Abstract
Our goal was to determine the most important predictors and construct a nomogram for overall survival (OS) in oral cavity squamous cell cancer (OCSCC) treated with primary surgery followed by observation, adjuvant radiation or chemoradiation. Multivariable analysis was performed using Cox Proportional Hazard model of 9258 OCSCC patients from Surveillance, Epidemiology and End Results Program (SEER) database treated with surgery from 2003 to 2009. Potential predictors of OS were age, gender, race, tobacco use, oral cancer sub-sites, pathologic tumor stage and grade, pathologic nodal stage, extra-capsular invasion, clinical levels IV and V involvement, and adjuvant treatment selection. Weighted propensity scores for treatment were used to balance observed baseline characteristics between three treatment groups in order to reduce bias. Following primary surgery, patients underwent observation (56%), radiation alone (31%) or chemoradiation (13%). All tested predictors were statistically significant and included in our final nomogram. Most important predictor of OS was age, followed by pathologic tumor stage. SEER based-survival nomogram for OCSCC patients differs from published models derived from patients treated in a single or few academic treatment centers. An unexpected finding of patient age being the best OS predictor suggests that this factor may be more critical for the outcome than previously anticipated.
Keywords: Oral cavity, Squamous cell cancer, Nomogram, Survival, Age
Introduction
Oral cancers constitute the sixth most common cancer worldwide [1] and represent 2.9% of all new cancer cases in the USA, with estimated 48,330 new cases and 9570 deaths for 2016 per National Cancer Institute Surveillance, Epidemiology and End Results (SEER) Program database [2]. Of these, half are oral cavity squamous cell carcinoma (OCSCC). Following primary surgery for OCSCC, a risk–benefit analysis is performed in order to determine if adjuvant radiation or chemoradiation is indicated. OCSCC patients with risk factors for recurrence [3] are often offered postoperative adjuvant treatment to reduce the rate of recurrence. Conversely, patients for which the risks from adjuvant treatment outweigh the expected benefits are placed on observation. However, the accurate risk-estimation and determination of adjuvant therapy benefit remains a substantial challenge.
To achieve risk estimation, nomogram-based prediction models have been proposed and developed [4, 5] to prognosticate overall and cancer-related mortality and risk of recurrence in individual patients with OCSCC based on host and tumor characteristics. A nomogram is a statistical tool that takes into account numerous variables to predict an outcome based on a variable of interest for an individual [6]. Nomograms have only sparingly been used for head and neck tumors, whilst they have been proven to be extremely useful in a variety of cancers, including prostate [7–11], breast [12–16], pancreas [17] and stomach [18], to name a few. The ability of nomograms to predict survivals has grown hand in hand with the success and availability of good data mining technologies and parallel growth in data mining software.
The goal of this study was determine the most important predictors and to construct a nomogram for overall survival in oral cavity squamous cell cancer treated with primary surgery, using data from SEER.
Materials and Methods
Study Design
The 2003–2010 Surveillance, Epidemiology and End Results Program (SEER) United Stated-national database was queried for patients with a diagnosis of OCSCC from 2003 to 2009 who had been treated with primary surgery and potential predictors of overall mortality were extracted. Patients were included if the site of cancer was an OCSCC site 00-10, surgery had been performed, and the diagnosis was made between 2003 and 2009 to allow for at least 1 year of follow-up. Data were extracted from the patient, physician/supplier, outpatient, hospital, and durable medical equipment SEER data files. Potential predictors of overall mortality included age, gender, race, tobacco use, oral cancer sub-site, pathologic tumor stage and grade, pathologic nodal stage, extra-capsular invasion, and level IV and V involvement, and adjuvant treatment. Surgical margin status was not available in the SEER database and could not be included in the model. Subjects with missing pathologic tumor or nodal stage were excluded from the analysis.
