Longitudinal Stability of Chronic Rhinosinusitis Endotypes

To the Editor:

The identification and characterization of chronic rhinosinusitis (CRS) inflammatory endotypes has advanced the understanding of CRS pathophysiology with potentially important clinical implications^1–3. Building upon these innovative studies, our group recently characterized CRS inflammatory endotypes in a North American population using cluster analysis of cytokines measured in sinonasal mucus^4,5. Cluster analysis of 147 adults with medically refractory CRS revealed phenotypic and inflammatory signatures unique to the different groupings. Certain clusters were characterized by low levels of most assayed cytokines, milder disease phenotypes with lower proportions of comorbid asthma and nasal polyps, and greater postoperative improvement in disease-specific quality of life (QoL) scores⁵. Conversely, other clusters demonstrated more severe disease and poor long-term response to surgical intervention. Despite the apparent clinical implications of these findings, the longitudinal stability of CRS endotypes is not yet known. The aim of this investigation was to assess the stability of previously validated inflammatory CRS endotypes over time and to determine the effect of surgical intervention on inflammatory burden and endotypic assignment.

We resampled twenty (10 CRSsNP, 10 CRSwNP) of the 147 subjects analysed in our initial study using the same minimally invasive approach⁴ during routine follow-up (Supporting Table S1). Mucus biomarkers have several advantages over surgically-obtained tissue in its minimally-invasive, yet comprehensive sampling of inflammatory mediators without being limited by potential variability in site-specific protein expression⁶. Repeat mucus collection was performed no sooner than 6 months after initial sampling (median = 15.5 months; range 10–41 months), the point following which few changes in postoperative disease-specific QoL are thought to occur and a suggested primary end point of CRS clinical studies⁷. As previously described, inflammatory burden in individual samples was measured using a multiplex assay that reflects various T_H1/T_H2/T_H17-associated cytokines⁵ (Table 1).

Table 1.

Cytokine and inflammatory mediator concentrations (pg/mL).

	CRS Phenotype						All
	CRSwNP (n=10)			CRSsNP (n=10)			(n=20)
	Preoperative	Postoperative	P value	Preoperative	Postoperative	P value	Preoperative	Postoperative	P value
IL-1β	120.8 (55.6–459.1)	174.0 (24.9–496.1)	.70	947.4 (121.1–4411)	177.2 (7.9–3985)	.32	175.4 (79.6–1404)	177.2 (10.3–637.1)	.65
IL-2	11.1 (2.2–47.8)	19.1 (2.2–34.2)	.91	2.6 (2.2–15.9)	15.2 (4.3–173.9)	.16	8.0 (2.2–27.7)	15.2 (2.4–38.2)	.25
IL-4	1.5 (0.3–3.0)	0.5 (0.3–1.9)	.28	0.3 (0.3–2.2)	1.2 (0.3–1.9)	.47	0.5 (0.3–2.3)	0.6 (0.3–1.7)	.68
IL-5	383.9 (67.1–474.9)	16.1 (2.7–275.3)	.01	1.4 (0.2–2.5)	1.4 (0.8–6.0)	.36	6.6 (1.1–385.7)	5.9 (1.1–55.8)	.08
IL-6	448.8 (72.8–1461)	596.6 (144.6–2792)	.77	218.0 (80.6–1045)	401.0 (98.1–2308)	.49	240.4 (82.3–1105)	462.7 (120.9–1972)	.45
IL-7	5.9 (2.8–18.3)	28.2 (13.7–73.3)	.03	14.6 (2.2–43.8)	34.9 (12.2–86.8)	.19	6.6 (2.9–21.6)	28.2 (13.9–77.8)	.02
IL-8	5817 (2506–14113)	38634 (27969–72604)	.08	45221 (8296–100717)	74844 (28813–101279)	.56	12236 (4229–46040)	47125 (28320–89654)	.11
IL-9	12.2 (0.6–29.6)	2.9 (0.6–8.9)	.46	1.6 (0.6–7.2)	3.6 (1.1–21.3)	.19	2.9 (0.6–18.5)	3.4 (0.8–10.7)	.71
IL-10	14.3 (3.7–27.1)	9.4 (6.0–19.8)	.63	6.4 (0.3–22.6)	13.1 (4.2–29.9)	.18	12.6 (1.0–23.1)	9.9 (5.6–20.4)	.74
IL-12	68.2 (6.9–91.8)	17.0 (9.1–50.1)	.11	86.5 (22.7–337.9)	42.0 (5.4–78.7)	.15	73.2 (13.8–104.6)	36.2 (7.2–55.7)	.07
IL-13	198.4 (64.0–388.0)	18.7 (11.1–85.0)	.01	12.1 (0.3–50.9)	9.1 (2.0–16.9)	.21	55.7 (4.3–211.0)	14.4 (5.0–23.0)	.01
IL-17A	2.8 (0.1–8.9)	3.2 (0.3–7.8)	.76	0.1 (0.1–5.5)	3.1 (1.5–6.9)	.74	1.6 (0.1–7.7)	3.1 (0.4–5.9)	.56
IL-21	75.7 (32.0–163.1)	182.0 (76.1–219.5)	.49	101.9 (6.9–259.2)	175.7 (73.6–325.5)	.99	75.7 (10.9–234.8)	175.7 (96.5–241.3)	.62
TNF-α	14.2 (3.0–30.1)	14.2 (7.7–40.1)	.80	5.4 (0.8–18.5)	11.9 (5.1–29.6)	.23	7.4 (1.4–18.5)	12.3 (6.1–38.1)	.35
IFN-γ	2.1 (0.7–4.7)	3.6 (2.0–9.5)	.06	0.5 (0.4–6.4)	4.6 (2.6–7.5)	.02	1.2 (0.4–4.7)	3.9 (2.6–7.5)	.003
Eotaxin	41.1 (8.3–83.5)	99.2 (18.5–253.0)	.49	27.1 (3.3–79.2)	63.3 (48.5–103.7)	.13	36.2 (7.1–79.9)	63.3 (35.0–176.9)	.11
RANTES	2240 (308.1–4855)	71.3 (24.1–278.9)	.01	495.4 (116.8–4929)	190.0 (103.4–308.7)	.11	1173 (192–4007)	162 (54.3–287.3	.001

