NSAIDs for PI3K-altered head and neck cancer: a biomarker-driven therapeutic opportunity

Epidemiology of non-steroidal anti-inflammatory drugs (NSAIDs) and head and neck cancers (HNCs)

NSAIDs are a heterogeneous class of agents characterized by inhibition of cyclooxygenase (COX) enzymes and subsequent suppression of prostaglandin synthesis. NSAIDs are frequently categorized as non-selective COX inhibitors (e.g., aspirin, ibuprofen, and naproxen), which inhibit both COX1 and COX2, or selective COX2 inhibitors (e.g., celecoxib)¹. Beyond their analgesic and anti-inflammatory effects, NSAIDs have been extensively investigated for their cancer chemopreventive and therapeutic properties, including suppression of prostaglandin E2 (PGE2), a mediator implicated in tumor proliferation, angiogenesis, immune evasion, and metastasis².

The earliest, most compelling epidemiologic evidence supporting NSAIDs as chemopreventive agents emerged in colorectal cancer (CRC). Observational studies in the 1980s and 1990s noted diminished incidence of colorectal adenomas and carcinomas among chronic aspirin users^3,4, including a landmark American Cancer Society cohort analysis showing diminished risk of fatal colon cancer among regular aspirin users⁵. These findings were subsequently supported by large cohort studies and randomized trials demonstrating diminished CRC incidence and mortality with long-term aspirin use^6,7.

These findings prompted the investigation of NSAIDs across multiple malignancies, including esophageal, gastric, breast, lung, prostate, and HNCs^8–10. However, outside CRC, the epidemiologic results have been inconsistent. In head and neck squamous cell carcinoma (HNSCC) specifically, a systematic review of observational studies published through 2011 has concluded that the overall evidence is conflicting, and only one study¹¹ showed a statistically significant inverse association between aspirin use and HNSCC risk [adjusted odds ratio (OR) = 0.75, 95% confidence interval (CI) 0.58–0.96]¹².

Since 2011, the results have remained mixed. Selected larger (N > 2,000 cases) post-2011 studies in Table 1 illustrate this heterogeneity. A Scotland-based general-practice case-control study¹³ has reported no significant association between aspirin prescription (at least 1 prescription) and upper aerodigestive tract cancers (including head and neck sites), although non-aspirin NSAID use did show an inverse trend with cancer risk (OR = 0.8). A Danish nationwide nested case-control study¹⁵ including 12,389 HNC cases has reported no association between low-dose aspirin use (≥2 prescriptions) and overall HNSCC risk (OR = 1.03, 95% CI 0.97–1.10). In contrast, Becker et al.¹⁴ have reported an association between ≥6 ibuprofen prescriptions (but not lower exposure) and diminished HNSCC risk (adjusted OR = 0.54, 95% CI 0.37–0.94). A more recent case-control analysis¹⁶ has reported an inverse association between aspirin intake and HNSCC (all subsites, OR = 0.48, 95% CI 0.26–0.91, all sub-sites), and similar estimates for laryngeal cancer (OR = 0.58). Finally, in a very large study including >5 million cases and controls each¹⁷, acetylsalicylic acid (ASA) exposure was associated with diminished HNSCC incidence (OR = 0.88, 95% CI 0.86–0.9).

For HNSCC disease outcomes (post-diagnosis NSAID use), the literature is more limited and also conflicting (Table 2). Aspirin use is associated with increased overall survival (OS), and hazard ratios (HRs) of 0.42¹⁹ to 0.59¹⁸ have been reported by 2 studies with moderate sample sizes (1,525 and 460 patients, respectively). In a large study¹⁷ including 51,640 matched cases and controls, ASA use has been associated with a modest increase in 5-year survival among patients with HNSCC (67.93% in ASA users vs. 65.54% in non-users) and a statistically significant but modestly diminished rate of death (22.7% vs. 23.6%, P = 0.0006). In contrast, another large cohort study (N = 10,770)²⁰ has reported an association between post-diagnosis use of non-aspirin NSAIDs and elevated HNSCC-specific mortality within the first year after diagnosis (HR = 1.68). Collectively, these findings illustrate the inconsistent evidence regarding the effects of NSAIDs on both HNSCC risk and survival outcomes in unselected HNC patient populations.

