Aetiological heterogeneity of head and neck squamous cell carcinomas: the role of human papillomavirus infections, smoking and alcohol

Abstract

Tobacco and alcohol consumption are the main risk factors for head and neck squamous cell carcinoma (HNSCC). In addition, human papillomavirus (HPV) infection plays a causal role in oropharyngeal cancer (OPC), a subset of HNSCC. We assessed the independent effects of tobacco, alcohol and HPV infection on OPC risk in the head and neck cancer (HeNCe) Life study, a hospital-based case-control study of HNSCC with frequency-matched controls by age and sex from four Montreal hospitals. Interviewers collected information on socio-demographic and behavioural factors. We tested exfoliated oral cells for HPV DNA by polymerase chain reaction (PCR). We included only OPC cases (n = 188) and controls (n = 427) without missing values for HPV, smoking or alcohol. We examined associations by estimating odds ratios (ORs) and corresponding 95% confidence intervals (CI) using unconditional logistic regression. Smoking (OR = 1.90, 95% CI: 1.04–3.45) and alcohol (OR = 2.74, 95% CI: 1.45–5.15) were associated with an increased risk of OPC independent of HPV status. Positivity for HPV 16 among heavy smokers and heavy alcohol users was associated with a 30.4-fold (95% CI: 8.94–103.26) and 18.6-fold (95% CI: 5.75–60.13) elevation in risk of OPC relative to participants who were HPV negative, respectively. Moreover, the combined effect of heavy smoking and alcohol comsumption with HPV 16 infection substantially increased OPC risk (OR = 48.76, 95% CI: 15.83–150.17) and (OR = 50.60, 95% CI: 15.96–160.40), respectively. Our results support the independent roles of smoking, alcohol and HPV infection in OPC risk and a possible combined effect. Efforts should be made to tackle these major risk factors simultaneously.

Introduction

Tobacco smoking and alcohol drinking are important risk factors for head and neck squamous cell carcinoma (HNSCC) (1–3). The overall HNSCC incidence has decreased in some populations in recent decades (3), paralleling a decrease in tobacco use. During the same period, incidence rates of oropharygeal cancers (OPC) have increased in developed coutries (4), probably a reflection of the role played by sexually transmitted human papillomaviruses (HPV) in this subset of cancers (4–9).

HPV-related HNSCC are biologically (10,11), clinically (4) and histologically (4) distinct. Cases of HPV-related HNSCC are generally younger (9,12–14) and more likely to be males (9,13). They usually present with late disease stages (15), but their tumours are smaller and prognosis and survival are better (4,10,13,14). Moreover, these individuals seem to consume less tobacco and alcohol (4,12,14,16,17) compared with HPV-negative HNSCC cases. Also, they are less likely to harbor p53 mutations (4,10,16,18) associated with smoking (16).

Two aetiological pathways for HNSCC have been suggested through smoking and alcohol consumption and alternatively through HPV (4,12,18,19). Limited conflicting results have emerged on interactions between traditional risk factors for HNSCC and HPV (15,17,19–23). While several studies observed lower levels of smoking in HPV-related HNSCC (4,12,16,17,24), others did not (21). This led some researchers to suggest that there is no synergy between HPV and smoking in HNSCC aetiology (17,19). Indeed, the combined effect of HPV 16 positivity and tobacco smoking/pan chewing in the International Agency for Research on Cancer (IARC) multicenter international study was lower than would be expected in a multiplicative model (17). Furthermore, Chaturvedi et al. 2016 pointed out that interaction studies from case series and case-control studies might lead to biased conclusions on the burden of HPV-related HNSCC among smokers and non-smokers (25). The authors further showed that although the risk of HNSCC attributable to HPV infection is higher among non-smokers compared with smokers, the burden of these cancers at population level is considerably higher among smokers (25).

Amid these conflicting results and opposite time trends for the incidence of HPV-positive and HPV-negative OPC, a question of high public health significance is to study the interplay among these risk factors on the risk of OPC. We addressed this question by estimating the risk associated with tobacco smoking and alcohol consumption independent of HPV infection and vice versa among a sample of Canadians.

