Cost-utility and clinical impact of endoscopic screening for esophageal and gastric neoplasia in patients with head and neck neoplasms

Abstract

Objective

Patients with head and neck neoplasms (HNN) are at an increased risk of esophageal neoplasia (EN) and gastric neoplasia (GN). We aimed to assess the clinical impact and cost-utility of endoscopic screening in this population in the Western setting.

Methods

In this single-center study HNN patients eligible for curative treatment underwent screening esophagogastroduodenoscopy. We assessed the frequency, clinical, and pathological outcomes of EN and GN. The cost-effectiveness of an annual endoscopic screening for EN was evaluated from a societal perspective, using a Markov model and probabilistic sensitivity analysis. In addition, we performed a sensitivity analysis using data on the prevalence of detected EN lesions in the four largest previous Western studies on this topic.

Results

Forty-six HNN patients met the inclusion criteria and underwent endoscopic screening. Six EN were detected in five patients (10.9%, 95% confidence interval: 1.9–19.9%). Additionally, five GN were detected in five patients. Most patients had early-stage EN or GN (90%) and were treated with endoscopic resection (80%). Endoscopic screening strategy had an incremental cost-effectiveness ratio of 39 357.8 €/quality-adjusted life years gained, being cost-effective at a willingness-to-pay threshold of two times the Portuguese gross domestic product per capita. In the sensitivity analysis, it remained cost-effective when considering the prevalence of EN reported in Germany, France, and Brazil.

Conclusion

An endoscopic screening program identified EN or GN in a fifth of HNN patients, most presenting at an early stage. The program implementation appears to be cost-effective in Portugal. These results may be applicable to other medium-to-high-income Western countries.

Introduction

Esophageal and gastric cancers are among the most prevalent and lethal malignancies worldwide [1]. Screening for gastroesophageal neoplasia (GEN) is recommended and was found to be cost-effective in high-incidence Asian countries [2]. In Western countries, according to the European Society of Gastrointestinal Endoscopy, gastric neoplasia (GN), screening by esophagogastroduodenoscopy (EGD) may be considered in intermediate-to-high-risk regions, depending on local settings and availability of endoscopic resources [2]. Screening for esophageal neoplasia (EN), namely squamous cell carcinoma (SCC), however, has only been suggested in high-risk populations such as patients with achalasia or with head and neck neoplasms (HNN) [2].

Patients with HNN are at increased risk for a second primary EN (either synchronous or metachronous), due to shared risk factors including smoking and alcohol consumption. The role of human papillomavirus (HPV) infection in this context is controversial, especially in the absence of additional risk factors [3,4]. The co-occurrence of an EN is associated with a poor prognosis of HNN, regardless of how appropriately the latter is managed [5]. Therefore, early detection and management of concurrent EN by minimally invasive endoscopic resection may potentially improve patients’ outcomes.

Several studies, mainly from Asia, have evaluated the diagnosis and management of EN in patients with HNN [6–12]. The applicability of these results in Western populations is, however, uncertain due to differences in the incidence of HNN and EN. Studies conducted in the Western setting showed contradictory results, and a wide debate is still ongoing regarding (1) the impact of screening programs on patients’ outcomes, (2) the timing and frequency of screening, and (3) the cost-effectiveness of screening all HNN patients vs. selected high-risk individuals [13–20].

Therefore, we aimed to evaluate the clinical impact and cost-effectiveness of an endoscopic screening program for EN in HNN patients, in a Western setting. Besides, we sought to evaluate the additional diagnosis of GN and to perform a complete histopathological characterization of all detected GEN and matched HNN.

Materials and methods

Study design

This study is a retrospective analysis of prospectively collected data that evaluated an endoscopic screening program for HNN patients in a tertiary referral center (Unidade Local de Saúde São João - ULSSJ, Porto, Portugal). The study was reported according to Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [21].

Eligibility criteria

We included all adult patients with HNN (from the oral cavity, nasopharynx, oropharynx, hypopharynx, larynx, and paranasal sinus) diagnosed between June 2020 and December 2023 and proposed for curative-intent treatment.

HNN diagnosis and staging were obtained by flexible transnasal endoscopy and computed tomography scan (including chest scan), which were performed in all patients. MRI (depending on the location) and PET were also performed when deemed necessary [22].

We excluded patients (1) with nonepithelial HNN or (b) with clinical or imaging suspicion of GEN before the first EGD.

Study endpoints

The main endpoints were (1) the frequency of EN in patients with HNN undergoing endoscopic screening and (2) the cost-utility of an endoscopic screening strategy for EN vs. no screening strategy in patients with HNN.

