Differences in clinical and imaging characteristics between p16-positive non-smokers and p16-positive smokers or p16-negative patients in oropharyngeal carcinoma

The eighth edition of the TNM classifies oropharyngeal squamous cell carcinomas (OSCCs) depending on p16 status. Some imaging features are reportedly associated more frequently with p16-positive (P16+) OSCC than p16-negative (p16−) OSCC. However, classical risk factors such as tobacco use were not specifically considered when assessing these imaging features. We aimed to evaluate whether P16+ OSCCs have different epidemiological, clinical, prognostic and imaging features depending on smoking status, and to compare P16+ and p16− groups. A retrospective study of data from 85 patients with P16+ OSCC (41 non-smokers, 44 smokers) and 36 with p16− OSCC from 2011 to 2020 was carried out, assessing epidemiological data, clinical aspects of the tumour and presence of adenopathy. Staging was assessed according to the seventh and eighth editions of the TNM. Compared with P16+ OSCC non-smokers, P16+ OSCC smokers had tumours that were less well-defined (36.6% vs 77.8%, p < 0.001), more ulcerated (85.4% vs 44.4%, p < 0.001) and more necrotic (53.7% vs 25%, p = 0.012). There was also less downstaging from N2 or N3 of the seventh edition of the TNM to N1 of the eighth edition for smokers than non-smokers (22.7% vs 43.9%, p = 0.042). Compared with p16− tumours, more P16+ tumours had well-defined contours (55.8% vs 22.2%, p = 0.001), were exophytic (89.6% vs 72.2%, p = 0.023), less necrotic (40.3% vs 80.6%, p < 0.001), less ulcerated (97.2% vs 66.2%, p = 0.006) and involved less muscle tissue (26.0% vs 47.2%, p = 0.027).P16+ OSCCs of smokers show clinical, imaging and prognostic differences with P16+ OSCCs of non-smokers.


Methods
Data from patients with head and neck cancer from January 2011 to January 2020 were retrieved from our institution database by using "oropharyngeal cancer", "palatine tonsil", "base of tongue" and "HPV16" as search terms. p16 status was determined by anatomopathological examination of a biopsy of the lesion or of the resected lesion after surgery. Exclusion criteria were no pre-treatment CT available and tumour in a location other than the oropharyngeal palatine tonsils or base of tongue.
Of the 101 patients with P16+ OSCC, 85 fulfilled our inclusion criteria ( Fig. 1) and two groups were considered: smokers and non-smokers (patients who had never smoked or had quit smoking for more than 20 years).
To obtain three groups of comparable size, we selected the first 40 consecutive patients with p16− status from January 2017 onwards (before this date, data are more often incomplete). Four patients were excluded because no pre-treatment CT images were available.