In 2012, Liao et al.²¹ suggested that tumor genomics might modify aspirin benefit. They conducted a retrospective cohort analysis in 964 health professionals diagnosed with CRC. Aspirin exposure was defined as regular use after diagnosis (≥2 times per week), with doses ranging from 81 mg to 325 mg; patients were categorized as regular vs. non-regular users. Tumor PIK3CA mutation status was determined with PCR followed by pyrosequencing of exons 9 and 20. The primary outcomes were CRC-specific survival and OS. Regular aspirin use after diagnosis was associated with significantly increased survival among patients with PIK3CA-mutant tumors (CRC-specific survival HR = 0.18; OS HR = 0.54). No survival benefit was observed among patients with PIK3CA-wild-type (WT) tumors, and a statistically significant interaction was found between aspirin use and PIK3CA status (P = 0.009).

In HNSCC, the best-developed molecularly stratified evidence to date has come from Hedberg et al.²², who have evaluated NSAID use in a PIK3CA-characterized cohort and reported marked improvements in disease-specific survival benefit (HR = 0.23) and OS (HR = 0.31) among regular aspirin users whose tumors bore PIK3CA alterations, but no clear benefit in patients with PIK3CA-WT tumors. These studies suggest that non-stratified population analyses might possibly obscure benefit in molecularly defined subsets and support PIK3CA as a plausible effect modifier of the effects of NSAIDs in cancer.

Collectively, these findings suggest that the relationship between NSAID exposure and HNSCC risk or outcomes cannot be fully explained through population-level analyses alone. The heterogeneous epidemiologic results might have been due to variability in exposure definitions; dosing; duration of use; and confounding factors, such as recall bias and unmeasured over-the-counter use. More importantly, emerging evidence in CRC and HNSCC suggests that tumor molecular context, including alterations in PIK3CA, might determine whether NSAID-mediated effects translate to meaningful antitumor activity. These observations support a growing view that the potential benefits of NSAIDs in cancer prevention or therapy might be confined to biologically defined tumor subsets, thereby highlighting the importance of integrating tumor biology into future epidemiologic and clinical investigations of NSAID use in HNSCC.

PI3K signaling and alterations in HNSCC

The PIK3CA gene encodes the p110α catalytic subunit of PI3Kα²³. A regulatory subunit, p85α, heterodimerizes with p110α and subsequently maintains PI3Kα in an inactive state under quiescent conditions. The PI3Kα enzyme is a member of a larger family of PI3K proteins consisting of 4 distinct subclasses²⁴. Class 1 PI3Ks consist of p110α, p110β, p110δ, and p110γ, which can form heterodimers with the cognate regulatory proteins p85α, p85β, p55γ, and p101, respectively. Class 1 PI3Ks phosphorylate phosphatidylinositol (4,5)-bisphosphate (PIP₂), thus generating the second messenger phosphatidylinositol (3,4,5)-trisphosphate (PIP₃). These enzymes play key roles in promoting cellular proliferation, survival, and migration²⁵. Class II PI3Ks (PI3KC2α, PI3KC2β, and PI3KC2γ) generate phosphatidylinositol (3)-phosphate and phosphatidylinositol (3,4)-bisphosphate, and have roles in endocytosis and metabolism. No regulatory subunits are known for class II PI3Ks. The lone member of the class III PI3K subfamily, the Vps34 protein, heterodimerizes with the negative regulatory subunit Vps15. Vps34 phosphorylates phosphatidylinositol, thus generating phosphatidylinositol (3)-phosphate, and is involved in autophagy and endocytosis. Unlike class I, II, and III PI3Ks, which are lipid kinases, the structurally related class IV PI3K proteins are serine/threonine kinases and are often referred to as PI3K-related kinases. Examples of class IV proteins include ATM, ATR, DNA-PK, and mTOR. Class IV proteins have diverse cellular roles, including DNA repair, and cellular proliferation and metabolism.

Among the genes encoding the catalytic and regulatory subunits of class I–IV PI3Ks, genetic alterations in the PIK3CA gene occur most frequently in human cancers²⁶. Analysis of a cohort of 504 patients with HNSCC (The Firehose Legacy) in The Cancer Genome Atlas (TCGA^27,28) has identified PIK3CA mutations in 13.7% of patients and PIK3CA amplification in 16.5% of patients; moreover, 34.5% of patients with HNSCC in this cohort exhibited PIK3CA gene alterations (Table 3). PIK3CB, encoding p110β, was altered in 10.5% of patients, whereas genes encoding the other catalytic and regulatory subunits were altered considerably less frequently (Table 3). Similarly, alterations in genes encoding class II and class III catalytic and regulatory subunits generally occurred infrequently (Tables 4 and 5).

