Background
The human papillomavirus (HPV) has transformed the landscape of cancer diagnosis and treatment. Oropharyngeal squamous cell carcinoma (OPSCC), a disease traditionally associated with tobacco and alcohol exposure, is now overwhelmingly a disease associated with HPV [
1‐
4]. This change in the epidemiology of the disease has resulted in dramatic improvements in treatment efficacy along with disease specific and overall survival [
1‐
6]. The 8th Edition of the American Joint Committee on Cancer (AJCC) Staging Manual dramatically downstages HPV mediated (HPV+) OPSCC tumors relative to the prior edition [
7]. Specifically, a significant fraction of HPV+ OPSCC tumors previously staged as stage III-IV has now been down-staged to stage I-II, consistent with the improved survival for HPV+ OPSCC patients compared to their HPV negative (HPV-) counterparts [
7].
In 2010, Ang et al. identified an interaction between HPV status and tobacco exposure in OPSCC patients and defined “low-risk” and “intermediate-risk” tumors based on differential survival [
1]. The principal distinction between the two risk categories was a reduced survival in patients who were smokers. Although this observation has been confirmed in subsequent studies, the interaction between HPV and cigarette smoking remains poorly characterized both from a biological perspective and with respect to clinical outcomes [
8]. Tobacco exposure is generally expressed as a function of volume and time (1 pack-year (PY) = 1 pack of cigarettes per day for 1 year). Ang et al. set a threshold of 10 PY to define a “smoker” [
1]. Although sufficient to impact survival, this threshold has not been fully validated in the context of HPV+ OPSCC. For certain patient populations a threshold of 10 PY results in defining nearly 90% of the patient population as smokers [
8]. Studies of Veterans with HPV+ OPSCC have quantified tobacco exposure as high as 150 PY [
8]. Although smoking rates have decreased over the last two decades, evidence suggests that tobacco exposure remains a significant consideration for OPSCC patients throughout the United States and varies widely based on race/ethnicity and socioeconomic strata [
3,
9].
The recent changes to the AJCC staging system for OPSCC create an urgent need to better define the potential impact of tobacco exposure on clinical outcomes for HPV+ OPSCC in order to generate appropriate pre-treatment risk stratification and provide appropriate patient counseling. Furthermore, given continued efforts at treatment de-escalation for HPV+ OPSCC tumors, it is critical to determine whether a subset of patients may be at risk for inappropriate de-escalation. We hypothesized that tobacco exposure has a significant negative impact on HPV+ OPSCC patient survival and sought to test this hypothesis in a large single-institution cohort of patients with robust tumor biological data (concordant p16/HPV testing status) who received radiation-based treatment.
Discussion
OPSCC incidence is rising at an alarming rate in the United States [
2,
3,
18,
19]. The most recent analysis completed in 2018 demonstrated a persistently low rate of HPV preventive vaccination among the US population [
20]. It is, therefore, reasonable to expect that the current increase in OPSCC incidence secondary to HPV is likely to continue for at least the next few decades, especially in North and South America, Central, Eastern, and Northern Europe [
21]. As the OPSCC patient population is expected to increase, it is critical to improve our understanding of how disease biology interacts with and/or determines treatment response [
2,
3,
8,
18,
19]. This requires not only an improved understanding of HPV-mediated effects on tumorigenesis and treatment response, but also a better understanding of the interaction between HPV exposure and other OPSCC risk factors such as tobacco exposure.
Despite a continued decrease in prevalence over the last half century, tobacco use remains associated with 4 out of 5 leading causes of death in the US. The interplay among smoking, HPV infection, other risk factors, and carcinogenesis is complex and multifactorial [
22]. Increasing tobacco exposure has been linked to greater risk of oral HPV infection [
23‐
25]. The interaction between tobacco exposure and HPV infection in OPSCC carcinogenesis and whether the risk of HPV-mediated OPSCC is higher or lower among smokers have been a matter of ongoing debate [
26‐
28]. Our data in the US Veteran population indicate that tobacco exposure is nearly ubiquitous and that approximately 75% of new HPV+ OPSCC diagnoses occur in patients with > 10 pack-year history of tobacco exposure [
8]. Similar data have been reported for other patient populations including indigent, uninsured and underinsured patients [
29,
30].
