Introduction

Cancers of the upper aerodigestive tract, such as the oral cavity, pharynx, nasopharynx, larynx, sinuses, and salivary glands, collectively constitute head and neck cancer (HNC)1,2. Head and neck cancer ranked among the dominant cancers globally, with 1.43 million newly reported cases and 0.51 million deaths in 2020. HNCs constituted 7.4% of the total 19.3 million new cancer cases worldwide, and caused 0.51 million deaths, for 5.1% of the total 10.0 million cancer deaths3. Over 90% of HNCs are squamous cell carcinomas3. The risk factors for HNC include exposure to carcinogens such as tobacco, alcohol consumption, and betel quid chewing, as well as male gender. Other risk factors include infectious agents such as human papillomaviruses and Epstein–Barr virus, poor oral hygiene, history of esophageal cancer, consuming hot beverages, occupational exposures (such as metal smelting and textile production), and consumption of preserved/processed foods containing nitrosamines4,5,6,7. However, only a fraction of HNC cases are accounted for by these known risk factors1.

The interaction between viral infection, the human immune system, and cancer biology is complex. Viruses like EBV, HPV, HBV, and HCV are implicated in 10–20% of human cancers. These viruses contribute to cancer through mechanisms like genomic instability and immune evasion. The immune system surveils and eliminates cancerous cells, but viruses can disrupt this process. This disruption can lead to unchecked cancer cell proliferation8. Oncogenic viruses have been valuable in cancer research, helping identify key molecular pathways involved in tumorigenesis and immune evasion, and guiding the development of targeted therapies and immunotherapies9.

Herpes zoster (HZ), also known as shingles, has been proposed as a likely risk factor for subsequent cancer, although the evidence remains controversial. HZ is caused by reactivation of latent infection with varicella-zoster virus (VZV), manifesting in the form of intermittent excruciating pain and discomfort along the affected unilateral dermatomes, followed by a maculopapular and vesicular rash in the area10. HZ can occur at any age, but it predominantly impacts older adults, with the incidence rising with advancing age11,12. The increase in HZ occurrence is notable from the age group of 50–60 years, reaching a prevalence of 11 cases per 1000 population aged 80 years and older13. VZV is thought to be restrained in a latent form by VZV-specific cell-mediated immunity11. Immunocompromised hosts face an elevated risk of developing HZ14. In a large, prospective follow-up cohort study in Australia by Qian et al. from 2006 to 2016 involving over 240,000 adults, patients with hematologic or solid cancer showed significantly elevated incidence of HZ prior to and after the cancer diagnosis compared to those without cancer, with adjusted hazard ratios of 3.74 and 1.30, respectively15. Anti-cancer therapy, such as radiation therapy, further increased the risk of HZ. The occurrence of HZ following radiation therapy was notably higher in the designated radiation field16. A nationwide population-based study also demonstrated that patients with HNC receiving radiation therapy are at an increased risk of herpes zoster17.

The exact connection between HZ and the risk of subsequent cancer incidence is not firmly established. Several epidemiologic studies have reported a possible link between HZ and subsequent cancer diagnoses; no study has specifically examined the association with subsequent HNC, which is a substantial contributor to the global cancer morbidity and mortality. Identifying factors with a potential link to HNC could enhance disease prevention, cancer downstaging, and assessing cancer prognosis. Further, exploring site-specific HNC classifications may facilitate better identification of cancer risk factors and the intricate pathways underlying cancer development.

The aim of this population-based case-control study was to examine HNC patients risk of having HZ in the five years preceding their HNC diagnosis compared to persons without HNC. Additionally, we assess site-classified HNCs to evaluate risk differences by cancer site.

Materials and methods

Database

We conducted a case-control study utilizing information from Taiwan’s Longitudinal Health Insurance Database 2010 (LHID2010). The LHID2010 has been extensively utilized by various researchers and scholars to investigate the epidemiology of diseases and treatments. The study was authorized by the Research Ethics Committee of National Taiwan University (202012EM075) and is in compliance with the Declaration of Helsinki. As we used deidentified administrative data provided by the LHID administration, informed consent was waived by the Research Ethics Committee of National Taiwan University.