Materials and methods

Data were collected at the Canadian site of an international hospital-based case-control study, the head and neck cancer (HeNCe) Life study which investigates the aetiology of HNSCC using a life course framework.

Study subjects and data collection

Between September 2005 and November 2013, consecutive incident histologically confirmed HNSCC cases were recruited from four major referral hospitals in Montreal (Notre-Dame Hospital, Jewish General Hospital, Royal Victoria Hospital and Montreal General Hospital). For participants to be eligible, they had to have no history of cancer; have no mental or immune suppression disorders; be born in Canada; be at least 18 years of age and live within 50 km of the hospital.

Cancers of the oral cavity, oropharynx and larynx were included in the study. They were identified using the International Classification of Disease version 10 codes (26): oral (C02-C06), oropharynx (C01, C02.4, C05.01, C05.2, C09, C10, C13) and larynx (C32). Other ill-defined sites (C14) were classified according to their description in the medical records. Lip, salivary gland and nasopharyngeal cancers were excluded because of their different aetiology. In the present analysis, only OPC cases were included.

Controls, frequency-matched to cases by sex and 5-year age group were recruited from outpatient clinics of the same hospitals. They were selected from a list of non-chronic diseases and those not related to tobacco or alcohol consumption. Ethical approval was obtained from the Research Ethics Office at McGill University, the Institut National de la Recherche Scientifique (INRS), and all participating hospitals. All subjects provided written informed consent.

Face-to-face interviews were conducted using a questionnaire and life-grid technique (27) to collect information on socio-demographic characteristics and various exposure domains, including a detailed history of tobacco smoking and alcohol drinking.

Oral cell sample collection, HPV detection and genotyping

Epithelial cells were collected from various sites in the mouth, including the tumour site for cases, using an OralCDx® brush and mouthwash. Samples were placed into a solution which was centrifuged at 1000×g for 10 min, then the pellet was transferred with a small amount of the supernatant to a 1.5 ml centrifuge tube (28). DNA was extracted with MasterPure™ (Epicentre, Madison, WI) (28,29). Purified DNA in 10 mM Tris–EDTA buffer, pH 7.5, was then kept at −70°C until testing for HPV DNA using a polymerase chain reaction (PCR) assay. First, the presence of HPV DNA was detected by amplifying 10 μl of extracted DNA with PGMY09-PGMY11 primers (30,31) and by testing amplicons with a generic probe mix (32). All HPV DNA samples were tested for β-globin by amplifying 10 μl of extracted DNA with PC04 and GH20 and detecting amplicons by agarose gel electrophoresis, to assess DNA integrity, absence of inhibitors, and presence of a sufficient quantity of cellular material for PCR analysis (33,34).

β-Globin-negative samples were considered inadequate for PCR analysis while positive samples were considered for HPV-genotyping. HPV amplicons generated with PGMY09-PGMY11 from samples positive with the generic assay for HPV were then genotyped with the linear array for 36 mucosal HPV genotypes including types 6, 11, 16, 18, 26, 31, 33, 34 (formerly known as 64), 35, 39, 40, 42, 44 (formerly known as 55), 45, 51, 52, 53, 54, 56, 58, 59, 61, 62, 66, 67, 68, 69, 70, 71, 72, 73, 81, 82, 83, 84 and 89 (35). HPV52 was detected in the linear array with a probe that also cross-reacts with types 33, 35 and 58. Samples reactive in the linear array with the cross-reactive probe for HPV52 and containing at least one of the cross-reactive types were further tested with a validated HPV52-specific real-time PCR assay (36). Negative, weak (10 HPV 18 DNA copies) and strong positive HPV controls were included in each amplification run. Precautions to avoid contamination were effective in our laboratory.

Statistical analyses

Participants who smoked tobacco products for at least 1 year were considered as ever smokers. They were asked to report their smoking history as periods defined by constant intensity of smoking (units per day), type of tobacco product (e.g. cigarette, cigar and pipes) and brand of product. For each period, the age at beginning and end of that period and intensity of smoking was collected and cross-referenced with participants’ other life events to increase accuracy of recall. Similarly, a detailed history of alcohol consumption was also collected based on type of beverage (e.g. wine, beer and hard liquor) and units drunk.