Statistical Analysis
The construction of the nomogram was based on 9258 patients in the SEER database that had OCSCC treated with primary surgery followed by observation, adjuvant radiation, or chemoradiation between years 2003–2009. The outcome measure was overall survival and was defined as the time from the date of diagnosis to death, last date known alive, or last date of follow-up. All statistical analysis was performed using SAS 9.4 and statistical significance was assessed using an alpha level of 0.05. Descriptive statistics by treatment type and by mortality status for all variables (OCSSC sub-site type, age, race gender, tobacco status, treatment, extracapsular invasion, nodal disease in levels IV or V, node stage, tumor stage, and grade) were determined and Chi square and t-tests were used to examine differences. Because baseline demographic and tumor characteristics are associated with choice of treatment and mortality, we used a propensity score weighted method to balance the observed baseline characteristics between the three treatment groups.
Propensity scores reflect the probability that a patient will receive a particular treatment based on their observed baseline characteristics and reduce bias of treatment selection to obtain a better estimate of the treatment effect on survival [19]. The inverse propensity score was assigned as a weight for each patient. These weights were then incorporated into Cox Proportional Hazards (CPH) models for calculating all-cause mortality. Risk factors were first examined individually in simple CPH models. All variables were entered into a more comprehensive multivariable CPH full model. A backward model building strategy was used to build the final model with the final model containing all variables that were statistically significant at the 0.05 alpha level.
Using the beta coefficients from the final CPH model, a nomogram for survival at 5 years was created for patients with OCSCC. The absolute value of the beta coefficient from the CPH model was ranked from high to low with the largest coefficient assigned a point value of 100. Point values for all other beta coefficients were assigned points based on the magnitude of the beta coefficient relative to the magnitude of the largest beta coefficient. Predicted probabilities of 5-year survival were determined using each patient’s nomogram total points and running a CPH model with the independent variable being the nomogram points.
Validation of this nomogram has not been performed at this time. Data external to the SEER data were not available for cross-validation purposes. As well, we decided not to split SEER dataset into the test and validation sets or use bootstrapping techniques for cross validation in order to maximize the statistical power. SEER data are different from other available OCSSC data from the centers of excellence and does not include some variables clinicians feel are important predictors of survival. However, the SEER data are rich in that Medicare patients from both private and academic providers across USA are included.
Results
Descriptive Statistics and Simple Model
Following primary surgery, out of 9258 patients, 5185 (56%) were observed whereas 2870 (31%) underwent radiation alone and the remaining 1203 (13%) underwent both chemo- radiation. Table 1 gives the distribution of baseline demographic and tumor variables by treatment group before and after weighting with the inverse propensity score. The model used to create the propensity scoring weights included age, sex, race, tobacco use, extracapsular invasion, nodal disease in levels IV and V, overall stage and grade (model not shown). Tumor stage and node stage were not included in the propensity score model as these were highly confounded with overall stage.
