👃 Nasal Polyp Recurrence Prediction Simulator

Post-FESS Recovery & Recurrence Risk Intake

Chronic rhinosinusitis with nasal polyps (CRSwNP) affects 2–4% of adults and is a chronic relapsing disease, not a curable one. Functional endoscopic sinus surgery (FESS) reliably clears obstructive polyp tissue and restores sinus drainage, but surgery alone does not treat the underlying mucosal inflammation. Within months of an anatomically successful operation, a substantial subset of patients begin regrowing polyps — the goal of recurrence prediction is to identify these patients before symptoms return.

• ~20%: 5-Year revision surgery rate (after primary FESS, all-comers)
• ~4×: AERD recurrence multiplier (vs non-AERD CRSwNP)
• 2–4%: Global CRSwNP prevalence (of adults)
• 85–90%: 12-month symptom improvement (before relapse risk accrues)

CRSwNP as a chronic relapsing disease

FESS mechanically removes polyps and widens sinus ostia to restore mucociliary drainage and allow topical therapy to reach the sinus mucosa. Surgical success rates for symptom control at 12 months are excellent — 85–90% of patients report meaningful improvement. But CRSwNP is fundamentally a mucosal inflammatory disease, most often driven by Type 2 (T2) immunity, and the mucosa that regrows after surgery is still exposed to the same inflammatory drivers that produced the original polyps.

Without risk-adapted maintenance therapy, published cohorts report endoscopic polyp recurrence in 40–60% of patients within 18–24 months, and roughly 1 in 5 patients undergo revision surgery within 5 years. Recurrence is not evenly distributed: a small subset of high-risk patients (eosinophilic, AERD, asthmatic) account for a disproportionate share of revisions, while low-risk patients may remain polyp-free for a decade on nothing more than saline irrigation.

Because recurrence risk is so unevenly distributed across the CRSwNP population, blanket post-operative protocols both under-treat high-risk patients and over-treat low-risk ones. Individualized risk prediction is what allows biologics and topical steroids — both expensive and non-trivial commitments — to be targeted to the patients who actually need them.

Clinical inputs captured at baseline

A recurrence-risk intake at the post-operative visit typically gathers:

• Tissue eosinophilia — the percentage of eosinophils among inflammatory cells in the resected polyp specimen, or absolute eosinophils per high-power field (HPF), from the surgical pathology report • AERD / asthma comorbidity — presence of aspirin-exacerbated respiratory disease (Samter's triad: asthma, nasal polyps, NSAID hypersensitivity) or asthma alone, both markers of systemic T2 airway disease • Prior revision surgery count — each previous sinus surgery is itself a marker of a more treatment-resistant phenotype • Lund-Mackay CT score — a 0–24 radiologic severity score summing opacification across sinus subsites, with higher preoperative scores correlating with more extensive disease burden

These four inputs are inexpensive, already collected in routine ENT practice, and together explain much of the variance in who relapses quickly versus who does not.

Why risk stratification changes management

Post-operative CRSwNP management sits on a spectrum from simple saline irrigation to systemic biologic therapy costing tens of thousands of dollars per year. Risk stratification is what makes this spectrum usable in practice:

• Low predicted risk → saline irrigation and routine endoscopic follow-up are usually sufficient • Moderate risk → high-volume topical corticosteroid irrigation becomes the backbone of therapy • High risk (eosinophilic + AERD, prior revisions, high Lund-Mackay) → early biologic therapy or a steroid-eluting stent may be considered even before symptoms recur, since endoscopic regrowth often precedes symptomatic relapse by months

The intake stage therefore is not just documentation — it is the raw material the downstream risk model consumes to route each patient toward the appropriate intensity of maintenance care.

Tissue Eosinophilia & the Type-2 Inflammatory Endotype

Not all nasal polyps are biologically identical. Histopathologic and molecular endotyping separates CRSwNP into eosinophilic (Type-2 driven) and non-eosinophilic (non-Type-2) disease — a distinction that is now recognized as one of the single strongest predictors of whether polyps will regrow after surgery.