Additional endpoints included (1) evaluation of the frequency of GN, (2) assessment of the treatment and clinical outcomes after the diagnosis of GEN, and (3) histopathological characterization of the detected GEN and matched HNN.

Screening and follow-up

The strategy consisted of the performance of a screening EGD in all included patients, within 12 months since the HNN diagnosis. Whenever possible, a synchronous strategy (≤6 months) was conducted. Careful inspection of the esophageal mucosa with white light imaging, narrow-band imaging (NBI), and Lugol’s staining chromoendoscopy [20 ml of Lugol’s staining (1.2% iodine solution)] was performed in all patients.

Visible lesions were characterized using the Paris Classification [23]. The surface and vascular patterns were carefully evaluated using NBI and magnification to assess the possibility of endoscopic resectability. Targeted biopsies were performed and located in separate vials in all suspicious lesions.

Random biopsies of the antrum and gastric body using separate vials were obtained at the endoscopist’s discretion.

When no lesion was identified, annual endoscopic screening was proposed. Treatment was pursued in the cases with histopathological confirmation of GEN [esophagus: low-grade dysplasia (LGD), high-grade dysplasia (HGD), SCC, or adenocarcinoma; stomach: LGD, HGD, or adenocarcinoma]. After GEN detection, staging and treatment ensued, according to a multidisciplinary team and patient decision.

Endoscopic submucosal dissection (ESD) was used to manage lesions amenable to endoscopic resection and curativeness of the resection was defined according to current guidelines [24].

Histopathological analysis

A full description of the histopathological analysis can be found in Supplementary Table 1, Supplemental digital content 1, http://links.lww.com/EJGH/B159. All HNN and GEN cases were reviewed by at least two expert pathologists (I.G. and F.C.). A detailed histopathological characterization of the detected GEN and matched HNN in each patient was performed, with evaluation of biomarkers of viral infections, namely p16 immunohistochemistry for HPV and in situ hybridization for Epstein–Barr virus (EBER-ISH). This analysis aimed to assess if an infection by one (or both) of these oncogenic viruses could explain the presence of synchronous or metachronous neoplasias, besides the traditional risk factors common to both tumor models.

Cost-utility analysis

Target population and model structure

A health economic analysis was performed to evaluate the cost-utility of an endoscopic screening program (vs. no screening) for the diagnosis of EN in patients with HNN in Portugal. The societal perspective was adopted in agreement with the recommendations for reporting cost-effectiveness analyses [25]. We opted for a cost-utility analysis to simultaneously account for the life years gained and the quality of life associated with the corresponding health states. The consequences of the strategies being compared were expressed in quality-adjusted life years (QALY). The Consolidated Health Economic Evaluation Reporting Standards 2022 was followed for reporting the present analysis [26].

The target population included patients ≥18 years with stage I–III HNN. A Markov state-transition cost-utility model was designed (Fig. 1 and Supplementary Figure 1, Supplemental digital content 1, http://links.lww.com/EJGH/B159). The screening strategy consisted of the performance of an EGD within 12 months of the HNN diagnosis. In the case of a negative initial examination, follow-up EGD was proposed yearly for 5 years (time horizon). The nonscreening strategy implied that patients were nonroutinely screened for EN. In this case, EN was diagnosed following (1) an EGD performed due to other reasons or (2) a complete evaluation following after EN signs or symptoms.

Fig. 1.

Markov model illustration of a screening (a) vs. no screening strategy (b) in HNN patients, for the diagnosis of esophageal neoplasia (EN). A Markov model was used to project the costs and clinical outcomes for a time horizon of 5 years. The state-transition model simulated EN natural history as annual transitions between different health status. We assumed death as an absorbing state. *Preoperative chemoradiotherapy followed by surgery according to the CROSS study [Supplementary Table 2]. BSC, best supportive care; CHT, chemotherapy; CRT, chemoradiotherapy; ESD, endoscopic submucosal dissection; HNN, head and neck neoplasia.

Costs were calculated using data from our center’s Financial Department (ULSSJ, Porto, Portugal) and, whenever necessary, from national sources (Ministry of Health, Portugal).

Costs were calculated in 2023 Euros (€), with inflation adjustment as necessary. In accordance with published recommendations, a discount rate of 3% was incorporated for both costs and effectiveness results [25].

Supplementary Table 2, Supplemental digital content 1, http://links.lww.com/EJGH/B159 summarizes the model inputs for the computation of costs, utilities, and transition probabilities. Modeling assumptions are described in Supplementary Table 3, Supplemental digital content 1, http://links.lww.com/EJGH/B159.