Notably, PIK3CA alterations are also observed in several other human cancers. Table 6 depicts cancers with PIK3CA alteration frequencies exceeding 10% in large patient cohorts. Interestingly, whereas PIK3CA mutation and amplification occur at approximately similar frequencies in HNSCC; in most cancers, mutation is the more common alteration. Exceptions include lung squamous cell carcinoma and colorectal cancer, in which amplification is the predominant PIK3CA alteration.

In resting cells, p110α is normally maintained in an inactive state through heterodimerization with p85α in the PI3Kα complex (hereafter referred to as PI3K). The binding of growth factors to their cognate cell surface receptors leads to physical disruption of p110α/p85α³⁵ and results in PI3K activation in cells (Figure 1). Mutation of p110α or amplification/overexpression also leads to the formation of free and active cellular p110α. The activation of PI3K in cells initiates a cascade of signaling that ultimately stimulates cellular proliferation and survival. The PI3K/AKT/mTOR signaling pathway is a major defined biochemical pathway leading from PI3K to cellular proliferation^25,26. In brief (Figure 1), active p110α phosphorylates PIP₂, thus generating PIP₃, which in turn provides a docking site for PDK1 and AKT (which has 3 isoforms, AKT1–3) on the inner leaflet of the plasma membrane. The proximity of these 2 proteins facilitates the phosphorylation and activation of AKT by PDK1. AKT is a serine/threonine kinase that phosphorylates and inactivates the TSC1/TSC2 protein complex. Consequently, relief of the inhibitory effect of the TSC1/TSC2 complex on RHEB, a GTPase protein important for activation of the mTORC1 complex, results in mTORC1 activation. The mTORC1 complex contains the catalytic subunit mTOR, a class IV PI3K serine/threonine kinase. Other components of the mTORC1 complex include the scaffolding proteins RAPTOR and mLST8, and the negative regulatory protein DEPTOR. mTORC1 activity is also negatively regulated by the PRAS40 protein. Activation of mTORC1 via the PI3K signaling pathway results in direct phosphorylation of S6K1 and 4E-BP1, proteins that regulate protein translation. In addition, mTORC1 promotes anabolic metabolism by enhancing the expression and activity of metabolic regulators. Ultimately, changes in cellular protein translation and metabolism facilitate the key processes of cellular proliferation and suppression of cell death.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

The PI3K/AKT/mTOR signaling pathway. Growth factor binding to receptor tyrosine kinases activates PI3K, a heterodimer composed of the catalytic subunit p110α and the regulatory subunit p85α. Activated PI3K catalyzes the phosphorylation of PIP2, thus generating PIP3 at the plasma membrane. PIP3 recruits and activates downstream signaling components, including PDK1 and AKT. Activated AKT phosphorylates multiple substrates that promote cell survival, proliferation, and metabolic reprogramming. One major downstream branch involves inhibition of the TSC1/TSC2 complex, thus leading to activation of the small GTPase RHEB and subsequent stimulation of mTORC1. mTORC1, composed of mTOR, RAPTOR, MLST8, and DEPTOR, promotes protein synthesis and cell growth through the phosphorylation of downstream effectors. PRAS40 acts as an inhibitor of mTORC1. The lipid phosphatase PTEN acts as a negative regulator of the pathway by converting PIP3 back to PIP2, thereby limiting PI3K signaling. Green arrows show net activation of the mTORC1 pathway. Red arrows show inhibition of the mTORC1 pathway. AKT, protein kinase B; DEPTOR, DEP domain containing mTOR-interacting protein; MLST8, mechanistic target of rapamycin-associated protein LST8; mTOR, mechanistic target of rapamycin; mTORC1, mechanistic target of rapamycin complex 1; PDK1, phosphoinositide-dependent kinase-1; PI3K, phosphoinositide-3-kinase; PIP2, phosphatidylinositol-4,5-bisphosphate; PIP3, phosphatidylinositol-3,4,5-trisphosphate; PRAS40, proline-rich AKT substrate of 40 kDa; PTEN, phosphatase and tensin homolog; RAPTOR, regulatory-associated protein of mTOR; RHEB, Ras homolog enriched in brain; TSC1, tuberous sclerosis complex 1; TSC2, tuberous sclerosis complex 2. Created with BioRender.com.

