Risk factors for second primary neoplasia of esophagus in newly diagnosed head and neck cancer patients: a case–control study

Squamous cell carcinomas of the head and neck (H&N) region and the esophagus are common dismal malignancies globally, especially in the Western Pacific regions where carcinogen uses such as drinking alcohol, cigarette smoking and betel quid chewing are prevalent [1]. Mucosa of the upper aerodigestive tract (UADT) may be exposed to common carcinogens and at risk of early molecular alterations without histopathological changes, followed by malignant transformation [2]. According to population-based analyses, the risk and incidence of second primary cancers of the index head and neck, or esophagus are quite high [3‐5], and without early detection, synchronous or metachronous carcinogenesis may lead to poor prognosis despite multidisciplinary treatment of the primary cancers [6‐8]. Even with traditional panendoscopy screening for synchronous esophageal cancer, the treatment result was still poor [6].

Recent advances in image-enhanced endoscopy (IEE) have enabled precancerous or early cancerous lesions visible more easily under endoscopic examination [9]. Using IEE examination, especially chromoendoscopy with Lugol’s solution and narrow-band imaging (NBI) system with high-resolution magnifying endoscopy (ME), dysplastic or cancerous lesions, and tumor invasion could be well delineated [9‐17]. The prevalence of high grade intraepithelial neoplasia or invasive carcinoma of the esophagus in population at high risk, such as alcoholics or H&N cancer patients, is around 3.2 to 28% by IEE screening [11]. To triage and allocate H&N cancer patients at higher risk for esophageal neoplasia to surveillance program becomes important and cost-effective. However, routine application of endoscopic surveillance of esophagus is not a consensus nowadays, and which population benefit from the screening policy has not been well defined.

Although previous studies have found certain risk factors for synchronous esophageal neoplasia in H&N cancer patients [12, 13], there was no prospective study using routine application of NBI-ME and Lugol’s chromoendoscopy screening with standardized pathological classification. The aim of this prospective study was to determine the prevalence and risk factors for synchronous esophageal neoplasia and the impact of routine IEE screening on the decision making for the management of newly diagnosed H&N cancer patients.

The demographic characteristics of the study population and risk assessment by univariate logistic regression analysis of these factors are shown in Table 1. A total of 122 male and 7 female head and neck cancer patients were enrolled. Sixty defined suspicious esophageal lesions were biopsied from 53 (41.1%) patients after IEE screening and 30 (23.3%) patients had the presence of dysplastic or cancerous lesions by histopathological examination.

Compared with H&N cancer patients without synchronous esophageal neoplasia, the mean age and sex ratio of those with synchronous esophageal neoplasia were not significantly different. However, the BMI is lower in the group with synchronous esophageal neoplasia, and every 1-kg/m² increment is associated with 0.86-fold lower risk (p value = 0.008) for esophageal neoplasia (Table 1). The frequency of patients with presence of synchronous esophageal neoplasia was higher in those with hypopharyngeal cancer (12/32, 37.5%) and laryngeal cancer (4/8, 50%) than those with oral cavity cancer (8/63, 12.7%) and oropharyngeal cancer (6/22, 27.3%) (Table 1). The advanced stages, including 7th AJCC TNM stage III and IV, were associated with increased risk (OR 2.30, 95% CI 0.98-5.42) for esophageal neoplasia, but not statistically significant (p value = 0.056). Regarding the three common carcinogen exposures, only drinking alcohol was associated with a higher risk (OR 4.10, 95% CI 1.16-14.56, p value = 0.029) of esophageal neoplasia with a stepwise dose–response relationship (p value for trend = 0.009). Betel quid chewing was found to be associated with a lower risk (OR 0.37, 95% CI 0.16-0.84, p value = 0.018) of synchronous esophageal lesions. Compared with the proportion of betel quid chewers in oral cavity cancer subgroup, there were less betel quid chewers in the hypopharyngeal (48.4% vs. 68.3%, p = 0.03) and laryngeal cancer (25.0% vs. 68.3%, p = 0.01) subgroups (not shown in Table 1). The number of concomitant carcinogens used was not associated with an increased risk for esophageal neoplasia.

In the multivariate logistic regression model (Table 2), age, gender, cigarette smoking, betel quid chewing and location of index H&N tumor were not associated with the risk for synchronous esophageal neoplasia. However, status regarding the drinking of alcohol (OR 5.90, 95% CI 1.23-26.44, p = 0.020), lower BMI (every 1-kg/m² increment with OR 0.87, 95% CI 0.76-0.99, p = 0.036), and advanced stages (stage III&IV v.s. I&II) of index H&N cancers (OR 2.98, 95% CI 1.11-7.99, p = 0.030) were associated with higher risk for esophageal neoplasia.

