183 / 220
Predictors of survival
For all patients with oropharyngeal SCC in this
cohort, the following factors were found to be statistically
significant predictors of survival in the univariate models
(Table 2): age, surgical approach, extranodal extension,
HPV status, marital status, mucosal margins, p16 status,
perineural invasion, primary tumor site, race, smoking
status, T classification, and low/intermediate/high-risk
stratification (by HPV or p16).
For HPV-negative patients, the following factors were
significant predictors of survival in the univariate models:
extranodal extension, perineural invasion, primary tumor
site, and T classification (Table 2). In the HPV-positive
cohort, the following factors were significantly associated
with survival (Table 2): surgical approach and T classifi-
cation. The presence of extranodal extension nearly
reached significance (
p
5
.0553).
In the multivariable analysis (
n
5
260), after adjust-
ment for all covariates, HPV status, extranodal extension,
and T classification were independent predictors of sur-
vival (Table 3). Of note, aside from HPV status, several
of the factors that were predictive of survival in the
RTOG 0129 (smoking status and neck disease) failed to
reach significance in this patient cohort after adjustment
in the multivariable model.
For HPV-negative patients (multivariable model
n
5
108), extranodal extension (HR of 2.322 for those with
extracapsular spread [ECS]; 95% CI
5
1.359–3.968;
p
5
.0021) and T classification (HR of 2.029 for those with
T3/T4 disease; 95% CI
5
1.054–3.906;
p
5
.0342) were
the only independent predictors of survival. For HPV-
positive patients (multivariable model
n
5
152), the sur-
gical approach and mucosal margins were the only 2 fac-
tors predictive of survival (Table 3). Within the HPV-
positive subgroup, those with open surgical approaches
had over 3 times the hazard of death than those with
transoral surgical approaches (HR
5
3.09; 95% CI
5
1.293–7.385;
p
5
.0111). Those with positive margins
had a HR of 2.519 (95% CI
5
1.101–5.766;
p
5
.0287).
There were no differences in positive margin rates
between the transoral and open surgical approaches (
p
5
.2868).
Recursive partitioning analysis
RPA of this patient cohort revealed that HPV status
was the most important predictor of overall survival (see
Figure 2). For HPV-positive patients, the best outcomes
were achieved in those who underwent transoral surgery
and had no evidence of perineural invasion at the primary
site (9 deaths/104 patients; 91.3% survival). In addition,
for those HPV-positive patients who were treated with an
open surgical approach, the margin status was the next
most important predictor of survival, with 17 deaths out
of 46 patients (63.0% survival) at last follow-up with
negative margins as opposed to 6 deaths out of 8 patients
(25.0% survival) if the margins were positive. For HPV-
negative patients, the most important predictor of survival
was the presence of ECS (58.2% vs 27.5% survival). Sur-
vival rates were worst for T3/T4 tumors with ECS
(13.8%).
TABLE 2.
Continued
Overall
HPV negative
HPV positive
Variables
HR
95% CI
p
value
No. of
patients
HR
95% CI
p
value
No. of
patients
HR
95% CI
p
value
No. of
patients
Smoking status
10 pack-years
Ref
Ref
Ref
>
10 pack-years
2.006 1.216–3.308
.0064
283
1.707 0.778–3.745 .1819
116
1.226 0.606–2.482 .5706
160
TNM classification
I/II
Ref
Ref
Ref
III/IV
1.025 0.498–2.111
.9461
294
1.918 0.828–4.445 .1287
117
0.739 0.175–3.114 .6799
171
T classification
T1/T2
Ref
Ref
Ref
T3/T4
2.946 1.999–4.341
<
.0001
296
2.158 1.321–3.525 .0021
117
3.192 1.643–6.201 .0006
172
Abbreviations: HPV, human papillomavirus; HR, hazard ratio; 95% CI, 95% confidence interval; BOT, base of tongue.
K
UMAR ET AL
.
HEAD & NECK—DOI 10.1002/HED APRIL 2016
163