Outcomes and analysis

The effect size used to present our results was the incremental cost-effectiveness ratio (ICER). The main willingness-to-pay (WTP) threshold was set at 50 000 €/QALY gained, as it approximately corresponds to two times the Portuguese gross domestic product (GDP) per capita (2 × 24 646 = 49 292 €). In addition, we considered WTP of one or three times the Portuguese GDP per capita as recommended by the WHO [27].

We performed a probabilistic sensitivity analysis using Monte Carlo simulation methods, with 10 000 simulations run per model. From the probabilistic sensitivity analyses performed, we retrieved the proportion of simulations identifying screening as cost-effective.

Additionally, we performed a sensitivity analysis considering the frequency of detected EN (at each oncological stage) in the four largest Western studies on this topic [13–16]. Results were expressed following probabilistic sensitivity analyses.

Statistical analysis

Categorical variables were described through absolute and relative frequencies, while continuous variables were reported as means and standard deviations or medians, interquartile ranges (IQR), and minimum and maximum values. The Fisher’s exact test or the chi-square test was used to compare categorical variables. Continuous variables with normal distribution were tested using the independent samples t test or the one-way ANOVA test, while continuous variables with nonnormal distribution were tested using the Mann–Whitney or the Kruskal–Wallis tests. All data were analyzed using SPSS v.28.0, IBM, Chicago, Illinois, USA and TreeAge Pro 2023 Healthcare Version, TreeAge Software, LLC, Williamstown, Massachusetts, USA.

Ethical considerations

Each patient gave written informed consent before EGD. The Ethics Committee of ULSSJ approved the study (project number 375/2023).

Results

During the study period, 207 patients diagnosed with HNN were evaluated in our institution, with 62 (30%) being evaluated in the gastroenterology outpatient clinic (median referral time since HNN diagnosis = 37 days, IQR = 8–82) (Fig. 2). The most common reason for nonreferral was the HNN not being eligible for curative treatment. After referral, 16 patients (25.8%) refused to undergo screening EGD. Forty-six patients underwent screening EGD after a median time of 4 months (IQR = 1–9) since HNN diagnosis [≤6 months (n = 31, 67.4%); 6–12 months (n = 15, 32.6%)].

Fig. 2.

Flow diagram of participant selection and clinical outcomes. CRT, chemoradiotherapy; EGD, esophagogastroduodenoscopy; ESD, endoscopic submucosal dissection; HNN, head and neck neoplasia; RT, radiotherapy. *Melanoma n = 3, central giant cell granuloma n = 2, ameloblastoma n = 1, and non-Hodgkin lymphoma n = 1. **Confirmed diagnosis of esophageal squamous cell carcinoma after EGD. +One patient had two synchronous esophageal neoplasias (one lesion with high-grade dysplasia, other with squamous cell carcinoma). ++The patient with two synchronous esophageal neoplasias underwent ESD of both lesions, with the one with squamous cell carcinoma being a noncurative resection. Note: None of the patients that underwent endoscopic screening had clinical or imaging suspicion of esophageal or gastric neoplasia.

Screened patients’ baseline characteristics

Most patients were male (n = 38, 82.6%) and had a history of tobacco use (n = 39, 84.8%) and alcohol consumption (n = 25, 54.3%), or both. The median age was 61 years (IQR = 60–67). We did not observe significant differences between patients with or without a diagnosis of GEN regarding demographic data, family history of GEN, or functional status (Table 1).

Table 1.

Demographic data and baseline head and neck neoplasms characteristics and staging of screened patients