A major negative regulator of the PI3K signaling pathway is the phosphatase and tensin homolog deleted on chromosome ten (PTEN) protein (Figure 1). PTEN dephosphorylates PIP₃ to PIP₂, thereby halting PDK1/AKT interaction and deactivating subsequent signaling in the pathway³⁶.

Given the high frequency of PIK3CA alterations in many human cancers, consideration of whether other components of the PI3K signaling pathway might also be frequently altered is warranted. Table 7 shows the frequency of genetic alterations in multiple components of the PI3K signaling pathway described above in patients with HNSCC. Notably, none of these genes are altered in more than 10% of HNSCC tumors, a frequency far less than that of PIK3CA alteration. However, a major limitation of this analysis is that the proteins’ expression levels in tumor cells were not evaluated. Up- or down-regulation of the proteins through epigenetic or posttranslational mechanisms might have major effects on signaling via the PI3K pathway. For example, whereas the PTEN gene was found to be altered in only 6.2% of HNSCC in the TCGA, Squarize et al.³⁷, through immunohistochemistry, observed a loss of PTEN protein expression in approximately one-third of analyzed HNSCCs.

Although PI3K mutation typically results in constitutive activation of the protein, amplification of PIK3CA leads to an abundance of free p110α, both of which result in chronic activation of the PI3K signaling pathway in cells. Collectively, PIK3CA alterations have oncogenic effects that drive proliferation, cell survival, and migration, and PIK3CA has been formally classified as an oncogene^38,39. In HNSCC, PIK3CA is the most frequently altered oncogene. In contrast, PTEN is a tumor suppressor gene, and mutation or loss of PTEN expression (via genetic or other mechanisms) also promotes oncogenesis.

Although cancer-associated mutations in PIK3CA are spread throughout the coding region, a preponderance of mutations occur at 3 hotspot locations: E542 and E545 in the helical domain and H1047 in the kinase domain⁴⁰. Mutations at these sites are termed canonical mutations and give rise to constitutively active p110α proteins with oncogenic activities³⁸. In HNSCC, 63% of mutations occurring in PIK3CA are canonical mutations, whereas 37% are noncanonical mutations⁴¹. Functional analyses of noncanonical mutations have revealed that most of these mutant proteins also exhibit hyperactive kinase activity and promote oncogenic phenotypes⁴¹.

Given the high frequency of PIK3CA alterations and the oncogenic effects of aberrant PI3K/AKT/mTOR signaling in cancer, considerable efforts have focused on targeting PI3K activity as a therapeutic approach. Many PI3K inhibitors have been developed and tested²⁶. However, results from clinical studies have been largely disappointing. Adverse toxicity associated with the administration of PI3K inhibitors has been a common limitation and is often ascribed to non-selective inhibition of multiple class I PI3K isoforms. Consequently, research has been undertaken to develop isoform-specific inhibitors, particularly p110α-selective inhibitors. Alpelisib is a p110α-selective inhibitor that has been evaluated in patients with breast cancer (BC) bearing hotspot PIK3CA mutations. In a phase III trial, the regimen of alpelisib plus fulvestrant (an estrogen receptor degrader) improved disease-free survival (DFS) over that with fulvestrant alone; subsequently, this combination received FDA approval for BC with hotspot PIK3CA mutations^42,43. In HNSCC, alpelisib has shown activity in a patient bearing a noncanonical PIK3CA mutation⁴¹. Further refinement of p110α-selective inhibitors and development of mutation-selective inhibitors offer considerable promise for the treatment of PIK3CA-mutant cancers, although issues of potency, specificity, and toxicity remain major obstacles.

Preclinical studies highlighting the mechanistic basis for NSAID inhibition of PIK3CA-altered HNSCC

The ability of NSAIDs to inhibit the growth of HNSCC cells and tumors has been directly demonstrated in preclinical models. Treatment with the NSAIDs sulindac or celecoxib has been observed to elicit greater growth inhibition in HNSCC cell lines with mutant PIK3CA rather than WT PIK3CA²². Similarly, sulindac, celecoxib, or aspirin treatment of mice with human HNSCC patient-derived xenograft tumors has been found to result in tumor growth inhibition in models with mutant PIK3CA but not WT PIK3CA²². In addition, aspirin treatment has been shown to inhibit the growth of HNSCC tumors engineered for loss of PTEN in immunocompetent mice⁴⁴.