• ~85%: Eosinophilic CRSwNP prevalence (in Western surgical cohorts)
• >55/HPF: High-risk eosinophil threshold (or >10% of inflammatory infiltrate)
• 60–80%: T2-high endotype recurrence (by 3 years without biologic therapy)
• 15–20%: Non-T2 endotype recurrence (by 3 years)

Reading the biopsy: counting eosinophils

Surgical pathology quantifies tissue eosinophils on hematoxylin & eosin (H&E) stained polyp sections, usually reported as eosinophils per high-power field (HPF, 400× magnification) or as a percentage of the total inflammatory cell infiltrate. Commonly used thresholds separate "eosinophilic CRSwNP" (tissue eosinophilia, often >10 eos/HPF as a minimum and >55 eos/HPF as a high-burden cutoff used in several outcome studies) from "non-eosinophilic CRSwNP," which is instead dominated by neutrophils or a mixed inflammatory infiltrate.

In Western cohorts, roughly 80–85% of CRSwNP is eosinophilic; in East Asian cohorts, the split is closer to even, with a larger non-eosinophilic, neutrophilic-predominant subgroup — a difference with real implications for how transportable Western-derived risk models are.

The Type-2 cytokine cascade

Eosinophilic CRSwNP is the end result of a Type-2 (T2) immune cascade:

• Epithelial alarmins (IL-25, IL-33, TSLP) are released by damaged sinonasal epithelium and activate group 2 innate lymphoid cells (ILC2s) and Type-2 helper T cells (Th2) • IL-5 drives eosinophil differentiation, recruitment, and survival — the single most eosinophil-specific cytokine in the pathway • IL-4 and IL-13 drive B-cell class switching to IgE, goblet cell hyperplasia, and epithelial barrier disruption, all of which favor polyp formation • Local IgE (often polyclonal, sometimes directed against Staphylococcus aureus enterotoxins) amplifies mast cell and basophil activation within the polyp stroma

This cascade is precisely the pathway targeted by the biologic therapies discussed in Stage 5: dupilumab blocks the shared IL-4/IL-13 receptor subunit, mepolizumab and reslizumab neutralize IL-5, and omalizumab neutralizes IgE.

Because the T2 cascade is self-amplifying and mucosal — not confined to the resected polyp tissue — surgical debulking removes the polyps but leaves the cytokine-producing epithelium largely intact. This is the biological reason eosinophilic disease recurs so much more often than non-eosinophilic disease after an anatomically identical operation.

Beyond histology: blood and molecular biomarkers

Because polyp biopsy at baseline is only available at the time of surgery, ongoing risk monitoring increasingly draws on more accessible biomarkers that correlate with the tissue endotype:

• Peripheral blood eosinophil count (>150–300 cells/µL is commonly used as a surrogate T2 marker) • Total serum IgE • Periostin, a matricellular protein induced by IL-13 signaling, used as a research-grade T2 biomarker • Nasal exhaled nitric oxide (nasal FeNO), elevated in T2-high sinonasal and airway disease

None of these single markers is sufficient alone — each has substantial overlap between endotypes — which is precisely why Stage 3 combines several inputs into a single weighted model rather than relying on any one threshold.

Multivariate Risk Model — Combining Clinical & Biomarker Inputs

No single risk factor reliably predicts recurrence on its own. Modern CRSwNP recurrence prediction combines several weighted clinical and biomarker inputs — using either classical regression or machine-learning methods — into a single personalized probability, analogous to cardiovascular risk calculators like the Framingham score.

• 0.78–0.85: Reported model discrimination (AUC) (typical published logistic models)
• 6–10: Variables in typical models (clinical + biomarker features)
• +8–12%: ML vs single-threshold uplift (AUC gain over single-variable cutoffs)
• Limited: External validation cohorts (few models validated beyond one center)

Regression-based scoring

The most widely published approach uses multivariable logistic regression, fitting a model of the form:

logit(P) = β₀ + β₁(eosinophilia) + β₂(AERD) + β₃(revision count) + β₄(Lund-Mackay) + …

Each coefficient (β) is estimated from a retrospective cohort with known recurrence outcomes, and the resulting probability P is interpretable and easy to communicate to patients ("your risk of regrowth by 12 months is approximately X%"). The strength of regression models is transparency — clinicians can see exactly how much each factor contributes — at the cost of assuming each factor acts independently and additively, which is only an approximation of the underlying biology.