	Total (N = 46)	GEN (n = 10)	No GEN (n = 36)	P
Demographic data
Sex, n (%)				0.055
Female	8 (17.4)	4 (40)	4 (11.1)
Male	38 (82.6)	6 (60)	32 (88.9)
Age, median (IQR), minimum–maximum, years	61 (60–67), 43–80	63 (63–64), 55–73	56 (54–74), 43–80	0.423
Family history of GEN, n (%)
Esophageal neoplasia	1 (2.2)	0 (0)	1 (2.8)	0.594
Gastric neoplasia	2 (4.3)	0 (0)	2 (5.6)	0.446
Personal history of cancer, n (%)*	6 (13)	2 (20)	4 (11.1)	0.598
CRC	2 (4.3)	2 (20)	0 (0)
Lung cancer	1 (2.2)	0 (0)	1 (2.8)
Bladder cancer	1 (2.2)	1 (10)	1 (2.8)
Non-Hodgkin lymphoma	1 (2.2)	1 (10)	1 (2.8)
Prostate cancer	1 (2.2)	1 (10)	0 (0)
Basal cell carcinoma	1 (2.2)	0 (0)	1 (2.8)
Tobacco use, n (%)				0.398
Current	20 (43.5)	3 (30)	17 (47.2)
Former	19 (41.3)	6 (60)	13 (36.1)
No	7 (15.2)	1 (10)	6 (16.7)
Tobacco pack years, median (IQR)	40 (29–50)	30 (20–40)	40 (30–50)	0.373
Alcohol consumption, n (%)				0.758
Current	14 (30.4)	4 (40)	10 (27.8)
Former	11 (23.9)	2 (20)	9 (25.0)
No	21 (45.7)	4 (40)	17 (47.2)
Alcohol grams per day, median (IQR)	50 (26–83)	26 (19–53)	50 (40–90)	0.199
Gastroesophageal reflux disease, n (%)	3 (6.5)	1 (10)	2 (5.6)	0.530
ECOG status				0.515
0	30 (65.2)	8 (80)	22 (61.1)
1	15 (32.6)	2 (20)	13 (36.1)
2	1 (2.2)	0 (0)	1 (2.8)
HNN characteristics and staging
Location of HNN, n (%)				0.853
Nasopharynx	1 (2.2)	0 (0)	1 (2.8)
Hypopharynx	7 (15.2)	2 (20)	5 (13.9)
Oral cavity	15 (32.6)	2 (20)	13 (36.1)
Larynx	19 (41.3)	5 (50)	14 (38.9)
Paranasal sinus	4 (8.7)	1 (10)	3 (8.3)
Histology, n (%)				0.134
Squamous cell carcinoma	42 (91.3)	8 (80)	34 (94.4)
Adenocarcinoma	3 (6.5)	1 (10)	2 (5.6)
High-grade squamous dysplasia	1 (2.2)	1 (10)	0 (0)
Stage†				0.190
I	8 (17.4)	3 (30)	5 (13.9)
II	8 (17.4)	2 (20)	6 (16.7)
III	11 (23.9)	1 (10)	10 (27.8)
IVa	18 (39.1)	3 (30)	15 (41.7)
NA	1 (2.2)	1 (10)	0 (0)
T				0.326
T1	10 (21.7)	3 (30)	7 (19.4)
T2	9 (19.6)	2 (20)	7 (19.4)
T3	12 (26.1)	2 (20)	10 (27.8)
T4	14 (30.4)	2 (20)	12 (22.2)
NA	1 (2.2)	1 (10)	0 (0)
N				0.044
N0	20 (43.5)	7 (70)	13 (36.1)
N1	7 (15.2)	1 (10)	6 (16.7)
N2	14 (30.4)	0 (0)	14 (38.9)
N3	4 (8.7)	1 (10)	3 (8.3)
NA	1 (2.2)	1 (10)	0 (0)
M				0.217
M0	45 (97.8)	9 (90)	36 (100)
M1	0 (0)	0 (0)	0 (0)
NA	1 (2.2)	1 (10)	0 (0)
HNN treatment, n (%)				0.480
CRT	15 (32.6)	2 (20)	13 (36.1)
Surgery	7 (15.2)	3 (30)	4 (11.1)
Surgery + RT	10 (21.7)	2 (20)	8 (22.2)
Surgery + CRT	14 (30.4)	3 (30)	11 (30.6)

Bold value indicates P < 0.005.

CRC, colorectal cancer; CRT, chemoradiotherapy; GEN, gastroesophageal neoplasia; HNN, head and neck neoplasms; IQR, interquartile range; NA, nonapplicable; RT, radiotherapy.

^*More than one cancer in two patients. All neoplasias were under remission at the beginning of follow-up.

^†Eighth edition of the American Joint Committee on Cancer/International Union Against Cancer TNM staging system.

Head and neck neoplasms characteristics

The most frequent HNN location was the larynx (n = 19, 41.3%), with the majority being SCC (n = 42, 91.3%). All patients underwent treatment, with the most common strategies being chemoradiotherapy (n = 15, 32.6%) and surgery plus chemoradiotherapy (n = 14, 30.4%).

At endoscopic screening, 31 patients (67.4%) were already in HNN remission, while the remaining (n = 15, 32.6%) were being treated with curative intent. Most patients underwent one screening EGD (n = 38, 82.6%), with the remainder undergoing two (n = 7; 15.2%) or three (n = 1; 2.2%) screening EGDs.

Gastroesophageal neoplasia detected with endoscopic screening

An EN or GN was detected in 10/46 patients (21.7%, 95% confidence interval (CI): 9.8–33.6%). All of these patients had a current or previous history of drinking, smoking or both. Histopathological characterization of these cases is reported in Table 2. Supplementary Table 4, Supplemental digital content 1, http://links.lww.com/EJGH/B159 describes a detailed individual analysis of each patient with GEN.