Lifetime exposure to tobacco smoking was computed as pack-years, where one pack-year is equivalent to smoking one pack of cigarettes per day for 1 year (37). For this purpose, cigars and pipes were converted to a standard cigarette scale based on their tobacco content (1 pack = 20 cigarettes = 4 cigars = 5 pipes) (2). Values for the pack-years of smoking variable were categorized based on tertiles of the distribution among smokers, while non-smokers were included as the reference category.

To compute the variable representing alcohol consumption, different types of alcoholic beverages were combined based on their concentration of ethanol (beer = 5%, wine and aperitif = 10% and hard liquor = 50%) (2). Lifetime alcohol consumption was computed as liters of ethanol consumed, then categorized into quartiles.

For HPV status, subjects were classified into four hierarchically constructed groups based on oropharyngeal carcinogenic potential: (i) HPV negative; (ii) infected exclusively with one or more low risk (LR) HPV types: HPV 6, 11, 26, 34, 40, 42, 44, 45, 52, 53, 54, 56, 58, 59, 61, 62, 66, 67, 68, 69, 70, 71, 72, 73, 81, 82, 83, 84, 89; (iii) (irrespective of harboring LR types) infected with high risk (HR) HPV other than HPV 16: HPV 18, 31, 33, 35, 39, 51; and (iv) infected with HPV 16, irrespective of other types (38). Infections with more than one HPV type were assigned to the category including the type with the highest risk.

To assess the effect of smoking, alcohol and HPV infection on OPC risk, we estimated odds ratios (ORs) and 95% confidence intervals (CI) using unconditional logistic regression, adjusting for age and sex and then further adjusting for years of education.

We assessed statistical interaction between HPV and each of smoking and alcohol drinking on OPC risk. Tobacco smoking and alcohol consumption were previously shown to have a non-linear association with risk of OPC, with consistently elevated risk at high exposure levels (39). Hence, we created binary variables for each of smoking and alcohol drinking and used the highest quartile among smokers (>37 pack-years) and the highest quartile among alcohol consumers (>369 L) as contrasts of interest to assess effect modification by HPV while maximizing statistical efficiency and precision when conducting stratified analyses by each of these factors.

Finally, to investigate the role of HPV in the association between smoking, alcohol and OPC, two models were compared. In the first model, the effects of HPV and smoking or alcohol on risk of OPC were considered to be independent of each other while in the second, we assumed that they modify each other’s effect. The latter model included product terms between the HPV infection variable and smoking or alcohol variables to account for the empirical value of the putative effect modification. A likelihood ratio test was performed to assess the statistical significance of the assumed effect modification between HPV and tobacco smoking and HPV and alcohol consumption. Models for tobacco smoking were adjusted for alcohol consumption and vice versa, in addition to age, sex and number of years of education.

Statistical analyses were conducted using Stata 12.1 (Stata Corp. LP, College Station, TX, USA).

Results

Of the 460 HNSCC cases and 458 controls recruited, only OPC cases and controls without missing data for the variables under study were included, constituting a total of 188 cases and 427 controls. The general participation rates in the HeNCe Life study were 47% and 54% of eligible cases and controls approached, respectively. Table 1 displays characteristics of the study sample. As expected, most cases were male and, by design, most controls were, as well. Cases reported higher levels of tobacco smoking and alcohol drinking. A much greater proportion of cases (48%) than controls (2%) was HPV 16 positive (Table 1).

Table 1.