Identification of study patients

We used a case-control methodology. The case group consisted of all patients in the LHID2010 aged ≥ 20 years old who received a first-time diagnosis of HNC, 9,191 persons. HNCs included all cancers defined as HNC, identified using the respective International Classification of Diseases (ICD) codes, oral cavity cancer (ICD-9-CM code 140, 141 (except 141.0; 141.6), 143, 144, 145.2; 145.3; 145.5, or 145 (except 145.2; 145.3; 145.5), ICD-10-CM codes C00, C02 except C02.4, C03, C04, C05, or C06), oropharynx (ICD-9-CM 141.0, 141.6 or 146, ICD-10-CM C01, C02.4, C09 or C10), larynx (ICD-9-CM 161, ICD-10-CM C32), hypopharynx (ICD-9-CM 148, ICD-10-CM C12 or C13), nasopharynx (ICD-9-CM 147, ICD-10-CM C11), sino-nasal (ICD-9-CM 160 (except 160.1), ICD-10-CM C31, C30.0), and salivary gland (ICD-9-CM 142, ICD-10-CM code C07, C08), during the period of 2011 to 2019. We assigned the date of their first-time HNC diagnosis as the index date.

We further employed propensity-score-matching method to retrieve the controls from the remaining beneficiaries aged ≥ 20 years old from the LHID2010. Beneficiaries with any claim before or after the selected cohort period showing a diagnosis of HNC were excluded. Using logistic regression modeling, propensity scores were calculated for all identified 9,191 patients with HNC and the remaining beneficiaries based on age, sex, monthly income, geographic location, urbanization level of the patient’s residence (5 levels, 1 most urbanized, 5 least urbanized), the presence of diabetes, hypertension, hyperlipidemia, and human papillomavirus (HPV) infection, tobacco use disorder and alcohol abuse/alcohol dependence syndrome. Finally, each sampled patient with HNC was matched with four control beneficiaries without HNC using the nearest neighbor random matching algorithm with caliper adjustment, using an a priori calipers value of +/-0.02. For cases, we assigned the index date as the date of first receipt of the HNC cancer diagnosis. For controls, the index date was assigned as the date when the corresponding control beneficiaries had their first ambulatory care visit during the index year of the matched case. The final study sample consisted of 9,191 patients with HNC and 36,764 controls without HNC.

Measures of exposure

We identified HZ based on ICD-9-CM code 0.53 or ICD-10-CM code B02 in any claim within a 5-year period prior to the index date.

Statistical analysis

We used the SAS system to conduct all statistical analyses. The differences in sociodemographic characteristics and the presence of medical comorbidities of patients with HNC and controls were studied using chi-square tests and t-tests. Multiple logistic regression modeling was utilized to evaluate the co-variate adjusted association between HNC and prior HZ taking the variables of sociodemographic characteristics and medical comorbidities (diabetes mellitus, hypertension or high blood pressure condition as well as hyperlipidemia or elevated levels of lipids in blood plasma, tobacco use disorder, HPV infection, and alcohol abuse/alcohol dependence syndrome) into consideration. We used odds ratios (OR) and their corresponding 95% confidence intervals (CIs) to present the difference in likelihoods of HZ occurrence among HNC patients compared to controls. We used a two-sided p-value < 0.05 as statistically significant.

Results

The sociodemographic characteristics and medical comorbidities among the sample patients are presented in Table 1. The HNC patient sample was dominated by men, 80.1%, and correspondingly the control group. The mean ages of the group with HNC and controls were similar, 60.4 ± 13.3 years and 60.5 ± 13.3, respectively, p = 0.493. The two groups were also similar on all the remaining sociodemographic and comorbidity variables, monthly income (p = 0.346), sex (p > 0.999), geographic location (p = 0.145), urbanization level (p = 0.758), diabetes (29.9% in both groups, p > 0.999), hyperlipidemia (38.7% in both groups, p > 0.999), hypertension (47.7% in both groups, p > 0.999), HPV infections (6.4% in both groups, p > 0.999), tobacco use disorder (10.3% in both groups, p > 0.999) and alcohol abuse/alcohol dependence syndrome (3.9% in both groups, p > 0.999).