Table 2.

Characterization of esophageal and gastric neoplasia

Esophageal neoplasia (n = 5)
Histological grading, n (%)
Low-grade squamous dysplasia	2 (40)
Squamous cell carcinoma*	3 (60)
Lesion size, median (IQR), mm	12 (10–30)
Morphology, n(%)
Polypoid	0 (0)
Nonpolypoid, nondepressed	3 (60)
Depressed	2 (40)
p16 expression
Negative	3 (60)
Focal/heterogeneous	1 (20)
Positive†	1 (20)
EBER-ISH
Negative	5 (100)
Positive	0 (0)

Esophageal SCC (n = 3)
Histological subtype – WHO classification, n (%)
Esophageal squamous cell carcinoma, NOS	3 (100)
T staging, n (%)
T1b	2 (66.7)
SM1	1 (50)
>SM1	1 (50)
T2	1 (33.3)
Lymphovascular permeation, n (%)	0 (0)
Differentiation, n (%)
Poor	1 (33.3)
Well/moderate	2 (66.7)

Gastric neoplasia (n = 5)
Histological grading, n (%)
Low-grade dysplasia	1 (20)
Adenocarcinoma	4 (80)
Lesion size, median (IQR), mm	13 (15–24)
Location, n (%)
Antrum	1 (20)
Body	1 (20)
Cardia	3 (60)
Morphology, n (%)
Polypoid	1 (20)
Nonpolypoid, nondepressed	2 (40)
Depressed	2 (40)
p16 expression
Negative	3 (75)
Focal/heterogeneous	1 (25)
Positive*	0 (0)
NA	1 (20)
EBER-ISH
Negative	3 (75)
Positive	1 (25)
NA	1 (20)

Gastric adenocarcinoma (n = 4)
Histological subtype – Laurén classification
Intestinal	3 (75)
Undetermined	1 (25)
Histological subtype – WHO classification, n (%)
Gastric adenocarcinoma
Tubular	3 (75)
Gastric carcinoma with lymphoid stroma	1 (25)
T staging, n (%)
Tis	1 (25)
T1b	3 (75)
SM1	2 (66.7)
>SM1	1 (33.3)
Lymphovascular permeation, n (%)	1 (25)
Differentiation, n (%)
Poor	0 (0)
Well/moderate	4 (100)

EBER-ISH, Epstein–Barr virus encoding region in situ hybridization; IQR, interquartile range; NA, nonapplicable; NOS, not otherwise specified; SCC, squamous cell carcinoma.

^*One patient had a synchronous esophageal lesion with squamous high-grade dysplasia.

^†Cases were considered positive when block-positive p16 staining was observed in dysplastic lesions and when moderate to strong nuclear and cytoplasmic p16 staining was observed in at least 70% of cells in invasive lesions.

Esophageal neoplasia

A total of six EN were detected in 5/46 patients (10.9%, 95% CI: 1.9–19.9%). The median lesion size was 12 mm (IQR: 10–30). All cases were identified in the index EGD, ≤6 months after HNN diagnosis. Histopathological characterization revealed three cases of SCC, one of squamous HGD, and two of squamous LGD.

One patient with an SCC not amenable to endoscopic resection (T2N0M0) was referred for chemoradiotherapy. In one patient with an LGD lesion, surveillance was decided due to the metastatic progression of HNN. In the remaining three patients, ESD of the lesions (n = 4) was performed, being noncurative in two: one received subsequent treatment with adjuvant radiotherapy and the other underwent esophagectomy (with evidence of residual lesion with HGD in the surgical specimen).

No adverse eventss (AEs) were reported after endoscopic/surgical resection. No patient had evidence of recurrence/progression of the EN during the follow-up. One patient died (20.0%) in the follow-up due to an unrelated cause (pneumonia).

Histopathological analysis – p16 and Epstein–Barr virus encoding region evaluation

One patient had diffuse p16 expression in the esophageal SCC but not in the matched HNN (Supplementary Figure 2, Supplemental digital content 1, http://links.lww.com/EJGH/B159). One patient showed simultaneous focal p16 expression in both the EN and HNN. No p16 expression was observed in the remaining EN (n = 3), with the correspondent HNN being also p16 negative. No esophageal neoplasms were positive for EBER-ISH.