Research of HPV infection status.
Research of HPV was done on demand before 2017 and systematically from 2017 on patients who had an OSCC. p16 was detected by immunohistochemistry and sometimes complementary high-risk HPV PCR is performed to confirm unclear p16 status. p16 overexpression was considered as a surrogate to HPV infection 7 . As reported by Pannone et al. 15 , the p16 immunohistochemistry method proved a 100% sensitivity rate in OSCC confirmed HPV positive by PCR and/or In Situ Hybridation and a 93.5% specificity.
Acquisition of images. CT with iodine contrast enhancement was used to acquire images from the base of the skull to the base of the cervicothoracic junction, followed by a thoracic CT. Dynamic maneuvers, such as opening of the mouth or opening of the mouth with extension of the tongue were occasionally performed to improve the image quality (in the case of metal artefacts) and the depiction of abnormalities 16,17 . Analysis of images. Images were displayed on a PACS workstation (Fujifilm Synapse v4.1.600, Fujifilm, USA) and assessed by a radiologist with 4 years' clinical experience. CT and medical records were read from April 2019 to May 2020.
Analysis of patient and tumour characteristics. The following data were collected from the medical files: sex, age at diagnosis of the cancer, tobacco consumption, pack-year, years since quitting smoking, alcohol habitus and p16 overexpression. Patients were considered drinkers or non-drinkers on the basis of clinicians' reports.
Classification of the local extent (T) of the tumour was carried out according to both the 7th and 8th editions of the TNM by pooling together T1-T2, and T3-T4 stages. The primary site of the tumour was recorded (palatine tonsils, base of tongue), and subsequent extension towards oropharyngeal sublocations (tonsils, base of tongue, floor of the mouth, soft palate). Presence of metastases was also noted and reported according to the TNM.
Adenopathy was also assessed based on N of the TNM staging system (both 7th and 8th TNM editions); as N0, N1 or N2-N3 (N2 and N3 were grouped together). Morphological characteristics of adenopathy were also recorded: bilaterality, extra-nodal extension and cystic nodes (Fig. 5). Keratinizing histological type was also assessed.
Statistical analysis. Quantitative parameters were the mean and standard deviation, and qualitative parameters were frequency and percentage.
Description of each parameter was assessed in P16+ and p16− group. Bivariate logistic regressions were performed to define parameters associated with P16+ status. Results were expressed as odds ratio and 95% confidence interval. Parameters with a p-value of less than 0.1 were introduced in a multivariate logistic regression with backward selection. The stability of the final model was investigated with Bootstrap resampling. The discriminant power of the final model was evaluated with the area under the curve (AUC) and was considered to be good if the AUC was greater than 0.8.
The same method was applied to compare data from P16+ smokers and non-smokers. Ethical approval. The study has been declared to the French National Commission on Information Technology and Liberties (CNIL) on April 12 2019 and has been registered as a CNIL compliance declaration (MR004-2203860) by the Data Protection Officer of the Cancer Institute of Lorraine French Region (Number 165). According to the CNIL MR004 compliance declaration, all patients are informed of the potential retrospective use of their clinical data for research purpose and of their right to refuse. Individual written consent was consequently waived in light of those elements, in compliance to CNIL MR004 declaration and the study was approved by the scientific and ethical committee of the Institut de Cancérologie de Lorraine, Nancy, France.

Results
Among the 85 patients with P16+ OSCC, 44 (51.8%) were smokers. Median tobacco consumption was 34 packyears in P16+ smokers and 40 pack-years in p16− patients. Sixteen patients among the 41 non-smokers with P16+ OSCC were former smokers (median time since quitting, 27.5 years). Characteristics of patients are presented in Table 1. Comparison of data from P16+ and p16− patients are presented in Table 2. Comparison of data from P16+ non-smokers and P16+ smokers are presented in Table 3.
Bilateral adenopathy was more frequent in p16− patients (46.4% vs 20.6%, p = 0.012), whereas cystic adenopathy was more frequent in P16+ patients (75% vs 53.6%, p = 0.043). Metastases were also more frequent in p16− patients (27.8% vs 5.9%, p = 0.002).     There was a significant relationship between the primary location of the tumour and HPV status (p = 0.002), with more primary tumours located in the palatine tonsils in P16+ OSCC. Involvement of the base of the tongue and floor of mouth was less frequent in P16+ OSCC (p < 0.001 and p = 0.004, respectively).
Multivariate analysis of data from the P16+ patients shows that two factors are significantly associated with smoking: in P16+ smokers, there were fewer tumours with well-defined contours (OR 0.3, 95% CI [0.1; 0.95]) and more tumours with ulceration (OR 4.1, 95% CI [1.22; 13.75]) than in P16+ non-smokers. The AUC was equal to 0.77.

Discussion
This study confirms previously reported differences in clinical, epidemiological and imaging features in P16+ OSCC and p16− OSCC that were distinctly identified in the 8th edition of the TNM classification. This study also sheds light on differences in P16+ OSCC in smokers and non-smokers that may be considered separately in future staging models.
Epidemiological review. In our study, P16+ smokers developed tumours sooner than P16+ non-smokers. Alcohol abuse was not significantly different between the two groups. A synergistic association of the two risk factors in the cancerogenesis could explain the occurrence of cancer at a younger age in our population.
We defined non-smokers as people who had never smoked or who had quit smoking more than 20 years, as described by Marron and al. 18 . This study has shown that the risk of oropharyngeal cancer after 20 years of cessation of tobacco smoking is equivalent to that of people who have never smoked. Imaging features of P16+ OSCCs and p16− OSCCs. Our study shares some similar findings with the previous studies 13, 14 : P16+ OSCCs are more often well-defined, exophytic, and less ulcerated and necrotic than p16− OSCCs. P16+ OSCCs also show less aggressive radiological features such as muscle tissue involvement. Adenopathy is also often more cystic and less often bilateral in P16+ patients than in p16− patients. P16+ OSCCs were significantly less often keratinized on anatomopathological examination than p16− OSCCs in our study, as described in the literature 19,20 . In multivariate analysis, three factors differed significantly between P16+ patients and p16− patients: alcohol abuse and tumour necrosis were associated less with P16+ patients than p16− patients, whereas exophytic tumours were associated more with P16+ patients than p16− patients. The AUC was equal to 0.817 and these three parameters could distinguish well the two groups.