The characteristics of the esophageal lesions detected by IEE screening are summarized in Table 3. 15%, 43.3% and 41.7% of the lesions were found at the upper, middle and lower third of the esophagus, respectively. Among them, 23.3% and 20.0% were high-grade intraepithelial neoplasia (HGIN) and invasive submucosal carcinoma, respectively. Three esophageal lesions were Barrett’s esophagus presented with intestinal metaplasia. To detect low-grade intraepithelial neoplasia (LGIN), HGIN and invasive carcinoma, the diagnostic performance of NBI system with ME examination which had the sensitivity, specificity, and accuracy of 97.3%, 94.1% and 96.3%, respectively, was better than those of WLI system and LC (Table 3). Seven patients (13.2%) had multifocal esophageal lesions.

The treatment strategy had been modified in a total of 20 (15.5%) H&N cancer patients after the IEE examination. The number needed to screen for synchronous esophageal neoplasia to have a modified treatment strategy was 6.45(1/15.5%, 95% CI = 4.60-10.90). The characteristics of these patients are shown in Table 4. Among them, 4 patients had oral cavity cancer, 4 oropharyngeal cancer, 9 hypopharyngeal cancer, and 3 laryngeal cancer. For the staging of the head and neck cancer, there were 4, 4, 4 and 8 patients with stage I, II, III and IV, respectively. The esophageal lesions of these patients were four LGINs, four HGINs, and twelve invasive carcinomas (3, 5, 1 and 3 patients with stage IA, IB, IIA, and IIIA, respectively). Adding esophageal neoplasia into consideration for treatment planning, five patients received extended radiation field involving the esophageal lesions, six patients received esophagectomy, and nine patients were suggested for endoscopic treatment of the early esophageal neoplasia, including radiofrequency ablation, endoscopic mucosal resection or submucosal dissection (Figure 1E and 1F).

The development of second primary malignancy of esophagus, H&N region or lung in UADT cancer patients is high [4, 5, 7, 22, 23]. A nationwide study in Taiwan, where the prevalence of drinking alcohol, cigarette smoking and betel quid chewing is high, has shown that the standardized incidence ratio (SIR) of esophageal cancer in patients with head and neck cancer is significantly increased [4]. Supported by the concept of “field carcinogenesis” [2], cancer may develop synchronously or metachronously in the UADT. Moreover, the survival rate of those with second primary tumor, especially with esophageal cancer, is significantly lower (5-year survival rate only 6%) than those without second primary malignancy [7, 24]. Therefore, it is important to perform surveillance of the esophagus in H&N cancer patients to improve overall outcome. In this study, we have found that drinking alcohol, lower BMI, and advanced stages of the index H&N cancer were associated with a higher risk for synchronous esophageal neoplasia, and the endoscopy examination by the NBI system with ME may be the best screening modality available.

Carcinogen consumption largely contributes to the development of the UADT malignancy, including the cancer of H&N region and the esophagus [1]. Among the common psychoactive substances, alcohol, tobacco and betel nut are well-known carcinogens for the squamous cell carcinoma of H&N region and the esophagus [1, 11]. In our study, we have found that only drinking alcohol was the independent risk factor for synchronous esophageal neoplasia by univariate and multivariate analyses (Table 1 and 2). Cigarette smoking and betel quid chewing were not associated with higher risk for synchronous second primary tumor of the esophagus in multivariate logistic regression model. Besides, higher cumulative doses of alcohol lead to a higher risk for esophageal lesions (Table 1). Acetaldehyde, the intermediate metabolite of the ethanol, is mainly responsible for the systemic carcinogenic effect of alcohol [25, 26]. The systemic effect of cigarette smoking and betel quit chewing is not as conspicuous as ethanol; by contrast, the inhaled tobacco smoke and topical contact of arecoline in betel nut contributes to the cancer of the respiratory tract, oral cavity and pharynx predominantly, rather than the esophagus [7, 25‐27]. This hypothesis could explain the reason why drinking alcohol, but not cigarette smoking or betel quid chewing, is associated with a higher risk for esophageal neoplasia in H&N cancer patients. In our study, betel quid chewing was found with lower risk for synchronous esophageal neoplasia in the univariate analysis (Table 1), but without statistical significance by multivariate logistic regression model (Table 2). Because in the univariate analysis, the hypopharyngeal and laryngeal cancers were associated with higher risk of esophageal neoplasia than oral cavity cancer, the interesting finding may result from lower proportion of betel quid chewers in the hypopharyngeal and laryngeal cancer patients than those in oral cavity cancer patients.