Despite epidemiologic evidence in humans and preclinical findings in mouse models, the mechanistic basis for NSAID inhibition of PIK3CA-altered (or PTEN-deficient) cancers cannot be deduced by considering only the components of the PI3K/AKT/mTOR signaling pathway. Importantly, activated PI3K also induces expression of COX2 (Figure 2)⁴⁵. COX2 catalyzes a rate-limiting step in the production of PGE2 and is a direct NSAID target (Figure 2)⁴⁶. Notably, NSAIDs also inhibit COX1, a constitutively expressed isoform not regulated by PI3K signaling; therefore, the selective sensitivity of PIK3CA-altered tumors is probably attributable to inhibition of inducible COX2 rather than COX1. HNSCC cell lines and patient-derived xenograft tumors with mutant PIK3CA or PTEN deficiency generate higher levels of secreted PGE2 than cells and tumors with WT PIK3CA and PTEN^22,44. Moreover, NSAID treatment results in a greater decrease in PGE2 levels in HNSCC cells and tumors with mutant rather than WT PIK3CA²².

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

PI3K signaling promotes COX2-dependent PGE2 production. Activation of PI3K signaling leads to the accumulation of PIP3 and the recruitment of PDK1 and AKT to the plasma membrane. Activated AKT stimulates downstream transcriptional and metabolic programs that increase expression of COX2. COX2 catalyzes the conversion of arachidonic acid to prostaglandin intermediates, thus ultimately leading to production of PGE2, which in turn promotes tumor growth. PTEN negatively regulates this pathway by dephosphorylating PIP3 to PIP2 and restraining PI3K pathway activation. NSAIDs inhibit COX2 activity. AKT, protein kinase B; COX2, cyclooxygenase 2; NSAID, nonsteroidal anti-inflammatory drug; PDK1, phosphoinositide-dependent kinase-1; PGE2, prostaglandin E2; PI3K, phosphoinositide-3-kinase; PIP2, phosphatidylinositol-4,5-bisphosphate; PIP3, phosphatidylinositol-3,4,5-trisphosphate; PTEN, phosphatase and tensin homolog. Created with BioRender.com.

PGE2 contributes to cancer development/progression via both cancer cell intrinsic and extrinsic mechanisms. In the cancer cell intrinsic mechanism, PGE2 produced and secreted by cancer cells subsequently binds to neighboring cancer cells in an autocrine or paracrine manner (Figure 3A). The cell surface receptors for PGE2, EP1–4, are G-protein coupled receptors. PGE2 binding to the EP2 and EP4 receptors results in PI3K activation and oncogenic signaling via the PI3K/AKT/mTOR signaling pathway^47–49 (Figure 3A). Therefore, NSAIDs, by preventing cancer cell production of PGE2, attenuate PGE2-mediated autocrine or paracrine activation of the PI3K/AKT/mTOR pathway. The extrinsic cancer-promoting mechanism of PGE2 activity involves inhibition of anti-tumor immunity (Figure 3B). Specifically, PGE2 inhibits the development of cytotoxic T-cells, and promotes the development/activity of Tregs, neutrophils, and M2 macrophages^50,51. NSAID inhibition of PGE2 production by cancer cells blocks this important immunosuppressive mechanism. Therefore, NSAID disruption of PI3K/COX2-mediated PGE2 production by cancer cells with PIK3CA alterations or PTEN deficiency disarms 2 distinct mechanisms for promoting cancer growth, thus highlighting the potential utility of NSAID treatment for therapeutic benefit.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

NSAIDs suppress COX-dependent PGE2 signaling and have tumor cell-intrinsic and tumor microenvironment effects. (A) Tumor cell-intrinsic effects. NSAIDs, including aspirin, inhibit COX1/2 activity and thereby suppress the synthesis of PGE2. COX1 contributes to constitutive PGE2 production under basal conditions, whereas COX2 is inducible and is frequently upregulated in cancer. PGE2 signals through EP receptors on the cell surface and activates downstream signaling pathways including PI3K. In tumor cells, activating PIK3CA mutations or amplification drive hyperactivation of PI3K signaling and lead to activation of AKT and increased COX2 expression, thus further increasing PGE2 production and establishing a feed-forward signaling loop. NSAIDs inhibit COX enzymatic activity, by blocking both COX1 and COX1-dependent PGE2 synthesis. NSAIDs preferentially counteract this tumor cell-intrinsic prostaglandin program in PIK3CA-altered cancers. (B) Tumor microenvironment effects. Beyond tumor cells, PGE2 signaling shapes the immune landscape of the tumor microenvironment by promoting immunosuppressive states. Elevated PGE2 promotes the recruitment and activation of immunosuppressive cell populations, including neutrophils, M2-like macrophages, and Tregs, and simultaneously inhibits the activation, infiltration, and cytotoxic function of CD8⁺ T cells. Green arrows show activation. Red arrows show inhibition. AKT, protein kinase B; CD8, cluster of differentiation 8; COX1, cyclooxygenase 1; COX2, cyclooxygenase 2; EP receptors, prostaglandin E receptors; NSAIDs, nonsteroidal anti-inflammatory drugs; PGE2, prostaglandin E2; PI3K, phosphoinositide-3-kinase; PIK3CA, phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha; Tregs, regulatory T cells. Created with BioRender.com.