Imaging features of OSCCs of P16+ non-smokers and P16+ smokers.
Our study reveals some differences in P16+ OSCCs depending on tobacco abuse that have not been reported previously to our knowledge. P16+ smokers have significantly fewer tumours with well-defined contours and their tumours are more often ulcerated and necrotic. In multivariate analysis of data from all eligible P16+ patients, two factors were found to be significantly associated with smoking: tumours with well-defined contours were less frequent in P16+ smokers and tumours with ulceration were more frequent. The AUC was equal to 0.77 and these two factors could help to distinguish the two groups. Non-keratinizing histology was dominant but not significantly different in non-smoker P16+ OSCC patients of our population.
Treatment and prognosis. P16+ OSCC is generally considered to respond best to radiotherapy and chemotherapy 21,22 . Several studies have proposed de-escalation of irradiation dose and chemotherapy to reduce induced toxicity and complications in P16+ OSCC 22 . In our study, p16− patients had more advanced tumours associated with a worse prognosis than P16+ patients, consistent with the literature 7-12 . According to classification with the 8th edition of the TNM, P16+ patients had better prognosis; significantly more p16− patients had stage III or IV tumours than P16+ patients in our study. Our study also shows that P16+ smokers tend to have higher tumour staging than P16+ non-smokers according to the 8 th TNM classification; there were significantly more P16+ smokers with stage III or IV tumours than P16+ non-smokers. Considering the link between staging and prognosis, prognosis could be worse in P16+ smokers than in P16+ non-smokers. If this observation is confirmed in larger populations, treatment choice may need to take into consideration the smoking status. As expected, there was significant downstaging of P16+ tumours from the 7th to the 8th TNM classification. As the T criteria were not modified, the downstaging is mainly based on the N criteria. N2 ipsilateral unique or multiple adenopathy in the 7th TNM classification is now N1 in the 8th edition. For example, a T1-T2N2b in the 7th edition became a T1-T2N1 in the 8th edition, and as a consequence the staging changes from IVa to I. However, this downstaging may in fact reflect the better prognosis of P16+ OSCC in the non-smoker group, as the P16+ smokers could have more aggressive lesions.
Our study also shows that there is a link between tobacco use and P16+ OSCC characteristics. There was also significantly more downstaging from N2 or N3 of the 7th edition to N1 of the 8th edition in the non-smoker group. This may mean that P16+ smokers have more aggressive lesion, either locally or by node involvement or metastasis.
Limitations. Our study has several limitations. Tumour assessments were from CT images, which have lower contrast resolution than MRI but better spatial resolution. Our center used to perform CT first instead of MRI to cover the thoracic area in the same examination time. Extension to muscle is consequently more difficult to evaluate even if muscle stranding or fat adjacent to muscle stranding indicates likely muscle involvement. Categorization of patients as drinkers or non-drinkers was also based on clinical reports, but no reliable quan- www.nature.com/scientificreports/ tification was possible in our study. Moreover, alcohol abuse may be under-estimated by the patient. However, alcohol status was not significantly different in our population of P16+ patients, whatever their smoking status.

Conclusions
OSCCs are currently stratified according to their p16 status in the 8th TNM classification, without taking into consideration historical risk factors such as tobacco and alcohol consumption. Our study shows that P16+ OSCCs associated with tobacco abuse may be more aggressive than P16+ OSCCs in non-smoker patients, although their estimated alcohol status was not significantly different. In future studies it will be interesting to assess the outcomes and overall life expectancy of the P16+ OSCCs depending on their smoking status. If large studies confirm our findings, historical risk factors could also be considered in the future elaboration of the next edition of the TNM classification.