Gastric neoplasia

A total of five GN were detected in 5/46 patients (10.9%, 95% CI: 1.9–19.9%). Four lesions were identified in the index EGD (≤6 months after HNN diagnosis) while one lesion was detected in the second follow-up EGD (14 months after HNN diagnosis). Histopathological characterization revealed four cases of adenocarcinoma and one of LGD.

All five GN underwent ESD, with four being curative. One patient with a noncurative ESD underwent total gastrectomy, without evidence of residual disease in the surgical specimen. No significant differences between groups were observed regarding the staging of gastritis when comparing patients with or without GN (Supplementary Table 5, Supplemental digital content 1, http://links.lww.com/EJGH/B159).

No AEs were reported after endoscopic/surgical resection. No recurrence or progression of the GN was observed and no patient died in the follow-up.

Histopathological analysis – p16 and Epstein–Barr virus encoding region evaluation

Focal p16 expression was observed in one patient with gastric adenocarcinoma, whose matched HNN was negative for p16 expression. One case of gastric adenocarcinoma, showing the morphologic features of gastric carcinoma with lymphoid stroma, was associated with Epstein–Barr virus (EBV) infection, with the corresponding HNN being negative for EBER-ISH (Supplementary Figure 3, Supplemental digital content 1, http://links.lww.com/EJGH/B159).

Clinical outcomes of the screened patients

The median follow-up after EGD was 13 months (IQR = 5–20). Three patients (6.5%) were diagnosed with additional neoplasms during follow-up (pancreatic cancer = 1, lung cancer = 2).

Four patients (8.7%) refused to be enrolled in subsequent screening after the first EGD. Eleven patients (23.9%) died during follow-up, with no significant differences in the mortality rate between patients with or without GEN (10.0 vs. 27.8%, P = 0.244) – six (54.5%) due to progression of the HNN and five (45.5%) from causes unrelated to HNN or GEN.

Cost-utility analysis

Main analysis

Screening would be associated with an average increased cost of €1327.2 per patient and an average gain of 0.03 QALY (Table 3). This corresponds to an ICER of 39 357.81 €/QALY, rendering screening cost-effective at the WTP threshold of 50 000 €/QALY gained (as well as the WTP of three times the Portuguese GDP per capita). Supplementary Figure 4, Supplemental digital content 1, http://links.lww.com/EJGH/B159 depicts the overall survival, cumulative QALYs, and cumulative costs associated with the screening and no screening strategies.

Table 3.

Deterministic one-way sensitivity analysis and probabilistic sensitivity analysis results of the screening strategy for esophageal neoplasia vs. no screening

Country of origin of the study	Frequency of esophageal neoplasia*, n (%)	Patients with dysplastic lesions/stage I esophageal cancer, n (%)	Costs screening (Euro)	Costs no screening (Euro)	Differences in costs (screening vs. no screening) (Euro)	Effectiveness of screening (QALY)	Effectiveness of no screening (QALY)	Differences in effectiveness (QALY – screening vs. no screening)	ICER
Deterministic one-way sensitivity analysis
Portugal (current study)	5/46 (10.9)	4/5 (80)	3205.08	1327.15	1877.92	1.68	1.64	0.0337	39 357.81
Probabilistic sensitivity analysis
Portugal (current study)	5/46 (10.9)	4/5 (80)	4008.16	1580.52	2427.64	1.65	1.61	0.0451	35 044.79
Sensitivity analysis with data from other Western countries
Netherlands [18]	13/202 (6.4)	12/13 (92.3)	2536.89	1354.28	1182.61	1.69	1.66	0.022	52 839.41
Brazil [20]	181/1888 (9.6)	148/181 (81.8)	3067.62	1748.30	1319.32	1.66	1.63	0.031	42 854.96
Germany [21]	12/148 (8.1)	7/15 (46.7)	3421.74	1975.10	1446.64	1.66	1.63	0.033	43 360.71
France [19]	39/392 (9.9)	22/39 (56.4)	3934.92	2356.83	1578.09	1.64	1.64	0.042	37 514.43

An additional sensitivity analysis was performed taking into consideration the data from the main Western studies on this topic.

GDP, gross domestic product; ICER, incremental cost per QALY gained; QALY, quality-adjusted life years.

^*Besides lesions with esophageal squamous cell carcinoma includes also lesions with low-grade/high-grade dysplasia and esophageal adenocarcinoma.

In the probabilistic sensitivity analysis using data from the current study, screening would be associated with an average increased cost of €1580.52 per patient, as well as with an average gain of 0.05 QALY (ICER = 35 044.79 €/QALY) (Table 3).

Screening would have a 63% probability of being cost-effective at a WTP threshold of 50 000 €/QALY gained (Fig. 3a). For the WTPs of one and three times the Portuguese GDP per capita, the probabilities would respectively be of 34 and 72% (Fig. 3b).