Table 1.
Descriptive statistics and tests for differences by treatment status with and without propensity score weights
| Variable | Level | Unweighted | Weighted using inverse propensity score | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Rad + Chemo 12.5% | Rad. only 31.1% | Neither 56.4% | p value | Rad + Chemo 12.5% | Rad. only 31.1% | Neither 56.4% | p value | ||
| Oral cancer sub-site (%) | Buccal mucosa | 7.4 | 6.5 | 5.1 | < 0.0001 | 5.6 | 6.0 | 5.6 | 0.9613 |
| Floor of mouth | 15.2 | 14.5 | 10.8 | 12.4 | 12.0 | 12.5 | |||
| Lip | 5.6 | 7.8 | 28.6 | 19.3 | 20.6 | 18.9 | |||
| Other mouth | 2.8 | 3.0 | 2.7 | 3.1 | 2.8 | 3.0 | |||
| Palate | 6.8 | 6.9 | 5.7 | 6.3 | 6.6 | 6.5 | |||
| Tongue | 46.5 | 44.2 | 35.5 | 39.0 | 37.9 | 38.7 | |||
| Gum | 15.7 | 17.1 | 11.7 | 14.2 | 14.2 | 14.8 | |||
| Age—mean (SD) | 69.9 (9.7) | 68.1 (11.8) | 71.3 (11.1) | < 0.0001 | 71.3 (10.3) | 3.2 (11.4) | 70.7 (11.8) | 0.1517 | |
| Gender (%) | Male | 60.5 | 62.3 | 58.6 | 0.0050 | 57.1 | 58.4 | 59.7 | 0.2031 |
| Female | 39.5 | 37.7 | 41.4 | 42.9 | 41.6 | 40.3 | |||
| Race (%) | Black | 5.8 | 6.8 | 3.9 | < 0.0001 | 4.4 | 5.3 | 5.3 | 0.7913 |
| Other | 7.4 | 8.4 | 7.4 | 8.2 | 8.2 | 8.0 | |||
| White | 86.8 | 84.8 | 88.7 | 87.4 | 86.6 | 86.7 | |||
| Tobacco use (%) | Yes | 57.9 | 36.8 | 28.8 | < 0.0001 | 33.5 | 36.5 | 36.1 | 0.1797 |
| No | 42.1 | 63.3 | 71.2 | 66.5 | 63.5 | 63.9 | |||
| Extra-capsular invasion (%) | Yes | 15.7 | 11.1 | 1.5 | < 0.0001 | 6.2 | 6.2 | 7.3 | 0.0996 |
| No | 84.3 | 88.9 | 98.5 | 93.8 | 93.8 | 92.7 | |||
| Node IV or V involvement (%) | Yes | 9.0 | 6.9 | 0.8 | < 0.0001 | 3.6 | 3.7 | 4.4 | 0.1947 |
| No | 91.0 | 93.1 | 99.2 | 96.5 | 96.4 | 95.6 | |||
| Stage (%) | I | 18.4 | 20.7 | 46.4 | < 0.0001 | 34.9 | 34.2 | 34.3 | 0.3317 |
| II | 12.8 | 17.2 | 13 | 13.3 | 14.3 | 14.1 | |||
| III | 17.4 | 17.8 | 5.4 | 10.6 | 10.6 | 11.3 | |||
| IV | 39.9 | 34.1 | 6.5 | 18.9 | 19.0 | 19.9 | |||
| X | 10.1 | 9.3 | 21 | 17.9 | 16.0 | 15.8 | |||
| 0 | 1.4 | 0.9 | 7.6 | 4.5 | 5.8 | 4.7 | |||
| Node stage (%) | N1 | 18.3 | 16.6 | 4.3 | < 0.0001 | 11.0 | 10.3 | 10.4 | < 0.0001 |
| N2/3 | 29.3 | 23.6 | 2.7 | 13.5 | 13.0 | 10.0 | |||
| NX | 2.2 | 1.2 | 2.4 | 4.6 | 2.2 | 2.3 | |||
| N0 | 50.2 | 58.6 | 90.6 | 71.0 | 74.6 | 77.4 | |||
| Tumor stage (%) | T1 | 30.0 | 31.3 | 50.0 | < 0.0001 | 42.2 | 41.1 | 40.8 | < 0.0001 |
| T2 | 30.5 | 32.1 | 15.4 | 22.5 | 23.2 | 20.2 | |||
| T3 | 10.9 | 10.1 | 3.5 | 7.1 | 6.1 | 7.5 | |||
| T4 | 16.5 | 15.2 | 4.2 | 8.1 | 8.6 | 11.5 | |||
| TX | 10.6 | 10.3 | 19.4 | 15.6 | 15.1 | 15.3 | |||
| T0 | 1.4 | 0.9 | 7.6 | 4.5 | 5.8 | 4.7 | |||
| Grade (%) | I | 15.6 | 16.0 | 28.4 | < 0.0001 | 23.7 | 22.6 | 22.3 | 0.4798 |
| II | 45.8 | 48.7 | 36.8 | 41.1 | 40.7 | 40.6 | |||
| III | 25.6 | 22.2 | 8.7 | 15.7 | 15.8 | 17.3 | |||
| IV | 0.8 | 0.9 | 0.5 | 0.6 | 0.6 | 0.8 | |||
| Unknown/NA | 12.3 | 12.3 | 25.6 | 19.1 | 20.3 | 19.0 | |||
All variables showed significant differences between treatment groups when weights were not used (Unweighted columns in Table 1). However, when the inverse PS weights were applied, no statistically significant differences were detected between treatment groups (Weighted Using Inverse Propensity Score columns in Table 1). Figure 1 shows the overall distribution of propensity scores by treatment group. There is sufficient overlap in the propensity scores indicating that the groups are comparable.