Table 1 Demographic characteristics of patients aged 20 years or older with head and neck cancer (HNC) and propensity score-matched control patients.
Full size table

Table 2 presents the rate of HZ occurrence among the total patients and the patients with HNC and controls. Among the full sample, 4.6% (2,122 patients) had a prior diagnosis of HZ, 7.9% among the HNC group and 3.8% among the control group (p < 0.001). Within HNC site classifications, significant differences between the cancer group and control group were observed in respect of oral cavity cancer (7.2% vs. 3.8%, respectively, p < 0.001), oropharynx (7.2% vs. 3.8%, p < 0.001), larynx (9.6% vs. 3.8%, p < 0.001), hypopharynx (6.2% vs. 3.8%, p < 0.001), nasopharynx (9.1% vs. 3.8%, p < 0.001), and salivary glands (11.6% vs. 3.8%, p < 0.001). We failed to observe a statistically significant difference between the sinonasal cancer group and controls (6.4% vs. 3.8%, p = 0.054).

Table 2 Prevalence rates of herpes zoster in the prior five years among patients with head and neck cancer (HNC)s vs. controls.
Full size table

Table 3 shows the OR for HZ of HNC patients versus controls, for all HNCs and HNC site classifications. After adjusting for sex, monthly income, age, geographic location, urbanization level, hypertension, HPV infection, hyperlipidemia, diabetes, tobacco use disorder and alcohol dependence syndrome, the OR for prior HZ was 2.198 (95% CI = 2.001 ~ 2.415, p < 0.001). Adjusted ORs were also significantly higher for oral cavity cancer 2.048 (95% CI = 1.798 ~ 2.333, p < 0.001), oropharynx 2.111 (95% CI: 1.616 ~ 2.758, p < 0.001), larynx 2.154 (95% CI = 1.654 ~ 2.805, p < 0.001), hypopharynx 1.744 (95% CI = 1.298 ~ 2.344, p < 0.001), nasopharynx 2.714 (2.311 ~ 3.186, p < 0.001), and salivary gland 2.993 (95% CI = 2.031 ~ 4.411, p < 0.001).

Table 3 Adjusted odds ratios of prior herpes zoster among patients with head and neck cancer (HNC) vs.controls.
Full size table

Table 4 estimate the ORs of prior HZ among patients with HNC vs. controls by the infection timing of prior HZ. We found that the OR was similar regardless of the HZ infection time (year).

Table 4 Adjusted odds ratios of prior herpes zoster among patients with head and neck cancer (HNC) vs. controls by the occurrence timing of prior herpes zoster.
Full size table

Discussion

Our study may be the first nationwide population-based study to report on HZ as a potential risk factor or prior harbinger for a subsequent diagnosis of head and neck cancer. We found about twice the odds of HZ in the prior 5 years among persons diagnosed with HNC. Of the overall sample, 4.6% had suffered from HZ in the prior five years, being 7.9% among patients with HNC, compared to 3.8% among controls. When categorized by the anatomic site, patients with cancer at all locations showed a higher rate of HZ than control patients, with the exception of sinonasal cancer, After adjustment for the covariates, the odds ratio for the HNC patients was 2.198 (95% CI = 2.001–2.415, p < 0.001), with statistically significantly increased odds observed for most subsite classifications (oral cavity, oropharynx, larynx, hypopharynx, nasopharynx, and salivary gland), and sinonasal cancer failing to attain statistical significant (p = 0.054).