Fig. 3.

(a) Scatter plot for probabilistic Monte Carlo sensitivity analysis of the endoscopic screening vs. no screening strategy. The scatter plot represents 10 000 simulated iterations in a Monte Carlo probabilistic sensitivity analysis where the x-axis represents incremental effectiveness in terms of quality-adjusted life years (QALY) and the y-axis represents incremental costs in Euros. The oval represents the confidence ellipse, in which 95% of the results of the simulations are present. Cost-effective simulations are in green and below the dotted line representing the willingness-to-pay threshold, set at €50 000/QALY gained. (b) Acceptability curve of the endoscopic screening vs. no screening strategy. The vertical lines represent the thresholds of acceptability [1–3 times the Portuguese gross domestic product (GDP) per capita]. The x-axis represents the willingness-to-pay in Euros per QALY and the y-axis represents the probability that screening is cost-effective.

Sensitivity analysis with data from other Western series

Probabilistic sensitivity analysis using data from other Western series showed that the screening strategy remained cost-effective at a WTP of 50 000 €/QALY gained if the prevalence of EN lesions was the same as reported in Germany, France, and Brazil (but not in the Netherlands) (Table 3).

Discussion

This study aimed to evaluate the impact and cost-utility of an endoscopic screening program in patients with HNN in a Western country with intermediate-to-high risk for GN. Our results showed an overall frequency of EN of 10.9%. The health economic analysis indicated that endoscopic screening for EN, performed within 12 months since diagnosis in stage I–III HNN patients, is cost-effective in Portugal. In addition, an unexpectedly high frequency of GN was evident.

Routine endoscopic screening in patients with HNN has yet to be implemented in most Western countries [2]. Our study showed a higher frequency of EN than previously reported in the other prospective Western studies (ranging between 1.8 and 10%) [13–20]. Nevertheless, many of these studies were performed before the advent of image-enhanced endoscopy, which is associated with a better diagnostic performance for detecting early EN, without a significant increase in cost per procedure [28]. In our study, all procedures were performed under sedation, with a standardized inspection of the esophageal mucosa using both virtual and conventional chromoendoscopy, which can partly explain our higher detection rates. Moreover, while the role of endoscopic screening for the diagnosis of EN has been extensively studied, very few studies addressed the opportunistic diagnosis of GN in patients with HNN (with rates ranging between 0.5 and 1.8%) [13,19,29], but in our cohort, it was unexpectedly high (10.9%). Portugal is a country with an intermediate-to-high risk for GN, and endoscopic screening for this neoplasia in the general population was found to be cost-effective [30]. A recent study from our group recently compared the frequency of GN among healthy individuals who underwent EGD at the same time of screening colonoscopy with that of patients with compensated advanced chronic liver disease who underwent EGD for screening of gastroesophageal varices. The frequency of GN was 1 and 4%, respectively [31]. In the present study, the synergic presence of additional risk factors for GN in patients with HNN, within a susceptible population may partially explain our findings. Interestingly, all patients with GN were negative for Helicobacter pylori infection. Unfortunately, systematic gastric biopsies were not performed in this cohort, and missing data prevented us from evaluating the relationship between GN frequency and gastritis stages.

To the best of our knowledge, no previous study has evaluated the cost-utility of an endoscopic screening program in HNN patients in the Western setting. Recently, a study from Taiwan reported EN screening among patients with oral cavity cancer (OCC) to be cost-effective [32]. In this study, four different strategies (annual, annual for 3 years, annual for 10 years, annual for life) were compared, with the latter being the most favorable from the payer’s perspective, particularly in patients with early-stage OCC. Nevertheless, it should be acknowledged that these findings may not be applicable in the Western setting for several reasons: (1) this study focused only on OCC excluding other HNN in which the risk of EN may be significantly higher (e.g. pyriform sinus); (2) differences in the incidence of EN; (3) significant divergence in costs (e.g. the cost of EGD in Taiwan is 47.42 € vs. 142.81 € in Portugal). In our analysis, an annual endoscopic screening program for the diagnosis of EN in patients with HNN was found to be cost-effective, for a WTP corresponding to between 2–3 times the Portuguese GDP per capita. Of note, while the gains in QALY associated with screening were relatively small, it is important to note that these patients (1) have a high risk of mortality in the follow-up due not only to the baseline HNN but also due to other risk factors such as second primary tumor (other than EN) and comorbidities and (2) tend to display a low quality of life. The probabilistic sensitivity analysis using data from other Western centers confirmed that the screening strategy remained cost-effective if the prevalence of detected lesions would be the same as the one detected in Brazil, Germany, and France. We, however, did not use specific cost data from these countries, so we cannot conclude whether screening would be cost-effective in those countries. The generalizability of the results to other Western countries should warrant further research. In the near future, screening may become even more cost-effective as a consequence of (1) the application of artificial intelligence tools [33] or (2) increased survival of HNN patients with novel targeted immunotherapies [34].

