the duvelisib arm vs ofatumumab (74% vs 45%). The final analysis, however, showed no OS benefit (HR: 1.09; 95% CI: 0.79–1.51). The duvelisib approval in R/R FL in 2018 was an accelerated one based on the response rates at 6 and 12 months and the approval mandated additional post marketing trial which never occurred. Finally, the approval in FL was voluntarily withdrawn in 2022. Further analysis of the DUO trial’s OS showed that the treatment arm crossed over to show potentially more death after ~45 months compared with the ofatumumab (HR: 1.06; 95% CI: 0.71–1.58).

Umbralisib is a dual inhibitor of PI3Kδ/casein kinase-1ϵ, thought to exhibit improved selectivity for PI3Kδ compared with other PI3K inhibitors. In a phase IIb open labeled multi cohort trial enrolling 208 patients designed to evaluate the efficacy and safety of umbralisib in patients with R/R indolent non-Hodgkin’s lymphoma who are CD20+ and not responding to at least one anti-CD20 treatment. Patients were given umbralisib 800 mg orally once a day until disease progression, unacceptable toxicity or withdrawal. The ORR was 47.1% and tumor reduction occurred in 86.4% of patients. The median time to response was 2.7–4.6 months. The median duration of response was not reached for marginal zone lymphoma (MZL), 11.1 months for FL, and 18.3 months for SLL. Median PFS was not reached for MZL, 10.6 months for FL, and 20.9 months for SLL. At least one serious AE occurred in 53.4% of patients and included neutropenia (11.5%), diarrhea (10.1%) and increased LFTs (~7%) with 31 patients discontinuing study due to SAEs (12). The FDA gave accelerated approval to umbralisib for patients with R/E (extranodal) FL who received three or more lines of treatment, or R/R MZL who had received one or more lines of anti-CD20 treatment. In a phase 3 trial comparing umbralisib plus ublituximab versus obinutuzumab and chlorambucil in untreated R/R CLL, the PFS was improved with HR 0.55 (13). However, interim analysis showed that OS was worse with HR 1.23 (14). The manufacturer eventually withdrew all applications from CLL, SLL, FL and MZL due to lack of OS benefit.

PI3Ks phosphorylate the inositol ring of phosphatidylinositol (PtdIns) lipids. Three classes are documented, and Class I is noted for its role in regulating cell signaling at and downstream of the plasma membrane. Class I PI3Ks consist of a regulatory subunit (p85) and a catalytic subunit (p110), which exists in four isoforms: α, β, γ, and δ. Whereas p110α and p110β are ubiquitously expressed in different tissues, p110γ and p110δ are mainly concentrated in leukocytes. Widespread high throughput sequencing discovered that p110α mutations were present in 14% of all solid tumors, strengthening the belief that targeting PI3Kα in solid tumors is sound (15). Initial successes in the clinics targeting PI3K in B cell malignancies using delta selective idelalisib ushered in an era of subtype targeting. The impact of different inhibitors on PI3Kδ and PI3Kα function is studied in BaF3 cells (16). It is noteworthy that Treg downregulation by PI3Kδ inhibitor emerged as an approach of TIME manipulation ( Figure 1 ).

From clinical trial experiences, a common theme has emerged. PI3Kδ inhibitors induce relatively quick response initially, resulting in improvement in ORR or PFS evaluated as an early clinical trial endpoint. However, in the long term, these responses have not translated into an OS advantage, raising the question of whether the predicative values of ORR/PFS are appropriate clinical endpoints for this class of drug (14). Efficacy in B cell lymphoma has to do with PI3Kδ’s ability to drive transformed B cells out of their protective lymph node or spleen niche which has the protective stroma to provide adhesion, survival or homing signal to CLL cells as evidenced by peripheral blood lymphocytosis initially after the inhibitor use (38, 39) although lymphocytosis is not always evident when combined with rituximab (40).

Evaluating the side effect profile in patients treated with PI3Kδ inhibitors revealed many serious adverse events (AEs), including infections such as pneumonia, and autoimmune AEs such as diarrhea/colitis, hepatitis and pneumonitis which accounted for the fatalities seen in the trial. Indeed, the FDA mandated a box warning in the approval of idelalisib and later for duvelisib, regarding the risk of death due to drug-induced toxicities. Fatal or serious infection happened in 21–48% of patients who received idelalisib and 9–11% for duvelisib per drug insert. The side effects are even more prominent when these drugs were used in untreated populations (40, 41). AEs associated with the use of PI3Kδ inhibitors are related to interruption of signaling mediated by the PI3Kδ. Given the role of PI3Kδ in B and T cell antigen receptor signaling and impaired innate natural killer cells, macrophage function and immune response anticipated in PI3Kδ use, infections such as pneumonia or sepsis are anticipated. This could be attributable to impaired CD4/CD8+ effector cell differentiation, which may lead to infections including bacterial, viral such as cytomegalovirus, or opportunistic like pneumocystis pneumonia. Diarrhea/colitis, pneumonitis and increased liver function tests are autoimmune in nature and have been linked to dysfunction of T-regulatory cells when PI3Kδ is suppressed (17, 42). It is noteworthy that copanlisib, which is considered a pan PI3K inhibitor, characteristically induced hyperglycemia and hypertension due to its effects on the alpha subunit of PI3K. Fortunately, these are transient (43). Analysis of 5 published trials including linperlisib failed to link other specific side effect(s) clearly to the subunit they are preferentially targeting possibly due to the overlapping nature of side effects these other subunits target simultaneously. Serious AEs led to a high rate of dose adjustment (58%) and discontinuation (39%) in clinical trials (44). Infection rates almost doubled from 9.8% per 100 person-years to 18.4% in the treatment group leading to shorter days on drug 173 vs 473 and a numerically higher mortality rate in Medicare beneficiaries receiving idelalisib (45). Obviously, if an infection is not taken care of properly, increased mortality becomes a genuine threat to improving survival. Indeed, patients on trial did well initially but the advantage was erased as the trial continued, with more cumulative incidence of serious AEs. It should be noted that the COVID pandemic may have contributed to less frequent follow up and infection control (6).

