Adding a COX-2 Inhibitor to Neoadjuvant Immunotherapy in dMMR/MSI-H Colorectal Cancer: What PICC-2 Tells Us
Alessandro Mannucci, MD, CGA-IGC Communications Committee Member
Source: Hu H, Shen X, Li Y, et al. Neoadjuvant toripalimab plus celecoxib versus toripalimab monotherapy for mismatch repair-deficient or microsatellite instability-high, locally advanced colorectal cancer (PICC-2): an open-label, multicentre, randomised, phase 2 trial. Lancet Oncol 2026;27:1020–30.
Why this trial matters
Over the last five years, neoadjuvant PD-1 blockade has completely changed the management of patients with mismatch repair-deficient (dMMR) or microsatellite instability-high (MSI-H) locally advanced colorectal cancer. Trials like KEYNOTE-177 (the original Memorial Sloan Kettering rectal cancer series) and NICHE-2 established that these immunologically "hot" tumours can respond dramatically to checkpoint blockade. Such treatments lead to pathological complete response (pCR) rates that would have been unthinkable in microsatellite-stable disease.
But even in the best series, roughly 1 in 5 to 1 in 2 patients still fail to achieve a pCR after neoadjuvant immunotherapy. That residual-disease population is where PICC-2 picks up the story and it does so with a strategy that is refreshingly simple: adding a cheap, generic, well-known drug (celecoxib) to a PD-1 inhibitor!
The biological rationale
Why celecoxib? Residual tumours after neoadjuvant immunotherapy are often enriched for IL-1β-expressing monocytes tied to COX-2–prostaglandin E2 (PGE2) signalling. Preclinically, COX-2-derived PGE2:
Promotes tumour-associated inflammation
Impairs type I interferon signalling
Induces cancer-promoting cytokines
Suppresses effective anti-tumour immunity, including T-cell dysfunction
In short, COX-2/PGE2 signalling looks like a mechanism by which tumours actively resist checkpoint blockade. Inhibiting COX-2 should, in theory, re-sensitise the tumour microenvironment to PD-1 blockade — and this isn't a purely theoretical leap: the same group's earlier PICC exploratory study (celecoxib + toripalimab) reported an 88% pCR rate, and their PCOX trial showed encouraging activity for the PD-1/COX-2 combination in metastatic dMMR/MSI-H CRC.
PICC-2 was designed to test this combination properly — as a randomised comparison.
Trial design at a glance
Design | Investigator-initiated, open-label, multicentre (3 academic hospitals, China), randomised phase 2 |
Population | dMMR or MSI-H colorectal adenocarcinoma, clinical stage T3–T4 or any T with N+, ECOG 0–1 |
Randomization | 1:1, stratified by tumour location (colon vs rectum) and clinical T stage (T4b vs non-T4b) |
Arm A (intervention) | Toripalimab 3 mg/kg IV q14d × 12 cycles + Celecoxib 200 mg PO BID (days 1–14 of each cycle) |
Arm B (Comparator) | Toripalimab (same schedule) |
Surgery | Within 4 weeks of last neoadjuvant dose, regardless of response |
Primary endpoint | Pathological complete response (pCR) — no residual viable tumour in primary tumour and all sampled lymph nodes — by central blinded independent pathological review, ITT population |
A key design choice worth flagging: this is a genuinely long course of neoadjuvant immunotherapy (six months, 12 cycles is much longer than the usual short 6–8 week used in NICHE-2/3, for example). The authors' hypothesis was that a longer duration of PD-1 blockade itself increases the chance of a complete response, and the celecoxib addition was layered on top of that already-extended backbone.
Who was enrolled
Between May 2022 and January 2025, 110 patients were randomised (55 per arm). Median age was 50 years, 59% were male, 34% had Lynch syndrome, 80% colon primaries, 69% cT4 tumours, and 95% clinically node-positive. This population represents a genuinely high-risk, locally advanced population, not an easy "early-stage" cohort cherry-picked for a good response.
The headline result
Pathological complete response (ITT population, central review):
Toripalimab + celecoxib: 49/55 (89%; 95% CI 78–96)
Toripalimab alone: 38/55 (69%; 95% CI 55–81)
Between-group difference: 19 percentage points (95% CI 4–34); p = 0.014
The per-protocol population (excluding 6 patients who did not undergo the planned surgery) told the same story: 92% vs 75% (difference 17 points; 95% CI 3–31; p=0.017). A post-hoc sensitivity analysis assuming all non-surgical patients had achieved pCR still preserved a significant 16-point difference (p=0.021) — reassuring given that missing surgical specimens are always a soft spot in pCR-based endpoints.
Secondary pathological outcomes moved in the same direction, even if some didn't reach formal significance individually: major pathological response (93% vs 85%), TRG0 (91% vs 71%), and tumour regression grade distribution generally favoured the combination arm (Table 2 of the article gives the full breakdown).