Fig. 1.
Distribution of propensity scores by treatment group
Table 2 gives the descriptive statistics by survival status and simple CPH model results. Variables that were significantly associated with survival in simple CPH models include age at diagnosis, race, tobacco use, radiation and/or chemotherapy treatment, extra-capsular invasion, nodal disease in levels IV and V, N stage, T stage, overall stage, grade. Gender was the only variable in simple models that was not statistically significant; however, it was kept in the full model for model building purposes.
Table 2.
Descriptive statistics [n (%), mean (SD)] and univariable CPH hazard ratios on survival
| Variable | Level | Dead N = 3480 (38.7%) | Alive N = 5678 (61.3%) | HR | 95% CI | p value |
|---|---|---|---|---|---|---|
| Oral cancer sub-site | Buccal mucosa | 241 (6.7) | 299 (5.3) | 1.037 | 0.895–1.201 | < 0.0001 |
| Floor of mouth | 514 (14.4) | 642 (11.3) | 0.991 | 0.884–1.111 | ||
| Lip | 538 (15.0) | 1244 (21.9) | 0.637 | 0.571–0.710 | ||
| Other mouth | 95 (2.7) | 162 (2.9) | 0.888 | 0.723–1.091 | ||
| Palate | 220 (6.2) | 355 (6.3) | 0.889 | 0.770–1.027 | ||
| Tongue | 1398 (39.1) | 2265 (39.9) | 0.913 | 0.832–1.002 | ||
| Gumb | 574 (16.0) | 711 (12.5) | ||||
| Agea | 73.6 (11.4) | 68.0 (10.6) | 1.044 | 1.041–1.047 | < 0.0001 | |
| Gender | Male | 2159 (60.3) | 3393 (59.8) | 0.968 | 0.909–1.031 | 0.3122 |
| Femaleb | 1421 (39.7) | 2285 (40.2) | ||||
| Race | Black | 215 (6.0) | 253 (4.5) | 1.255 | 1.100–1.432 | 0.0032 |
| Other | 265 (7.4) | 447 (7.9) | 1.001 | 0.891–1.125 | ||
| White | 3100 (86.6) | 4978 (87.7) | ||||
| Tobacco use | Yes | 1502 (42.0) | 1729 (30.5) | 1.268 | 1.190–1.351 | < 0.0001 |
| Nob | 2078 (58.0) | 3949 (69.6) | ||||
| Radiation or chemotherapy | Rad. only | 1219 (34.1) | 1657 (29.2) | 1.081 | 1.007–1.161 | < 0.0001 |
| Rad. + Chemo. | 805 (22.5) | 351 (6.2) | 1.898 | 1.746–2.062 | ||
| Neitherb | 1556 (43.5) | 3670 (64.6) | ||||
| Extra-capsular invasion | Yes | 349 (9.8) | 230 (4.1) | 2.456 | 2.218–2.719 | < 0.0001 |
| Nob | 3231 (90.3) | 5448 (96.0) | ||||
| Node IV or V involvement | Yes | 194 (5.4) | 150 (2.6) | 2.236 | 1.958–2.554 | < 0.0001 |
| Nob | 3386 (94.6) | 5528 (97.4) | ||||
| Stage | I | 978 (27.3) | 2257 (39.8) | 1.489 | 1.231–1.801 | < 0.0001 |
| II | 538 (15.0) | 784 (13.8) | 1.916 | 1.569–2.339 | ||
| III | 487 (13.6) | 509 (9.0) | 2.550 | 2.085–3.119 | ||
| IV | 990 (27.7) | 793 (14.0) | 3.934 | 3.251–4.760 | ||
| X | 496 (13.9) | 988 (17.4) | 1.595 | 1.307–1.945 | ||
| 0 | 91 (2.5) | 347 (6.1) | ||||
| Node stage | N1 | 465 (13.0) | 451 (7.9) | 1.667 | 1.516–1.832 | < 0.0001 |
| N2/3 | 636 (17.8) | 522 (9.2) | 2.359 | 2.164–2.571 | ||
| NX | 72 (2.0) | 110 (1.9) | 1.372 | 1.138–1.655 | ||
| N0b | 2407 (67.2) | 4595 (80.9) | ||||
| Tumor stage | T1 | 1225 (34.2) | 2634 (46.4) | 1.544 | 1.279–1.865 | < 0.0001 |
| T2 | 928 (25.9) | 1154 (20.3) | 2.317 | 1.912–2.809 | ||
| T3 | 343 (9.6) | 257 (4.5) | 3.986 | 3.242–4.901 | ||