Machine-learning approaches

More recent studies have applied random forests, gradient-boosted trees, and small neural networks to the same prediction task, sometimes adding higher-dimensional inputs such as full blood eosinophil trajectories, mucus cytokine panels, or radiomic features extracted directly from CT imaging.

These models can capture interactions that additive regression misses — for example, the combination of high tissue eosinophilia with AERD may carry more than the simple sum of their individual risks, since AERD amplifies the eosinophilic cascade through disrupted arachidonic acid metabolism (excess leukotriene production, prostaglandin D2 dysregulation). Reported gains in discrimination (AUC) over single-variable thresholds are modest but consistent — typically 8–12 percentage points — and come with a well-recognized tradeoff: ML models are harder to interpret and are more prone to overfitting when trained on the modest sample sizes typical of single-center ENT cohorts.

Whichever modeling approach is used, external validation remains the field's biggest gap — most published CRSwNP recurrence models have only been tested at the single center where they were developed, so a given risk score should be treated as a decision-support estimate, not a certainty, until broader validation exists.

From weighted inputs to a single score

Regardless of the underlying statistical method, the practical output is the same: each input is transformed into a relative contribution to risk, and these contributions are combined into one score. The relative risk (RR) or hazard ratio (HR) associated with each factor — drawn from published CRSwNP outcome cohorts — gives a sense of how strongly each input should be weighted.

Risk factors and their approximate relative risk for recurrence

Modeling the Recurrence Trajectory — 12 to 24 Months

A single risk percentage understates how recurrence unfolds — the hazard of polyp regrowth is not constant after surgery, but rises through a defined window and then plateaus. Projecting the personalized risk score forward in time produces a trajectory that is far more useful for planning follow-up intervals and maintenance therapy than a single number.

• ~40%: 12-month recurrence (eosinophilic) (endoscopic regrowth without maintenance)
• 60–70%: 24-month recurrence (eosinophilic) (without maintenance therapy)
• ~20%: 24-month recurrence (non-eosinophilic) (substantially lower baseline hazard)
• 6–18 mo: Peak hazard window (post-FESS)

Time-to-event modeling in CRSwNP

Because not every patient is followed for the same length of time, and because "recurrence" is an event that happens at a particular moment rather than a fixed attribute, CRSwNP outcome studies typically use survival-analysis methods: Kaplan-Meier curves to describe population-level time-to-recurrence, and Cox proportional-hazards regression to estimate how each risk factor accelerates or delays that event.

The mucosa is not equally vulnerable at every timepoint after surgery. Immediately post-op, topical steroid penetration is excellent and the surgical cavity is actively healing under close endoscopic surveillance. The hazard of clinically apparent regrowth rises over the following months as healing completes and any untreated inflammatory drive re-establishes itself, typically peaking between 6 and 18 months, before plateauing as the "high responders" have already relapsed and a more indolent remainder remains.

From population curves to individualized trajectories

A Cox model allows the population-average survival curve to be individualized: a patient's combined hazard ratio (the product of the hazard ratios for each risk factor present) shifts their personal curve up or down relative to the population baseline. A patient with eosinophilic disease and AERD does not just have a higher 12-month number — their entire curve is compressed, reaching a given cumulative risk months earlier than a low-risk patient reaches the same threshold.

This has direct clinical implications for follow-up scheduling: high-projected-hazard patients merit endoscopic re-examination as early as 8–12 weeks post-op, while low-risk patients can often be safely spaced to 6- or 12-month intervals.

Endoscopic polyp regrowth is frequently detectable months before a patient reports symptomatic obstruction, because early polypoid change in a wide post-FESS cavity does not immediately compromise airflow. A forward-projected risk trajectory is what allows a clinician to intervene at the "silent regrowth" stage rather than waiting for symptoms to return.