Who benefited most: the subgroup story
Prespecified subgroup analyses are where this trial gets clinically interesting, because two subgroups stood out with notably wide effect sizes:
cT4b disease. These are the tumours where you'd most want an alternative to straight-to-surgery: pCR 92% (12/13) with the combination vs 50% (6/12) with monotherapy alone — a 42-point difference. This matters because NCCN guidelines already recommend neoadjuvant checkpoint blockade as a preferred option specifically for cT4b dMMR/MSI-H colon cancer, precisely because cT4 tumours have historically shown lower rates of pathological response and downstaging after immunotherapy. If celecoxib genuinely rescues that high-risk subgroup, it could reshape how this population is managed.
Sporadic (non-Lynch) tumours. pCR 92% vs 67% with the combination. This is notable because sporadic dMMR/MSI-H tumours have generally been considered less immunogenic and less responsive to checkpoint blockade than Lynch-associated tumours (NICHE-2 found pCR was less common in sporadic disease, 61% vs 79% for Lynch). If celecoxib specifically helps close that gap, it suggests the COX-2/PGE2 axis may be a particularly relevant resistance mechanism in the sporadic subset.
By contrast, benefit in Lynch syndrome-associated tumours was more modest (83% vs 74%, CI crossing zero), plausibly a ceiling effect, since Lynch tumours already respond so well to PD-1 blockade alone that there's less room for celecoxib to add much.
Safety: no real toxicity trade-off
This is arguably as important as the efficacy data, because a strategy this simple only earns broad practice change if it's genuinely low-cost in toxicity, too.
Grade 3 treatment-related AEs: 5% (combination) vs 7% (monotherapy) — no meaningful increase
No grade 4/5 treatment-related AEs in either arm
Discontinuations for toxicity: 7% vs 5%; in the combination arm, every discontinuation was attributed to toripalimab, not celecoxib
Serious treatment-related AEs: identical, 7% in each arm
One postoperative death from sepsis (combination arm), judged unrelated to either study drug
How does this compare to the rest of the field?
Dual checkpoint blockade (PD-1 + CTLA-4/LAG-3): NEOSHOT, NICHE-2, NICHE-3 achieved pCR of 68–78%, but only in short 6–8 week courses, presumably to limit immune-related toxicity.
PD-1 + anti-angiogenic (NEOCAP: camrelizumab + apatinib): 61% pCR, but with "substantial toxicities."
Short-course single-agent PD-1 blockade across multiple series: 44–58% pCR.
PICC-2 (toripalimab + celecoxib, 6-month course): 89% pCR — among the highest reported in this disease setting to date, achieved with a similar safety profile to monotherapy and using an inexpensive, generic, widely available drug.
That accessibility point is worth dwelling on for a moment: celecoxib is off-patent, cheap, and already familiar to virtually every clinician. Compared with dual checkpoint blockade or novel combination agents, this is about as low a barrier to implementation as oncology gets — assuming the phase 3 confirmation holds up.
The bigger picture: where is this field heading?
A few threads worth pulling about the broader trajectory of dMMR/MSI-H CRC management:
Duration matters, and this trial pushed it further. Both PICC-2 arms used a full 6-month, 12-cycle neoadjuvant course, longer than most prior trials. Even the monotherapy arm's 69% pCR is numerically higher than most short-course series. The ongoing OPEN study (6 vs 3 months of neoadjuvant sintilimab) should help clarify whether duration alone, independent of celecoxib, is doing meaningful work here.
Organ preservation comes next. With pCR rates like these, a nonoperative "watch and wait" strategy becomes increasingly plausible for dMMR/MSI-H colon cancer too, an area pioneered by Dr. Cercek and colleagues' dostarlimab nonoperative management study. PICC-2's authors are already running a follow-on trial, PICC-3, specifically testing whether toripalimab + celecoxib can support selective nonoperative management guided by ctDNA dynamics. The gap between clinical complete response (60% in this trial) and pathological complete response (89%) underscores why better real-time response assessment (ctDNA, refined imaging) is such an active area — radiological/endoscopic assessment alone still meaningfully underestimates true pathological response.
This remains phase 2, single-country, single-ethnicity evidence. All 110 patients were Chinese, enrolled at three affiliated academic centres, with a relatively young median age (50 years), not necessarily representative of the broader, often older, dMMR/MSI-H CRC population seen internationally. There is also no upfront-surgery control arm, so the trial cannot address whether neoadjuvant immunotherapy is even worth the treatment burden for patients (especially colon cancers without T4/bulky disease) for whom upfront surgery remains the default.
Bottom line
PICC-2 provides the first randomised evidence that adding a COX-2 inhibitor to neoadjuvant PD-1 blockade can meaningfully increase pathological complete response in dMMR/MSI-H locally advanced colorectal cancer without any discernible toxicity trade-off, using a drug that is cheap and already sitting in most formularies. Confirmatory phase 3 data, longer follow-up for survival endpoints, and broader population validation are the next steps, but this is a strategy worth watching closely. And one our patients may start asking about sooner rather than later.
Comments