Demographic and behavioural characteristics of OPC cases and controls, HeNCe Life study, Canadian site

Characteristic	Cases, n = 188		Controls, n = 427
	n (%)	Mean ± SD	n (%)	Mean ± SD
Age, years		59.91 (9.48)		61.1 (10.88)
Age
≤50 years	25 (13.30)		64 (14.99)
51–70 years	132 (70.21)		278 (65.11)
>70 years	31 (16.49)		85 (19.91)
Sex
Male	136 (72.34)		295 (69.09)
Female	52 (27.66)		132 (30.91)
Education, years		12.30 (3.73)		13.86 (4.37)
Smoking, pack-years		29.22 (35.48)		24.70 (38.79)
Smoking
Non-smokers	44 (23.40)		118 (27.63)
Light smokers (≤15 pack-years)	37 (19.68)		102 (23.89)
Moderate smokers (>15–37 pack-years)	41 (21.81)		103 (24.12)
Heavy smokers (>37 pack-years)	66 (35.11)		104 (24.36)
Alcohol consumption, ethanol (L)		569.49 (1187.39)		359.14 (812.92)
Ethanol
Up to 10 L	38 (20.21)		106 (24.82)
>10–124 L	40 (21.28)		107 (25.06)
>124–369 L	45 (23.94)		107 (25.06)
>369 L	65 (34.57)		107 (25.06)
HPV status
HPV negative	69 (36.70)		368 (86.18)
LR-HPV	13 (6.91)		34 (7.96)
HPV 18, 31, 33, 35, 39, 51	15 (7.98)		15 (3.51)
HPV 16	91 (48.40)		10 (2.34)

Table 2 presents the crude and adjusted ORs for the associations of smoking, drinking and HPV status with OPC. Heavy smoking (OR = 1.73, 95% CI: 1.08–2.77) and heavy alcohol consumption (OR = 1.70, 95% CI: 1.01–2.87) were associated with an increased risk of OPC, with a small attenuation of the magnitude of the associations or of their corresponding precision upon adjustment for age, sex, years of education and HPV status. We observed a 5-fold increase in risk with HR-HPV types other than HPV 16 (OR = 4.78, 95% CI: 2.14–10.65) and a 54-fold increase with HPV 16 infections (OR = 53.62, 95% CI: 25.74–111.70) adjusted for covariates including smoking and alcohol.

Table 2.

Adjusted ORs (and 95% CI) for the effects of smoking, alcohol drinking and HPV infection on OPC risk, HeNCe Life study, Canadian site

Variable	Adjusted only for age and sex	Adjusted for additional variablesa
Smoking
Non-smokers	1.00	1.00
Light smokers (≤15 pack-years)	0.93 (0.55–1.55)	1.08 (0.55–2.12)
Moderate smokers (>15–37 pack-years)	1.03 (0.62–1.71)	1.19 (0.61–2.33)
Heavy smokers (>37 pack-years)	1.73 (1.08–2.77)	1.62 (0.85–3.08)
Ethanol
Up to 10 L	1.00	1.00
>10–124	1.02 (0.60–1.72)	0.94 (0.48–1.84)
>124–369	1.17 (0.69–1.98)	1.09 (0.55–2.16)
>369 L	1.70 (1.01–2.87)	1.92 (0.97–3.80)
HPV status
HPV negative	1.00	1.00
LR-HPV	2.21 (1.1–4.46)	1.86 (0.90–3.82)
HPV 18, 31, 33, 35, 39, 51	5.77 (2.66–12.48)	4.78 (2.14–10.65)
HPV 16	51.41 (25.21–104.81)	53.62 (25.74–111.70)

^aAdjusted for age, sex, number of years of education (continuous), smoking (categorical) and ethanol consumption (L) (categorical).

To estimate the independent effect of smoking and alcohol on OPC risk, we stratified the subjects by HPV status (Table 3). Among the HPV-negative participants, the effects of heavy tobacco smoking and alcohol consumption on OPC risk were statistically evident, whereas the associations were considerably reduced or lost precision among those who were HPV positive.

Table 3.

Adjusted ORs (and 95% CI) for the associations between smoking, alcohol drinking and OPC, stratified by HPV status, HeNCe Life study, Canadian site

Characteristic	HPV negative	LR-HPV	HPV 18, 31, 33, 35, 39, 51	HPV 16
Smoking pack-years
Non- to moderate smokers (≤37 pack-years)	1.00	1.00	1.00	1.00
Heavy smokers (>37 pack-years)	1.90 (1.04–3.45)	2.79 (0.57–13.79)	0.61 (0.08–4.50)	0.42 (0.09–1.93)
Alcohol drinking
Non- to moderate drinkers (≤369 L)	1.00	1.00	1.00	1.00
Heavy drinkers (>369 L)	2.74 (1.45–5.15)	0.55 (0.12–2.59)	6.73 (0.94–48.37)	0.50 (0.11–2.29)

Models of each stratum were adjusted for age, sex and number of educational years (continuous), in addition to the variables in the table.