Interventional clinical trials of NSAIDs in HNC

The notion that NSAIDs have anti-tumor activity is challenging, given the traditional considerations for developing drugs against cancer. Most systemic agents used in the treatment of most cancers, including HNSCC, target proteins that are critical for dividing cells. The toxicity of these drugs (including platinum chemotherapy in the cases of HNSCC), despite their activity against tumor cells, remains a consideration. Here, we summarize clinical trials evaluating NSAIDs as anti-cancer agents (Table 8).

The COX2-selective agent celecoxib has been the most widely tested NSAID in HNSCC. The first related clinical trial, reported in 2005, combined celecoxib with the EGFR tyrosine kinase inhibitor gefitinib in patients with recurrent or metastatic (R/M) HNSCC. Wirth et al.⁵² found that this combination was well tolerated, and 4 of 18 patients (22%) experienced a partial response. In 2011, the combination of celecoxib and erlotinib (another EGFR tyrosine kinase inhibitor) was tested in combination with re-irradiation for R/M HNSCC⁵³. Although the combination was feasible, the emergence of dose-limiting toxicities is likely to preclude use of this regimen outside the setting of a poor prognosis.

Oral premalignant lesions (OPLs) can arise in the oral mucosa before the development of HNSCC. Therefore, the chemopreventive activity of many agents has been tested in patients with OPL. Two studies have reported the feasibility of NSAIDs in this setting. In 2008, Wirth et al.⁵⁴ conducted a biomarker-driven pilot study of celecoxib in patients with OPL. They reported the feasibility of measuring activity with surrogate tissue biomarkers (such as Ki-67 and COX2 expression), but no placebo group was included to assess drug-specific activity. In the same year, Papadimitrakopoulou et al.⁵⁵ tested 2 doses of celecoxib (100 mg or 200 mg twice daily) in a phase 2 randomized pilot study, and found no activity of either concentration of celecoxib with respect to placebo in OPLs.

In other solid tumors, Guo et al.⁵⁶ compared chemotherapy alone vs. chemotherapy plus celecoxib in patients with R/M gastric cancer. Although COX2 expression in tumors showed some indication of an association with a response to celecoxib-containing therapy, no significant differences in progression-free survival or OS were detected. The addition of celecoxib to chemoradiation for non-small cell lung cancer (NSCLC) in a phase 2 randomized trial reported no survival benefit with the addition of this NSAIDs⁵⁷.

The effects of celecoxib on disease recurrence have been studied in breast cancer, most notably in the Randomized European Celecoxib or REACT Trial⁵⁸. This phase 3, randomized, double-blind, placebo-controlled study enrolled 2,639 patients who were followed for 10 years. Patients received 400 mg daily of celecoxib or placebo without evidence of an effect on DFS. More recently, a phase 3 randomized controlled trial compared aspirin (300 mg daily) vs. placebo in 3,020 patients with high-risk non-metastatic breast cancer⁵⁹. At an early follow-up time point (median follow-up 33.8 months), no benefit was detected in the aspirin treated group, and the trial was stopped early for futility.

Celecoxib (or placebo) was added to adjuvant (postoperative) standard of care in a phase 3 trial of patients with stage 3 CRC⁶⁰. The addition of celecoxib had no effects identified in this cohort.

Approximately 20% of patients with CRC characterized by DNA mismatch repair deficiency do not respond to immune checkpoint inhibitors. Given the absence of predictive biomarkers, a phase 2 trial has tested the combination of COX inhibitors (agents unspecified) with anti-PD-1 therapy in patients with metastatic disease⁶¹. Although the combination was well tolerated, the median progression-free survival and OS were not reached.