Regarding the location of H&N cancers, some researchers have found the tendency of esophageal cancer in patients with oropharyngeal and hypopharyngeal cancers, and the tendency of lung cancer in patients with laryngeal cancer [7, 13]. Although the results were based on the hypothesis of the cancerization in respiratory and digestive axes [22], other investigation did not report the similar result [4]. In a nationwide database study [4], the incidence of second primary cancer is higher for esophageal cancer (SIR 8.71, 95% CI 7.55-10.01) than for lung cancer (SIR 1.56, 95% CI1.34-1.80) in laryngeal cancer patients. In our study, the fact that higher proportion of betel quid chewing, which was negatively associated with risk of second primary esophageal neoplasm (OR 0.37, 95% CI 0.16-0.84) for H&N cancer patients, in the oral cancer group than the hypopharyngeal and laryngeal cancer patients might be the reason why the esophageal neoplasia less coexisted with oral cancer patients by univariate analysis, and the locations of index H&N cancer were not an independent risk factor for synchronous esophageal neoplasia in multivariate analysis (Table 1 and 2). Moreover, cancers of hypopharynx and larynx are more closely associated with alcohol consumption than oral cancers [28], and alcohol was highly associated with second primary neoplasia of esophagus in H&N cancer patients by both univariate and multivariate analyses in our result. The underlying pathogenesis relationship between esophageal cancer and different location of H&N cancer is still questionable.

Being overweight is a well-established risk factor for several types of cancers, except for UADT cancer [29]. In a collaborated cohort study conducted in the Asia-Pacific region, the cancer risk and mortality rate were all higher in those with excess bodyweight gain among 424,519 participants, excluding lung and UADT cancer [29]. Another multicenter case–control study conducted in European countries has also revealed the inverse association of BMI gain (BMI changes ≧5%) with UADT cancer risk (OR 0.74, 95% CI 0.62-0.89) [30]. Similar results have been reported in a population-based case–control study [31]. BMI gain of more than 25% was associated with a lower risk for lung cancer (OR 0.53, 95% CI 0.33-0.84) and UADT cancers (OR 0.44, 95% CI 0.27-0.71), especially in cigarette smokers and alcohol drinkers [31]. In our study, we have found that every 1-kg/m² increment in BMI was inversely associated with risk for synchronous esophageal neoplasia (OR 0.87, 95% CI 0.76-0.99) in H&N cancer patients (Table 2). Advanced stage of index H&N cancer was also an independent risk factor for esophageal lesions. Because UADT cancer related symptoms, such as odynophagia, dysphagia or oral pain, may be accompanied with decreased amount of oral intake, a lower BMI is usually observed in UADT cancer patients. However, in our study population, both lower BMI and advanced stage of index H&N cancer were independent risk factors, and most of the esophageal neoplasia detected were early superficial lesions without obstruction signs. This inverse association may indicate the body weight-related tumor biological pathway in cancerization of both H&N and esophageal cancers.

IEE is a useful screening tool for precancerous or cancerous gastrointestinal neoplasia. To detect esophageal neoplasia in high risk population, Lugol’s chromoendoscopy which is one of the dye-based IEE techniques has the sensitivity and specificity about 80.0-96.0% and 63.0-72.2%, respectively [11, 32‐34]. However, many side effects, such as mucosal irritation, retrosternal chest pain or burning sensation, laryngospasm or bronchospasm, and even cardiac arrest, have been reported [18, 35]. These uncomfortable adverse effects and the low specificity hinder the Lugol’s chromoendoscopy from being the screening tool of esophageal neoplasia. On the contrary, NBI system which is one of the equipment-based IEE techniques has a higher sensitivity (88.9%) and specificity (97.2%) for detecting esophageal neoplasia in high risk populations [34]. By pushing a button on the handgrip of the endoscope, esophageal neoplasia with hypervascularity could be easily well-delineated without uncomfortable side effects [9, 17]. Moreover, the NBI system is not only applied to the detection of early esophageal neoplasia (accuracy 88.9%), but also superficial cancers of H&N region (accuracy 86.7%) as well [14, 15]. With the combination of magnifying endoscopy, the margin and invasiveness of the neoplasia can be predicted accurately [9, 16, 33, 34]. In our study, about 97.3 ~100% of esophageal neoplasia were detected by NBI system with ME and Lugol’s chromoendoscopy. Only 51.4% of the esophageal neoplasia were detected by the traditional WLI system. To screen LGIN, HGIN and invasive carcinoma, NBI systems with ME were shown to demonstrate a better diagnostic performance than both WLI system and Lugol’s chromoendoscopy (Table 3). Thus, routine IEE surveillance of the esophagus, especially using NBI system with ME avoiding the unpleasant side effects from Lugol’s solution, should be part of initial staging of H&N cancer patients.

The presence of synchronous esophageal neoplasia in H&N cancer patients is not uncommon. Routine IEE surveillance of the esophagus is very important to initial treatment strategy of newly diagnosed H&N cancer patients, especially for those with alcohol drinking habit, lower BMI, and advanced stage of index H&N tumor. We recommend the routine surveillance of esophagus by performing NBI system with ME examination in the initial staging workup of newly diagnosed N&N cancer patients with risk factors identified by this study, and the treatment of synchronous neoplasia should be taken into consideration.