| T4 | 511 (14.3) | 335 (5.9) | 4.180 | 3.429–5.096 | ||
| TX | 482 (13.5) | 950 (16.7) | 1.638 | 1.342–1.999 | ||
| T0b | 91 (2.5) | 348 (6.1) | ||||
| Grade | I | 690 (19.3) | 1434 (25.3) | 1.175 | 1.056–1.306 | < 0.0001 |
| II | 1627 (45.5) | 2224 (39.2) | 1.574 | 1.434–1.727 | ||
| III | 692 (19.3) | 696 (12.3) | 2.306 | 2.078–2.559 | ||
| IV | 32 (0.9) | 27 (0.5) | 2.515 | 1.811–3.494 | ||
| Unknown/NA | 539 (15.1) | 1297 (22.8) | ||||
| Survival in yearsa | 2.3 (1.7) | 3.7 (1.7) |
aContinuous variable—mean (SD) presented and HR is for a one-unit increase
bReferent level
Final Model
The final model (Table 3) contained all risk factors examined in simple models except overall stage. Because overall stage was highly correlated with tumor and node stage, two different models were examined, one with overall stage plus all other potential risk factors and the other with tumor and node stage and all other potential risk factors. The final CPH model presented contains tumor and node stage as this model had better model fit criteria than the model with overall stage. Table 3 also contains the CPH model for the nomogram point variable. For every one-point increase in the nomogram the risk of death increases by 1.04.
Table 3.
Multivariable CPH model on survival
| Risk factor | Level | Beta | SE | HR | 95% CI | p value |
|---|---|---|---|---|---|---|
| Individuals risk factor model | ||||||
| Oral cancer sub-site | Buccal mucosa | 0.228 | 0.076 | 1.256 | 1.082–1.457 | < 0.0001 |
| Floor of mouth | 0.326 | 0.060 | 1.385 | 1.232–1.556 | ||
| Lip | − 0.007 | 0.061 | 0.993 | 0.880–1.12 | ||
| Other mouth | 0.144 | 0.106 | 1.154 | 0.938–1.42 | ||
| Palate | 0.096 | 0.075 | 1.101 | 0.951–1.274 | ||
| Tongue | 0.238 | 0.049 | 1.269 | 1.153–1.396 | ||
| Gumb | ||||||
| Agea | 0.056 | 0.002 | 1.058 | 1.054–1.061 | < 0.0001 | |
| Gender | Male | 0.117 | 0.034 | 1.124 | 1.052–1.202 | 0.0006 |
| Femaleb | ||||||
| Race | Black | 0.160 | 0.069 | 1.174 | 1.025–1.345 | 0.0433 |
| Other | − 0.047 | 0.060 | 0.954 | 0.847–1.074 | ||
| Whiteb | ||||||
| Tobacco use | Yes | 0.208 | 0.034 | 1.231 | 1.153–1.315 | < 0.0001 |
| Nob | ||||||
| Radiation or chemotherapy | Rad. only | 0.042 | 0.037 | 1.043 | 0.970–1.121 | < 0.0001 |
| Rad. + Chemo. | 0.583 | 0.043 | 1.791 | 1.646–1.949 | ||
| Neitherb | ||||||
| Extra-capsular invasion | Yes | 0.331 | 0.063 | 1.392 | 1.229–1.576 | < 0.0001 |
| Nob | ||||||
| Node IV or V involvement | Yes | 0.281 | 0.075 | 1.324 | 1.142–1.535 | 0.0002 |
| Nob | ||||||
| Node stage | N1 | 0.265 | 0.055 | 1.303 | 1.171–1.450 | < 0.0001 |
| N2/3 | 0.500 | 0.060 | 1.648 | 1.465–1.854 | ||
| NX | 0.195 | 0.097 | 1.215 | 1.005–1.469 | ||
| N0b | ||||||
| Tumor stage | T1 | 0.177 | 0.104 | 1.194 | 0.973–1.465 | < 0.0001 |
| T2 | 0.533 | 0.108 | 1.704 | 1.380–2.104 | ||
| T3 | 0.927 | 0.115 | 2.527 | 2.016–3.168 | ||
| T4 | 1.127 | 0.112 | 3.085 | 2.475–3.846 | ||
| TX | 0.314 | 0.107 | 1.369 | 1.111–1.688 | ||
| T0b | ||||||