Polyps, particularly in the hereditary CRC condition Lynch syndrome, make this setting highly relevant for studying the activity of drugs that might prevent CRC. NSAIDs, primarily aspirin, have been the most widely studied drugs in this context. The double-blind randomized trial on cancer prevention with aspirin in hereditary colorectal cancer (CAPP2) has tested the effects of aspirin (600 mg daily) and observed significantly diminished CRC incidence in the aspirin-treated group at the 10-year follow-up timepoint⁶². In Japan, a double-blind, placebo-controlled randomized trial (the J-FAPP Study IV) has compared aspirin (100 mg per day) plus placebo; mesalazine (a topical anti-inflammatory agent) plus placebo; aspirin plus placebo; and placebo⁶³. Low dose aspirin suppressed polyps, and mesalazine had no added benefit. Aspirin concentration has been widely studied. A meta-analysis published in 2024 comparing low dose (<300 mg a day) vs. high dose (≥300 mg a day) aspirin has found a consistent benefit of only low dose aspirin in preventing colon adenomas⁶⁴.

Aspirin has also been studied in the adjuvant setting after treatment with curative intent for CRC. The ASCOLT trial, an international phase 3, randomized, placebo-controlled trial, enrolled 1,587 patients with CRC⁶⁵. Although the tested dose of aspirin (200 mg daily) was well tolerated, it had no effects on DFS (after patients were treated for 3 years). Another study has evaluated the effects of intermittent aspirin dosing (daily vs. 3 weeks on followed by 3 weeks off) on biomarkers in the rectal mucosa⁶⁶. In general, continuous aspirin administration more effectively decreased Ki-67 labeling and COX2 expression in the rectal mucosa.

Notably, all the above trials were performed in unselected populations. Participants were enrolled according to cancer or cancer predisposition diagnoses. Elucidation of the somatic alterations in human cancer has revealed driving oncogenes and oncogenic pathways that might ideally be leveraged to deliver precision cancer therapy. Cumulative evidence from preclinical models (discussed earlier) suggests that genetic alterations that activate the PI3K signaling pathway might mediate an effective response to NSAID therapy. Martling et al.⁶⁷ have recently reported positive findings of NSAID therapy in the setting of PI3K-altered localized CRC. In this landmark trial, patients with somatic PIK3CA mutations or alterations in other pathway genes, including PIK3R1 or PTEN, were randomized to receive low dose aspirin (160 mg a day) or placebo for 3 years. Participants who received aspirin experienced significantly diminished disease recurrence, thus validating the use of a predictive molecular biomarker (PI3K pathway alteration). Another trial (the SAKK 41/13 study) has compared aspirin (100 mg daily) vs. placebo in patients with PIK3CA mutated CRC⁶⁸. Although the study was ended early because of financial constraints, emerging evidence supported the use of aspirin in this population.

To date, only data from a single-institution retrospective study support prolonged DFS and OS in patients with PIK3CA mutated or PIK3CA amplified HNSCC who were long-term NSAID users²². Although these results are compelling, the effects of NSAIDs must critically be tested in a prospective study. Future clinical trials should prioritize expanding enrollment of patients bearing the PI3K pathway alterations (PTEN loss or PIK3CA alterations) with the strongest biological rationale and preclinical evidence. In this context, aspirin is the most practical agent, given its established pharmacology and clinical availability. Importantly, NSAID-associated toxicities, particularly gastrointestinal and intracranial bleeding, must be carefully considered. Although no HNSCC-specific safety signals have been reported to date, large population-based studies (e.g., ASPREE⁶⁹) have demonstrated that in unselected populations, particularly older adults, the bleeding risk of aspirin might outweigh its benefit. This finding underscores the importance of biomarker-driven trial design in HNSCC, for which restriction of enrollment to patients with PI3K pathway alterations might enhance therapeutic benefit while improving the risk-benefit ratio. Future trials should incorporate prospective stratification by molecular alterations (PI3K pathway), careful patient selection (excluding high bleeding risk), and standardized monitoring for adverse events.

Author contributions

Conceived and designed the analysis: Jennifer R. Grandis, Daniel E. Johnson

Collected the data: Jovanka Gencel-Augusto, Jennifer R. Grandis, Daniel E. Johnson

Wrote the paper: Jovanka Gencel-Augusto, Jennifer R. Grandis, Daniel E. Johnson.