Studies regarding whether HZ or not is associated with increased cancer likelihood subsequently have been inconsistent in their findings. Ragozzino and colleagues published the first population-based cohort study in 1982 assessing the risk of any type of cancer following a diagnosis of HZ18. They followed 590 patients in Rochester, Minnesota, for 9389 person-years after their diagnosis of herpes zoster, and reported a relative risk of 1.1, 95% CI 0.9–1.3. The study was limited by small sample size. However, they observed a slight elevation in the relative risks of colon and bladder tumors in women. Sørensen and colleagues reported on the risk of all types of cancer in a population-based study, using follow-up data on 10,588 hospitalized patients with HZ between 1977 and 1996 to investigate their subsequent risk of cancer19. In the first year, the relative risk (RR) for all types of cancer was 1.3 (95% CI 1.1–1.5), with a higher relative risk for hematologic cancer, 3.4 (95% CI 2.3–4.9). Buntinx and colleagues investigated the risk of a subsequent cancer diagnosis over 311,000 patient-years (1-year follow-up) of a general practitioner patient population in Belgium from 1994 to 2000 (most patients being women aged over 65 years). Among women aged 65 years or older, they identified a significantly increased risk of cancer among those with a HZ diagnosis. Wang and colleagues conducted a population-based retrospective study in 2012, examining 35,871 newly diagnosed patients with HZ in Taiwan from 2000 to 2008. They observed no association with subsequent cancer compared to the general population20. Their analysis did not account for confounding factors such as socioeconomic status and comorbidities. One systematic review and meta-analysis conducted by Schmidt et al. in 2016 reported an association between HZ and occult cancer. The pooled relative risk for any cancer was 1.42 (95% CI: 1.18, 1.71) across all follow-up durations, and 1.83 (95% CI: 1.17, 2.87) during Year 1 of follow-up19. A similar result was reported in a Taiwanese nationwide cohort study in 2020. The risk of subsequent cancer was modestly increased for one year or more after a diagnosis of HZ20. Multiple studies have explored the risk of specific cancers among patients with HZ. Compared to individuals without cancer, Qian et al. found that the risk of prior HZ diagnosis was increased among persons with hematological cancer (adjusted hazard ratio for 1–2 years prior, 2.01 [95% CI, 1.31–3.09]), but not for persons with solid tissue cancers15. Other studies have shown an association between HZ and subsequent lymphoid (or hematologic) neoplasms10,20,21.

Only one study had investigated the association between HZ and subsequent cancer in the head and neck region. In 2011, Ho et al. showed that individuals with herpes zoster ophthalmicus (HZO) had a 9.25-fold higher risk (95% CI, 5.51–15.55) of a subsequent cancer diagnosis compared to a propensity score-matched control group. They concluded that HZO could serve as a marker for increased cancer risk of cancer22. In their study, however, out of 658 cases of HZO, only 32 patients (4.86%) developed subsequent cancer during the study follow-up period of one year. Limited sample size and limited follow-up precluded additional exploration into specific cancer sites. Our study may be the first in the literature to demonstrate that HZ infection represents a higher likelihood of a subsequent diagnosis of head and neck cancer. Our findings held up within cancer sites, except for sinonasal cancer which could be due to statistical power limitations (p = 0.054) Further studies are needed for a more detailed understanding of the risk patterns by site.

Several potential mechanisms for the HZ-cancer link have been addressed in the literature. First, both HZ and HNC are associated with diminished cell-mediated immunity. The varicella zoster virus remains dormant until the cell-mediated immune system is weakened, allowing the virus to proliferate19. The reactivation of VZV further triggers immunologic mechanisms that weaken immune surveillance for cancer cells, enabling tumor escape22,23,24,25. Second, host immunity dysfunction may lead to the onset of either HZ or malignancy, with the possibility of either condition manifesting first22,25. Third, there is also a suggestion that HZ directly contributes to carcinogenesis26. Viruses can introduce alterations in the genome through incorporation of viral genes into the host genome21. The viral infection by itself can induce a state of chronic inflammation27. Both of these changes disturb normal cellular functions and promote cell proliferation and malignant change28,29.