| Grade | I | 0.038 | 0.058 | 1.039 | 0.928–1.163 | < 0.0001 |
| II | 0.257 | 0.053 | 1.293 | 1.166–1.435 | ||
| III | 0.451 | 0.060 | 1.569 | 1.394–1.767 | ||
| IV | 0.799 | 0.171 | 2.224 | 1.591–3.109 | ||
| Unknown/NAb | ||||||
| Nomogram point model | ||||||
| Nomogram pointsa | 0.041 | 0.001 | 1.042 | 1.040–1.044 | < 0.0001 | |
aContinuous variable—HR is for a one-unit increase
bReferent level
Using the nomogram points, the predicted probability of survival was determined for 5-year survival and the nomogram is given in Fig. 2. Unexpectedly, in the final model and nomogram, the age at diagnosis was decisively the most important predictor of OS, accounting for 100 of total 150 points in the nomogram. The age of diagnosis had hazards ratio of 1.75 for a 10-year increase in age (calculated from 1.06 hazards ratio for a 1 year increase).
Fig. 2.
Nomogram for 5-year survival of oral cancer subsites for lip, gum, palate, tongue, buccal mucosa, floor of mouth and other mouth
Discussion
With the nomogram created in this study, we estimated the predicted 5-year survival for patient who underwent primary surgery for OCSCC, so that subsequent management can be tailored appropriately based on individual patients’ risk versus benefit. Although nomograms cannot replace evidence-based data from large randomized prospective clinical trials, they provide ample support for assessing individual risk and optimizing management. TNM staging cannot be utilized for risk estimation in an individual patient [20]. Existing publications describing prognostic tools to predict the outcome of patients with head and neck cancer are not very applicable to postoperative setting. These initial studies primarily employed regression analysis techniques to stratify patients with head and neck cancer into risk groups based on TNM stage and other pathologic variables, without taking into account individual patient factors or stratifying them based on individual risk [21–23]. Hence, they offered little advantage to the individual patient management while guiding therapy.
Subsequent studies developed prognostic tools with individual patient characteristics to predict outcomes, such as Baatenburg de Jong et al. [24], Van den Broek et al. [25], but the application of these tools was limited since the data was not externally validated and they did not discuss postoperative risk. More recently, Gross et al. [5], published the first prognostic model for an individual using OCSCC nomogram in a postoperative setting which was rigorously validated using an appropriate external dataset. They constructed data from a cohort of 590 patients with OCSCC who were treated at Memorial Sloan-Kettering Cancer Center (MSKCC) and validated it using 417 patients with OCSCC who were treated at Hospital do Cancer AC Camargo (HACC) in Sao Paulo, Brazil. Our study examines similar outcomes and predictors using data from the general population obtained from the United States-national SEER database.