Stratified analyses by smoking and alcohol status (Table 4) showed that HPV 16 had a stronger effect on OPC risk among lower intensity tobacco and alcohol consumers (OR = 75.49, 95% CI: 29.72–191.77), (OR = 89.07, 95% CI: 34.82–227.84), respectively, than among heavy tobacco and alcohol consumers (OR = 30.38, 95% CI: 8.94–103.26), (OR = 18.59, 95% CI: 5.75–60.13), respectively.

Table 4.

Relationship between HPV status in exfoliated cells and OPC, stratified by smoking and alcohol drinking, HeNCe Life study, Canadian site

	n cases	n controls	OR (95% CI)	n cases	n controls	OR (95% CI)
	Non- to moderate smokersa			Heavy smokersa
HPV negative	44	283	1.00b	25	85	1.00b
LR-HPV	6	25	1.58 (0.60–4.19)	7	9	2.31 (0.73–7.35)
HPV 18, 31, 33, 35, 39, 51	11	9	8.40 (3.10–22.77)	4	6	1.76 (0.43–7.23)
HPV 16	61	6	75.49 (29.72–191.77)	30	4	30.38 (8.94–103.26)
	Non- to moderate drinkers alcohol drinkersc			Heavy alcohol drinkersc
HPV negative	42	286	1.00d	27	82	1.00d
LR-HPV	7	17	3.50 (1.31–9.33)	6	17	0.99 (0.34–2.82)
HPV 18, 31, 33, 35, 39, 51	8	11	4.96 (1.77–13.91)	7	4	5.03 (1.30–19.36)
HPV 16	66	6	89.07 (34.82–227.84)	25	4	18.59 (5.75–60.13)

^aSmoking: Never smokers and smokers up to 37 pack-years/heavy smoking (>37 pack-years).

^bEstimates were adjusted for age, sex, number of educational years (continuous) and total ethanol consumption (none to moderate vs. heavy).

^cAlcohol consumption (ethanol categories): Never drinkers and drinkers up to 369 L ethanol/heavy drinking (>369 L ethanol).

^dEstimates were adjusted for age, sex, number of years of education (continuous) and pack-years (none to moderate vs. heavy).

As a next step, we estimated the joint effects of smoking and HPV exposure (Table 5). In the models assuming independence, the OR related to OPC for heavy smoking among HPV-negative individuals was 1.48 (95% CI: 0.91–2.42), whereas the ORs for non- to moderate smoking and for heavy smoking among HPV 16-positive participants were 52.77 (95% CI: 25.47–109.33) and 78.24 (95% CI: 31.98–191.41), respectively. We found no evidence that the model assuming interaction had a better fit over the independence assumption model (P = 0.228); however, this could be attributed to the small numbers in the cells.

Table 5.

Joint effect of smoking and HPV infection on the risk of OPCs, HeNCe Life study, Canadian site

Smoking pack-years	HPV negative		LR-HPV		HPV 18, 31, 33, 35, 39, 51		HPV 16		Combined effects
	nca	nco	nca	nco	nca	nco	nca	nco	HPV negative OR (95% CI)	LR-HPV OR (95% CI)	HPV 18, 31, 33, 35, 39, 51 OR (95% CI)	HPV 16 OR (95% CI)
Assuming independance
Non-to moderate smokers (≤37 pack-years)	44	283	6	25	11	9	61	6	1.00 (referent)	1.86 (0.91–3.82)	4.82 (2.16–10.73)	52.77 (25.47–109.33)
Heavy smokers (>37 pack-years)	25	85	7	9	4	6	30	4	1.48 (0.91–2.42)	2.76 (1.15–6.59)	7.14 (2.73–18.70)	78.24 (31.98–191.41)
Assuming joint effectsa
Non- to moderate smokers (≤37 pack-years)	44	283	6	25	11	9	61	6	1.00 (referent)	1.44 (0.55–3.78)	7.21 (2.71–19.17)	70.39 (28.12–176.17)
Heavy smokers (>37 pack-years)	25	85	7	9	4	6	30	4	1.71 (0.95–3.05)	4.50 (1.50–13.49)	3.69 (0.94–14.48)	48.76 (15.83–150.17)
Pinteraction = 0.228