There are some implications of our findings. Firstly, healthcare professionals should be alert to actively investigate patients presenting with HZ for cancer especially if they have symptoms of oncological significance or risk factors for cancer12,30,31. Secondly, additional evaluation of the patient’s immune response to zoster infection may be in order. Thirdly, administration of the HZ vaccine to immunocompromised individuals, particularly those with cell-mediated immune deficiency, may mitigate the risk of HNC by preventing HZ32.

Our study has three strengths. First, patients with HNC and those in the control group were obtained from the LHID2010 claims dataset. The utilization of this population-representative sample minimizes concerns of recall bias, selection bias, non-response bias (inherent in survey-based studies) and surveillance bias, as well as bias due to loss to follow-up. Second, the large, 99.7% Taiwanese population coverage by the National Health Insurance Program provided us with a substantial enough sample to investigate the association between HZ and HNC, and particularly sub-site cancers within HNC. Many prior studies exploring HZ and cancer have adjusted for a limited set of confounders, with only a few accountings for comorbidities. Our study carefully accounted for many relevant sociodemographic and disease covariates not examined in previous research (e.g., monthly income, geographic location, and urbanization level, other sociodemographics and comorbidities). Use of propensity scores to match HZ cases with comparison patients to achieve an appropriately comparable group improves the validity of the findings by maximally overcoming confounding. Finally, the identification of cases of malignant head and neck tumors in our study is both valid and definitive because the National Health Insurance Program administration mandates biopsy and histological confirmation to declare a cancer diagnosis for reimbursement and subsequent referral for treatment.

Despite the above strengths, some limitations should be noted. First, the HZ patients identified in the LHID2010 claims dataset represents only patients who had sought treatment for HZ and HNC. This concern may be largely mitigated in respect of HZ, given the intense cutaneous symptoms—burning pain, itching, hyperesthesia, and paresthesia over protracted periods of time, several days to weeks or months, which is likely to constrain them to seek medical attention. Given the comprehensive benefits covered by the National Health Insurance Program which includes coverage of cancer treatment, it is unlikely that persons with HNC will refrain from seeking treatment. Second, the study was unable to further explore the relationships with different anatomic locations of HZ, particularly facial and other anatomical areas. Finally the data on important patient characteristics and lifestyle factors (e.g., tobacco and alcohol consumption, betel quid chewing, poor oral hygiene) associated with HNC and HZ were unavailable in the claims data used in our study.

It should be noted that the study lacks verification methods to ensure the herpes zoster infection, as our study employed a blend of clinical and molecular diagnosis codes, incorporating including the PCR test, serum antigen, and antibody tests to detect Herpes Zoster infection cases. This methodology was selected due to non-obligatory use of the laboratory test in our healthcare system for managing associated illnesses. The research design aimed to encompass as many instances of Herpes Zoster as feasible while delving into the potential association with head and neck cancer, thereby resulting in diagnostic heterogeneity regarding Herpes Zoster detection. The identification of human Herpes Zoster virus infections was reliant on medical documentation, which could be impacted by the frequency of medical visits and testing procedures, consequently resulting in detection prejudice. Even though the laboratory examination is not obligatory for managing related illnesses in our healthcare network, it would be preferable if all Herpes Zoster patients were confirmed through PCR testing, with subtypes specified for further analysis.

Our study presents an opening by showing that within a 5-year period prior to an HNC diagnosis, HZ infection may serve as a potential risk factor for cancer or at least a marker of occult cancer. Moving beyond the epidemiologic evidence presented in this population-based dataset, it is necessary to verify the findings and confirm the direction of association, and potentially a causal link between the two entities. Future studies are warranted to clarify the temporal association between HZ and HNC. While the histological data for cases of HNC in our study based on health insurance is currently unavailable, it is imperative that future investigations consider conducting subgroup analyses to determine if there exist any disparities between cases of squamous cell carcinoma (SCC) and those of non-SCC. In addition, further studies are suggested to elucidate the biomedical pathways connecting HZ to HNC, and further, the potential for the HZ vaccine to influence the causal linkages that may be involved between HZ and cancer.