The 8th Edition of the AJCC staging manual recognized the more favorable prognosis of patients with HPV-mediated OPSCC. As such, the current manual down-stages patients with what was historically regarded as locally advanced disease. However, while recognizing that patients with HPV-mediated disease have better prognoses, it has also been recognized that current and former smokers tend to have worse survival rates compared to non-smokers [
7]. The AJCC was challenged with incorporating smoking into the staging, which was later described in the manual as follows: “the role of tobacco as a negative prognostic factor is well established. However, exactly how this could be codified in the staging system is less clear” [
31]. Even prior to the AJCC 8th Edition era, similar challenges were reported when smoking was introduced into the prognostic framework. The model generated by Huang et al. which included smoking appeared robust for Stages I and II (by criteria used for 8th edition) at a threshold of 20 PY but did not hold for Stage III patients. That was attributed to the detrimental influence of age in the model, potentially related to inability to tolerate intensive treatment for these anatomically more extensive HPV+ lesions (T4 and N3) [
32]. Our data strongly suggest that, although survival is greatly impacted by HPV status, tobacco exposure also plays a very important role. This has been strikingly demonstrated by a poorer survival in both HPV+ and HPV- heavy smokers (i.e. more than double the hazard of death) compared to < 30 PY smokers. This suggests that not all HPV+ OPSCC tumors should be expected to demonstrate the same excellent outcomes we have come to expect. Perhaps most concerning is the significant survival decrease in patients with heavy tobacco exposure. For overall survival, tobacco exposure erases the favorable survival impact of HPV positivity, generating an absolute survival decrement of approximately 16%. As a point of reference, the absolute survival benefit for the addition of chemotherapy to radiation was only ~ 8%, based on the most recent MACH-NC meta-analysis [
33]. We recently updated survival for patients with oral cavity SCC, and identified the relative effect size for nodal metastasis at ~ 15% and extra-nodal extension at ~ 20% [
34]. These data place the impact of tobacco exposure within the range of other treatment modifying clinical-pathologic parameters.
Furthermore, our results showed that the impact of heavy tobacco exposure is alarmingly impacting the outcomes of even early stage disease as defined by the newest edition of the AJCC staging system. Our results showed that the proposed tobacco exposure cutoff of 30 PY clearly stratified patients at each AJCC (8th edition) stage in terms of overall survival outcomes. However, statistical significance was only reached in stage I, probably because of smaller numbers of patients with more advanced disease (i.e. ~ 60% of the whole cohort was categorized as stage I). Nonetheless, heavy smokers with stage I or II disease had 5-year outcomes that were approximating or -in some cases- even worse than outcomes of patients with lower smoking index and more advanced disease stage. This observation must be considered in the inclusion criteria for future dose de-escalation studies in early stage HPV+ OPSCC. In agreement with these results, a recent study by Vawda et al. [
35] has demonstrated that higher intensity of smoking exposure was associated with poorer outcomes in a cohort of exclusively HPV+ oropharyngeal cancer patients treated with primary radiation or surgery. The study, however, lacked the comparison with an HPV- cohort. As a result, the relative effect size of tobacco exposure in the HPV+ OPSCC population remains only partially contextualized. Our findings track closely with those of this recently published data. Moreover, our data indicate that HPV+ patients with smoking index above 30 have surprisingly comparable outcomes to HPV- patients, highlighting the importance of considering this very important risk factor in the treatment decision making process.
In the meantime, the putative impact of smoking on cancer-specific mortality in HPV+ OPSCC cannot be easily interpreted given the known adverse effects of smoking on general comorbidity and interaction with co-existing risk factors (i.e. alcohol consumption) [
36,
37]. Our competing risk analysis indicated that the worse survival outcomes of HPV+ heavy smokers were attributable to two main factors; the increased risk of locoregional failure that leads to more cancer related deaths as well as the overall increased risk of non-cancer related deaths compared with smokers below the identified threshold (Figs.