Uncertainty and confidence bands

Any projected trajectory carries uncertainty, driven by both statistical sampling error in the underlying model and genuine biological variability between patients. Well-reported risk models present a projected curve alongside a confidence band, widening with time since surgery — reflecting the reality that near-term predictions (3–6 months) are considerably more reliable than distant ones (24+ months), where unmeasured factors such as environmental allergen exposure, treatment adherence, and intercurrent respiratory infections increasingly dominate the outcome.

Personalized Post-Operative Maintenance Therapy

The entire purpose of recurrence prediction is to act on it. Once a patient's risk tier is established, maintenance therapy is matched to that tier — ranging from simple saline irrigation, through high-volume topical corticosteroid irrigation and steroid-eluting stents, up to systemic biologic therapy for the highest-risk eosinophilic, AERD-associated phenotype.

• −2.06 pts: Dupilumab NPS reduction (SINUS-52) (24-week nasal polyp score vs placebo)
• ~83%: Dupilumab surgery/OCS need reduction (vs placebo over 52 weeks)
• ↓~35%: Steroid-eluting stent reintervention (vs standard care, pooled trial data)
• ↓~50%: High-adherence irrigation benefit (relative recurrence risk vs low adherence)

Topical corticosteroid irrigation — the maintenance backbone

For low-to-moderate risk patients, high-volume budesonide or mometasone sinus irrigations (as opposed to simple nasal sprays, which poorly penetrate a post-FESS cavity) remain the first-line maintenance therapy. Irrigation delivers topical steroid directly onto the sinus mucosa at volumes (200 mL) far exceeding spray delivery, and adherence is one of the strongest modifiable predictors of outcome — cohorts comparing high versus low irrigation adherence report roughly a 50% relative reduction in recurrence risk in adherent patients.

Saline irrigation alone, without steroid, still provides mechanical clearance of mucus, allergens, and biofilm, and is appropriate maintenance for the lowest-risk, non-eosinophilic patients.

Biologics matched to the Type-2 endotype

For high-risk eosinophilic and AERD patients, biologic therapy targets the T2 cascade described in Stage 2 directly:

• Dupilumab (anti-IL-4Rα, blocking both IL-4 and IL-13 signaling): in the pivotal SINUS-24/SINUS-52 trials, dupilumab reduced nasal polyp score by roughly 2 points more than placebo and reduced the need for further surgery or systemic corticosteroid courses by approximately 83% over 52 weeks • Omalizumab (anti-IgE): reduced polyp score and improved sinus opacification in the POLYP1/POLYP2 trial program, particularly effective in patients with elevated IgE • Mepolizumab (anti-IL-5): reduced need for surgery and improved polyp score in the SYNAPSE trial, most effective in patients with high blood/tissue eosinophilia

Biologic selection is increasingly endotype-matched — a patient with markedly elevated tissue eosinophils but modest IgE may respond preferentially to an IL-5 pathway agent, while a patient with high IgE and comorbid allergic disease may be better matched to anti-IgE therapy.

Biologics are not curative and are typically continued indefinitely — trial data show that clinical benefit regresses toward baseline within roughly 6 months of discontinuation, reinforcing that maintenance therapy, once started in a high-risk patient, is a long-term commitment rather than a short course.

Steroid-eluting stents and the risk-tiered algorithm

Bioabsorbable steroid-eluting sinus stents, placed in the ethmoid cavity at the time of surgery or during a subsequent office visit, provide sustained local mometasone release over roughly 30–90 days directly at the site most prone to early scarring and polypoid regrowth. Pooled trial data (including the PROPEL program) show approximately a 35% relative reduction in the need for post-operative intervention (adhesiolysis, polypectomy, or oral steroid courses) compared with standard post-operative care.

Bringing the stages together, a simplified risk-tiered maintenance algorithm follows the projected 12-month recurrence risk from Stage 4:

• <20% (low risk) → saline irrigation, routine follow-up • 20–50% (moderate risk) → high-volume topical steroid irrigation ± steroid-eluting stent • >50% (high risk, typically eosinophilic + AERD/asthma) → topical steroid irrigation plus early biologic therapy, with shortened endoscopic follow-up intervals

This tiering is precisely what the risk model in Stage 3 and the projected trajectory in Stage 4 are built to support — matching the intensity, and cost, of therapy to each patient's actual biological risk rather than treating every post-FESS patient identically.