Many questions remain unsolved regarding patient selection, timing, and frequency of endoscopic screening in patients with HNN. A recent study conducted in the Netherlands evaluated screening for EN in HNN patients [13]. Compared with van Tilburg et al., we had almost twice the prevalence of EN (10.9 vs. 6.0%), if we account for the LGD cases in their work. The authors evaluated different timings for EGD, but most lesions were detected at the index EGD with the highest yield observed in synchronous screening (7.1%). Most of the patients (80%) were amenable to endoscopic treatment. Regarding our results, all the EN were detected at the index EGD and 66.7% underwent ESD. In our sample, 8.7% of patients refused subsequent screening and 23.9% died during our short follow-up period of 13 months. Nonetheless, 15.2% performed two more EGD and 2.2% performed three additional EGD with no additional lesion detection. This data suggests that at least one screening EGD should be recommended following an HNN diagnosis. Ideally, this index EGD should be performed within 6 months but may be acceptable up to 12 months. A clear definition of timing is crucial, as our findings, in line with those of van Tilburg, indicate that a significant proportion of patients in remission from HNN have a window of opportunity during which they are willing to undergo further procedures, like screening EGD. Delaying initial screening for 1–2 years should be discouraged because a lesion could progress to a stage where it is not amenable to endoscopic resection anymore [15]. Regarding long-term follow-up, previous studies reported metachronous EN rates ranging between 1.5 and 5.1%, which may continue to rise even after 10 years [18,35]. The decision to recommend follow-up EGD should be based on further studies that assess the yield of subsequent metachronous screening and the outcomes associated with therapeutic options that ensue. Therefore, taking all these factors into account, together with the findings of our study, follow-up EGD should be considered on a year-to-year basis, considering the natural history and location of HNN (particularly in those with a piriform sinus involvement, which is a known risk factor for EN [36]) and patients’ functional status and preferences.

In this study, we aimed to explore the role of viral infection as a driving factor of the co-occurrence of HNN and GEN, with the potential aim of more precisely selecting patients for screening. The viral-related pathogenesis associates high-risk HPV infection to oropharyngeal SCC and EBV infection to gastric carcinoma, particularly to the histological subtype of gastric carcinoma with lymphoid stroma [37,38]. Nevertheless, a previous study has suggested that patients with HPV-positive HNN have a lower risk of second primary tumors [4]. In our study, no significant matching between EBV or HPV infection in HNN and GEN was found, supporting that viral infection plays a limited role in the field cancerization phenomenon. Therefore, we are not able to suggest targeted recruitment of patients for EGD based on EBV/HPV status of the HNN tumor.

Our study has some limitations. A selection bias cannot be excluded since the screening program was implemented in a single tertiary reference center, which may limit the generalization of these results. Furthermore, the series included a small number of patients with HNN. While this is a consequence of the strict inclusion and exclusion criteria (which may have contributed, at least partially to an overestimation of the frequency of GEN), this also shows the difficulties of implementing a screening program in a difficult-to-manage population with significant comorbidities and frequent social issues. Finally, the short follow-up and the possible lead time bias limit our ability to assess the long-term impact on the patients’ prognosis. Regarding the cost-utility analysis, while we tried to evaluate all possible clinically relevant scenarios, the simultaneous presence of two neoplasias with distinct behaviors implied some model simplifications and assumptions. Nonetheless, it is important to highlight that a conservative approach was used in the estimations performed. In addition, the model did not consider EN recurrences in the follow-up. Considering that the risk of recurrence is higher in more advanced stages and is associated with additional costs and loss of productivity, this may have led to an underestimation of the cost-effectiveness of the model.

In conclusion, our study showed that the implementation of endoscopic screening led to the identification of EN or GN in one-fifth of HNN patients, mostly discovered at an early stage, allowing a curative-intent treatment. Annual endoscopic screening for EN may be considered cost-effective in the Portuguese setting. Future studies should focus on the optimization of patient selection, evaluation of the long-term clinical impact of screening, and assessment of its cost-effectiveness in other Western countries and of alternative strategies, such as one-time screening.

Acknowledgements

None.

Conflicts of interest

There are no conflicts of interest.