3 and
4). That’s to say HPV+ heavy smokers who don’t die of smoking-related comorbidities (the dominant detrimental effect of smoking per Fig.
3d) will more probably die of loco-regional failure (Fig.
4d).
Although large, this is a single institution patient cohort and as such our findings must be validated in additional patient cohorts. It is also important to note, that our outcomes for HPV- patients in this series are dramatically better than historical data even from our own institution [
1,
2,
6]. This, combined with the relative small size of the HPV- cohort may confound the comparison between the 3 groups outlined in Fig.
1. Moreover, chemotherapy, despite being associated by multiple randomized controlled trials with better treatment outcomes, was not shown to be an independent prognostic factor even in HPV- subpopulation where more benefit from chemotherapy would be expected [
38]. This can be attributed in part to the retrospective nature of the study where treatment decisions followed the institutional multidisciplinary protocol. That’s to say, patients with early stage disease, with more favorable prognosis, received no chemotherapy in contrast to the more advanced HPV- tumors which still showed worse outcomes despite receiving chemotherapy. The current study focused on patients treated primarily with radiation in order to maximally homogenize the cohort thus allowing us the opportunity to most accurately quantify the impact of tobacco exposure on survival. Clearly, given variable trends in surgery-based treatment for HPV + OPC, additional studies will be required to validate our findings in surgically treated cohorts.
In addition, the tobacco exposure range for the HPV+ cohort is substantially skewed toward the lower range of exposure (Supplementary Figure
2), likely limiting our ability to generate a more granular, dose-dependent effect for tobacco exposure on survival. Moreover, data suggests that current smoking status adversely affects LRC and OS in patients with HNSCC [
39,
40]. This is further exacerbated by the smoking-induced reduction of radiation-induced tumor killing with subsequent worsening of locoregional control [
41,
42]. The unavailability of smoking status for patients while on-treatment –as is the case with our study- might raise a question on the weight that should be assigned to the carcinogenic effect of smoking compared to its antagonistic impact on radiotherapy efficacy. Nonetheless, our results showed no differential disease control or overall survival between current and former smokers, even in the heavy smoker subset (PY > 30). We acknowledge that our study lacks a detailed categorization of history of tobacco exposure per the International Classification of Diseases, Tenth Revision (ICD-10) diagnosis code, as a result of inherent flaws of retrospective data collection [
43]. However, it is still intuitive to extricate from our results the useful public health message relating to advice concerning smoking cessation, especially during radiotherapy course [
44]. Benefits can include limiting radiotherapy/chemotherapy treatment prolongation or interruption and associated heavy symptom burden during and following treatment, in addition to the rapid return of carboxyhemoglobin levels in patients who quit to that of light /never smokers [
45,
46].
Conversely, by limiting our analysis to only those patients with concordant p16 and HPV testing data, we can feel very confident that however limited the data, it is in fact reflective of the underlying tumor biology and not simply a testing artifact. The data generated here conform to what we have come to expect from OPSCC defined by conventional risk factor exposure (i.e. tobacco exposure), namely decreased treatment response and a high rate of loco-regional failure. We previously showed that > 90% of recurrence/progression occurs loco-regionally in a patient cohort with significant tobacco exposure [
8]. The fact that tobacco exposure is a distinct competing risk for loco-regional failure not only suggests an impact on treatment effectiveness, but is particularly concerning when considering current efforts to de-escalate treatment for HPV+ OPSCC patients.
Based on this dataset, we strongly recommend development of a multi-institutional cooperative group focused on characterizing and quantifying the relative impact of tobacco exposure on HPV + OPSCC clinical outcomes. Until such time that definitive national datasets can be generated, we recommend strong consideration of tobacco exposure in the context of ongoing institutional and cooperative group trials aimed at de-escalation regimens for HPV+ OPSCC along the same lines as the phase II/III PATHOS and the recently completed NRG HN002 randomized controlled trials [
47,
48]. Furthermore, we recommend a dedicated multi-institutional effort aimed at validating the current dataset and developing additional guidance for consideration of tobacco exposure in the context of the AJCC 8th Edition Staging Manual.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.