Both models adjusted for age, sex, education years (continuous) and ethanol (L) (none to moderate vs. heavy).

n_ca, number of cases; n_co, number of controls.

^aEntering cross-product terms for smoking pack-years (none to moderate vs. heavy) and HPV variable.

Similarly, the model assuming joint effects between alcohol consumption and HPV on OPC risk did not present a better goodness of fit than the model assuming independence of the exposures, which could be attributed to the small sample size (Table 6). In the model assuming independence of effects, the OR for heavy alcohol consumption on OPC risk among HPV-negative subjects was 1.95 (95% CI: 1.17–3.25), whereas that for non- to moderate drinking among HPV 16-positive participants was 52.77 (95% CI: 25.47–109.33). Moreover, the combined effect of heavy drinking and HPV 16 positivity was (OR = 102.80, 95% CI: 39.73–265.98).

Table 6.

Joint effect of alcohol drinking and HPV infection on the risk of OPCs, HeNCe Life study, Canadian site

Ethanol (L)a	HPV negative		LR-HPV		HPV 18, 31, 33, 35, 39, 51		HPV 16		Combined effects
	nca	nco	nca	nco	nca	nco	nca	nco	HPV negative OR (95% CI)	LR-HPV OR (95% CI)	HPV 18, 31, 33, 35, 39, 51 OR (95% CI)	HPV 16 OR (95% CI)
Non- to moderate drinkers (≤369 L)	42	286	7	17	8	11	66	6	1.00	1.86 (0.91–3.82)	4.82 (2.16–10.73)	52.77 (25.47–109.33)
Heavy drinkers (>369 L)	27	82	6	17	7	4	25	4	1.95 (1.17–3.25)	3.62 (1.56–8.40)	9.39 (3.59–24.55)	102.80 (39.73–265.98)
Non- to moderate drinkers (≤369 L)	42	286	7	17	8	11	66	6	1.00	3.37 (1.28–8.92)	4.86 (1.76–13.40)	85.14 (33.63–215.59)
Heavy drinkers (>369 L)	27	82	6	17	7	4	25	4	2.73 (1.48–5.05)	2.73 (0.97–7.69)	13.47 (3.56–50.95)	50.60 (15.96–160.40)

Both models adjusted for age, sex, education years (continuous) and smoking pack-years (none to moderate vs. heavy).

n_ca, number of cases; n_co, number of controls.

^aEntering a joint effect term for ethanol (none to moderate vs. heavy) and HPV variable.

Discussion

Our results suggest that the traditional OPC risk factors, tobacco smoking and alcohol drinking, and HPV infections have independent effects on OPC risk. They might also be interacting in ways that need further investigation, as we noted a relatively small effect of smoking and alcohol on OPC risk among HPV-positive subjects.

In our study, heavy smoking and heavy alcohol drinking independently increased the risk of OPC by 1.90- and 2.74-folds, respectively. Heavy smoking (37 pack-years) corresponds to smoking 3.7 packs per day for at least 10 years. Similarly, individuals who consumed more than 369 L of ethanol in their lifetime, which is equivalent to 2.3 standard servings of hard liquor per day, or approximately 16 standard drinks per week, for 10 years (1 standard drink = 44 ml of hard liquor = 341 ml of beer = 142 ml of wine), were considered heavy drinkers.