CRS, chronic rhinosinusitis; CRSsNP, chronic rhinosinusitis without nasal polyposis; CRSwNP, chronic rhinosinusitis with nasal polyposis; TNF, tumor necrosis factor

Data are represented as medians with interquartile ranges. Boldface text indicates a p value of less than .05.

We first performed paired comparisons between matched samples to determine whether individual cytokine postoperative levels changed over time. The overall cohort was marked by significant increases in IL-7 (p=.02) and IFN-γ (p=.003) and significant reductions in IL-13 (p=.01) and RANTES (p=.001). Most changes were driven by patients with nasal polyps. Longitudinal evaluation of inflammatory burden in CRSwNP patients showed substantial reductions in IL-5 (p=.01), IL-13 (p=.01), and RANTES (p=.01), while CRSsNP patient were characterized by an increase in IFN-γ (p=.02).

We next sought to determine whether temporal changes in inflammatory biomarkers affect the endotypic assignments of each resampled subject. As previously reported, principal component analysis was performed on the entire data set to evaluate associations between immunologic variables^4,5, with a 5-factor solution explaining 67% of the variance (Supporting Table S2). Matched baseline and follow-up sampling of representative cytokines for each rotated factor are presented in Figure 1A. Using the factor scores calculated for each patient, hierarchical cluster analysis was again performed for all 147 patients repeated after replacing the original data for the 20 resampled patients with their repeat cytokine measurements (Figure 1B). Consistent with our prior study, analysis again converged on a five cluster model, validated using statistical modelling and bootstrapping. Alteration of the dataset with the resampled patients resulted in minor changes to dendogram nodal patterns and subject allocation, but a similar architecture to the original analysis was maintained. The majority of patients (89%) who were not resampled were assigned to the same disease cluster as the original analysis, suggesting stability of the model and previously validated endotypes.

Figure 1.

(A) Concentrations of representative inflammatory mediators from each of the five rotated factors at baseline and after repeat sampling. (B) Side-by-side comparison of original baseline dendogram derived from hierarchical cluster analysis from initial sampling (left) and dendogram after repeat sampling of 20 subjects (right).

Conversely, 13 of 20 resampled subjects (65%) switched disease clusters (Supporting Table S3). There was a notable convergence of resampled subjects into cluster 5, which is characterized by mild-moderate disease severity, good olfactory function, and low tissue eosinophilia^4,5. This cluster is further defined by low overall inflammatory burden, but the highest levels of IL-7 and lowest levels of RANTES. This disproportionate reassignment of subjects to cluster 5 may derive from relatively consistent changes in these cytokines in the overall cohort and from a general reduction in inflammatory burden in postoperative patients. The likelihood of cluster switching was independent of demographic factors, symptom burden, and both polyp and asthma status (Supporting Table S4).

To our knowledge, this is the first study to evaluate the long-term stability of CRS inflammatory endotypes and following surgical intervention. Surgery in patients with nasal polyps appears to result in profound and long-term reductions in the T_H2-associated cytokines IL-5 and IL-13. These changes are comparable or greater to changes in T_H2 mediators observed in CRS patients after treatment with dupilumab⁸ and may reflect substantial changes in the underlying immune environment after surgical and medical management. In contrast, a majority of patients with or without polyps experienced a significant elevation in IFN-γ. Taken together, these changes in cytokine levels suggest surgical intervention may ultimately shunt inflammatory burden from a T_H2- to T_H1 pattern. Collectively, our data suggests that surgery results in permanent alterations in CRS inflammatory burden that can change endotypic assignment. This observation has clinical relevance given that patients with a history of prior surgery accounted for approximately 30–60% of subjects in recent CRS endotyping studies^2–5.

Limitations to the current study include its small sample size and variable follow-up. Future investigations with increased study subjects may more specifically characterize to what degree differences in the interval between surgery and resampling among patients may impact cluster analysis. Because all patients in the current study underwent endoscopic sinus surgery between initial and follow-up sampling, it is difficult to dissociate temporal effects and other factors from surgery itself. Additional work in this area would also benefit from a comparison group of recalcitrant, nonsurgical CRS patients and from a focused evaluation of potential associations between cytokine levels, cluster assignment, and long-term symptom burden.

Recent investigations have expanded understanding of CRS pathophysiology and biomarkers largely by improving CRS classification schemes and defining putative CRS endotypes¹. These methods have potential both in directing patients to optimal therapeutic approaches based on inflammatory profiles and in following response to treatment. Sampling of nasal mucus is minimally invasive, easily collected, and facilitates monitoring of inflammatory changes over time. While the current study analysed a comparatively small sample size, it suggests that endotypic assignments are not static and can potentially be impacted by clinical and environmental factors. Future prospective studies to evaluate endotype stability over time and in response to corticosteroids, immunomodulating therapy, and surgery are needed. Ultimately, incorporation of dynamic (cytokines, inflammatory biomarkers, etc.) and static (polyp/asthma status, sex, etc.) variables into CRS endotyping may improve stability and its predictive ability⁹.