As of January 1, 2025, all FDA approved PI3Kδ inhibitor indications for lymphoma have been withdrawn in the US and ten trials for PI3Kδ are recruiting: Roginolisib in melanoma (NCT07203391, NCT06717126), myelofibrosis (NCT06887803), CLL(NCT06644183), R/R peripheral T cell lymphoma (NCT07018752) and non-small cell lung cancer (NCT06879717). HMPL-689 for relapsed or refractory (R/R) lymphoma (NCT05713110), and linperlisib in R/R peripheral T cell lymphoma (NCT06083701; NCT05949944) and R/R cutaneous T cell lymphoma (NCT06037239). Of note, linperlisib received approval in China in 2022 based on a phase 2, single arm study with ORR as the primary end point (46), but it is only available in China. Study of this drug in the US will need solid efficacy and safety data. Further hope for the development of PI3Kδ inhibitors in hematological malignancies and solid tumors has suffered due to FDA withdrawal of approval. The authors suggest that the potential of PI3Kδ inhibitors for overcoming immune tolerance is under explored. It is not clear whether the effects of PI3Kδ inhibitor are mediated solely by enforcing the antitumor immunity in heme malignancy (47). The important events from discovery of PI3K pathway to approval of its many inhibitors is depicted in Figure 2 . PI3K inhibitors tested in a variety of solid tumors are summarized in Table 1 .

The development of PI3K inhibitors is marked by the initial excitement of high objective response rate and progression-free survival (PFS) targeting diseases that are of indolent nature. However, longer observation failed to establish OS benefit from the treatment intervention and showed potentially harmful effects when these drugs were given on a continuous daily dosing schedule (9, 12, 48). This demonstrates a large gap between early efficacy (tumor-based) measures (ORR, PFS) and long-term survival which has been the gold standard of oncology trials (14). It is in sharp contrast with trials involving ICI in which OS benefits are seen without the ORR and PFS benefit perhaps characteristic of immunotherapy (49, 50). The treatment groups receiving PI3Kδ invariably had higher incidence of SAEs leading to the FDA recognition of this class-specific fatal AE with high rate of toxicity. With only studies of continuous dosing of these agents, with very high cumulative SAE and AEs, the FDA was left with no choice but to withdraw approval.

Activation-induced cytidine deaminase (ACD) is a B cell specific enzyme that targets immunoglobulin genes to initiate class switch recombination and somatic hypermutation. Its activity is tightly regulated by PI3K pathway normally. Chronic PI3Kδ inhibitor use is shown to increase somatic hypermutation and chromosomal translocation. This safety concern is especially relevant if patients stay on PI3K inhibitors long term (51, 52). From a small subcohort study of APDS patients who have constitutively active PI3Kδ pathway (n=9), continuous leniolisib long term (~4.5 years) conferred clinical benefit and seemed to be safe when carefully selected and monitored (53). Several questions remain: 1) Do we need continuous dosing for the agent to be effective? 2) Have we established the minimal effective dose instead of the conventional maximum tolerated dose? 3) Can we mitigate the potentially lethal side effects by simply dosing less frequently? 4) What are the most appropriate areas to further explore PI3K inhibitors in cancer therapy?

Nearly 30 years after the discovery of the PI3Kδ gene, targeting PI3Kδ signaling proved to be a safe and feasible for B cell malignancies. In addition, it has drawn attention to translational researchers as a promising entry point for modulation of the TIME to favor cytotoxic tumor killing by T lymphocytes. However, treating this kinase as a simple on and off switch has been met with excessive and lethal AEs including death. Promising drugs were left with unproven efficacy due to toxicity and withdrawal from the market, leaving many outstanding concepts yet to be tested in solid tumors. Although clinic setbacks have led to the withdrawal of PI3K inhibitors, current evidence suggests that these are temporary challenges rather than a definitive endpoint for this therapeutic approach. Significant changes in the way we drug patients may be necessary, however, and further exploration of intermittent dosing or metronomic dosing should be considered by the scientific community.