The prevalence of HPV infections (63%), and of HPV 16 (48%), in OPC was higher in our study compared with previous ones. Detection rates of any HPV infection in epithelial cells of OPC reported in previous studies ranged from 8.9% to 37% (17,19,21), and the prevalence of HPV 16 DNA has been estimated to be 32% in exfoliated cells of these cancers (19).

Although our interaction analysis may have been underpowered, our results concur with those of other studies suggesting that HPV and smoking are independent risk factors for OPC risk, including a recent pooled analysis of two large studies assessing the interaction between smoking and HPV 16 seropositivity on HNSCC risk (23). In addition, others have observed a lack of effect of HPV (or HPV 16) on the association between tobacco and alcohol and OPC (17,19,20). Similarly, another study reported no synergy between HPV 16 and cumulative pack-years nor with alcohol consumption on an additive scale, but did observe a synergy with current smoking (22). A synergistic effect on OPC risk has also been observed between HR-HPV detected in exfoliated cells and alcohol intake, but not with smoking (21). However, the latter study included histologic types other than squamous cell carcinomas, and the association of alcohol, smoking or HPV with such types is not well established (21). Additionally, the two studies above combined oropharyngeal and oral cancers (21,22).

In addition to HPV 16, our study assessed LR- and HR-HPV types. The small number of participants with these HPV types must be considered, as it could have lead to low precision in the results. As expected, there was a clear gradient in the risk associated with OPC by category of HPV infection. Participants who harbored HPV 16 had a much higher risk of OPC compared with all other risk groups, regardless of smoking and alcohol comsumption. These findings are in agreement with other studies (6,19).

The explanation for the lack of an interaction effect of tobacco and alcohol on HPV-related HNSCC is unclear. However, there are several hypotheses to consider. First, because of the case-control design, the prevalence of HPV among OPC cases across smoking and drinking strata does not capture the population-level burden of HPV-positive disease since it does not consider the underlying incidence of cancer (25). The large ORs observed for the associations between HPV 16 and OPC regardless of smoking and alcohol consumption level indicate that HPV is a strong risk factor for these cancers in our sample. In other words, it is conceivable that the effect of HPV may be so strong as to exert a masking effect on the putative contributory effect of smoking and alcohol. Second, tobacco and alcohol usually inactivate p53 and pRB (16), creating cells with altered phenotype (40). In HR-HPV HNSCC, E6 and E7 oncoproteins also inactivate these tumour suppressor proteins (15,41). Therefore, any additional carcinogenic effects of tobacco and alcohol may act on pathways other than the inactivation of the tumour suppressors (20). Third, a recent study showed a positive association between smoking and secretory leukocyte protease inhibitor (SLPI) expression in HNSCC (42). High SLPI expression protects against HPV infection (43) by reducing its entry into the cells (44). Thus, smoking induces SLPI expression, which in turn inhibits HPV infection (42), and possibly prevents HPV-related HNSCC. Higher SLPI levels were found in well and moderately differentiated HNSCC (45), which is usually the histological presentation of HPV-negative HNSCC (18). SLPI protein levels have also been negatively associated with tumour progression (45), which might explain why HPV-related cancers are usually diagnosed at advanced stages of disease (15).

Our study has some methodological limitations. Our relatively low participation rate was not unexpected and is similar to that of other studies (46–48). However, participation is unlikely to have been associated with HPV status, which was not known to subjects at the time of admission. Moreover, we observed no statistically significant differences in the age and sex of participating and non-participating subjects (data not shown). Although the use of exfoliated cells to assess HPV infection was shown to be predictive of HPV in tumours (12,21,49,50), others reported a low concordance between HPV detected in exfoliated cells and in tumours (17,22). These results may be attributable to the inclusion of cases after treatment (22), as cancerous tissue harboring HPV could have been surgically removed, or chemo- and radio-therapy mucosal effects could have influenced HPV detection. In another study, exfoliated cells were only collected using a brush, and not mouthwash, which limited the collection area (17). In the current study, participants were recruited before they received any treatment, and both mouthwash and a brush were used to collect exfoliated cells. HPV DNA detection in biopsy specimens is undeniably the gold standard (17). Our use of exfoliated cells for HPV detection could have led to exposure misclassification. We believe that any misclassification if present would be non-differential, as the same method of HPV detection was used in the cases and controls. Some studies showed an agreement between HPV detection in exfoliated cells and in H&NC (50–53). Others showed a low concordance, but methodological issues could have affected these results, such as the sampling technique used (only a brush and not mouthwash was used, which could have limited the cell collection area) (54) and the study methodology (recruiting subjects after cancer treatment) (55). Furthermore, it is reassuring that our results are generally in agreement with the literature, for example, the reported stronger association for HPV with OPCs, and the most prevalent HPV type among the cases being HPV-16. HPV detection in biopsy specimens is currently underway in the “HeNCe study”, and results using it will be forthcoming.

Similarly to other case-control studies, we were unable to evaluate the time of HPV infection. Given that most oral HPV infections may be transient in nature (41,56), we cannot assess the exact mechanism of interaction between HPV, smoking and alcohol. Also, as is common in interaction studies, our study had small numbers of participants in some cells after stratification; thus, our findings are less precise.

Some strengths of our study merit mention. To reduce selection bias, cases were recruited from the main referral hospitals in Montreal, and controls were recruited from clinics within the same hospitals. Also, controls were selected from a list of non-chronic diseases unrelated to tobacco smoking or alcohol drinking. Attempts were made to obtain a distribution of disease categories among the controls such that no disease group represented more than 20% of all controls. We compared the age- and sex-stratified smoking and alcohol consumption habits of the HeNCe Life controls with data from the Canadian Community Health Survey (CCHS)-2012 (57), and no major differences were observed (data not shown). This suggests that our hospital controls were representative of the source population. Our study is the first large study to address the effect of HPV status on the relationship between smoking, alcohol and OPC in a Canadian population. The smoking and alcohol exposure variables were created from detailed information, and the accuracy of collected information was optimized by the life-grid technique (27). Most previous studies assessing the interaction between smoking, alcohol and HPV categorized HPV into two broad groups or only addressed HPV 16 (15,17,19–22). Our study goes a step further by classifying HPV into four categories, based on their oncogenic potential in HNSCC (38).

In conclusion, our results highlight that although HPV is a strong risk factor for OPC, smoking and alcohol consumption are important risk factors. We did not detect evidence of synergy between either smoking or alcohol and HPV, but we found a possible suggestion of sub-multiplicative interaction, without evidence of statistical significance. Our results also suggest that the association between HPV 16 and OPC risk is much stronger than with the other HR types. Prospective studies will provide a better understanding of the natural history of oral HPV infections, clarifying mechanisms of interaction with behavioural risk factors and their biological explanations. Finally, we underscore that smoking and drinking reduction efforts in addition to HPV prevention could greatly reduce the burden of OPC.

Funding

BN holds a Canada Research Chair in life course oral epidemiology. MCR was a recipient of a Career Award from the Fonds de recherche du Québec – Santé (FRQS). NJF was a scholarship recipient from the Saudi Cultural Bureau in Canada. This study was funded by CIHR grant numbers: MOP-69062, MOP-201009, DCA-122700 and the Ministère du Développement économique, de l’Innovation et de l’Exportation du Québec.

Abbreviations

 
• CI

  confidence interval

 
• HeNCe

  head and neck cancer

 
• HNSCC

  head and neck squamous cell carcinoma

 
• HPV

  human papillomavirus

 
• HR

  high risk

 
• LR

  low risk

 
• OPC

  oropharyngeal cancer

 
• OR

  odds ratio

 
• PCR

  polymerase chain reaction

 
• SLPI

  secretory leukocyte protease inhibitor

Acknowledgements

Our first acknowledgement goes to the subjects who participated in this study. We also thank the staff at the Department of Otolaryngology — Head and Neck Surgery as well as the control clinics at the Notre-Dame, Jewish General, Royal Victoria and Montreal General Hospitals for their support in the recruitment of study subjects. We would also like to thank Dr Mariam El-Zein and Dr Claudie Laprise for their comments on the manuscript. Finally, we would like to thank the HeNCe Life team and the granting agencies.

Conflict of interest statement: None declared.