Our results found statistically significant associations of mortality with OCSCC sub-site, age, smoking status, symptom severity such as extracapsular invasion, radiation or chemotherapy administration and initial tumor grade. Similar results have been found in other studies where patient factors have been found to be important for determining outcome survival after treatment of OCSCC, such as age, initial symptom severity, exposure history and preoperative comorbidity [20, 26]. However, our study found a higher association with age compared to Gross et al. [5], with hazards ratio of 1.75 for a 10-year increase in age in our study (calculated from 1.06 hazards ratio for a 1 year increase) compared to hazards ratio of 1.14 for a 10-year increase in age in Gross et al. The effect of age on overall survival as an independent prognostic marker has been studied little for oral cavity cancers. Recent analysis performed using National Cancer Database found poor overall survival with increasing age (p < 0.001) [27]. The effect of age in head and neck cancers in general has been controversial, but recent studies have demonstrated slightly improved overall survival in patients < 45 years of age. One study found survival rate following long-term treatment outcome of head and neck cancer patients after 3–6 years for younger patients (45–60 years) to be 36%, as compared to 31% in the older patient group (>= 70 years) [28]. More conclusive evidence was shown by a prospective study performed in patients with advanced head and neck cancer who underwent surgery (with or without postoperative radiation); multivariate analysis of the surgical group and elective radiation subgroup showed that N stage and age were the strongest predictors of survival and that the method of therapy was not significant [29]. However, another study did not find a statistically significant difference in survival between young and older patients with oropharyngeal cancer, attributing it to human papillomavirus effect [30]. In general, cancer survival is generally higher in people diagnosed aged under 40 years old, with the exception of breast, bowel and prostate cancers, where survival is found to be highest in middle age [31–35].
Our study has a several limitations. The validly of our study relies on the information gleaned from the SEER dataset. SEER data pool has limitations in identifying specific clinical predictors of outcomes are due to the lack of precision of diagnoses and procedure codes inherent to the database. This is due in part to the varied provider experiences and the multi-institutional input to the database. We sought to ameliorate this limitation through application of a stringent inclusion/exclusion criteria. As previously mentioned, risks/benefit analyses are performed following patients’ primary surgery for OCSCC. Those with risk factors for recurrence are often offered post-operative adjuvant therapies. One of the major risk factor for recurrence, which directly impacts survival, is the status of surgical margins post-operatively. Since this important data is not captured by the SEER database, we must infer that our cohort all had margins of resection free of tumor at completion of surgery. Our main outcome measure was overall survival and not disease-free survival; the negative impact of this limitation in essence may have a lesser consequence on arrived-at conclusions. Another limitation of our study, which is also inherent to the SEER database, is the influence of heterogeneity of treatment strategies among different institutions. Also intrinsic to this limitation is the patients’ preferences for different treatment strategies. Due to the lack of an external data set for cross-validation of this nomogram, the results of this study should be interpreted with caution. Despite presenting some limitations of data interpretation, the SEER database as used in our study, has been a key component of many peer-reviewed publications across the board. We believe that with mindful adjustments to these limitations, meaningful and instructive conclusions could be gleaned that could help guide head and neck cancer management.
In conclusion, we have constructed new nomogram predicting overall survival in patients with oral cavity squamous cell cancer treated with primary surgery followed by observation, adjuvant radiation, or chemoradiation based on 9258 patients from Surveillance, Epidemiology, and End Results Program United States-national Database. This SEER-based survival nomogram for OCSCC patients differs from models derived from the patients treated in a single or few academic institutions in that our model shows a higher association of patient’s age to overall survival. Patient related factors, such as age, may be more critical for the outcome than previously anticipated and should be emphasized in the future research. Nomograms such as ours present an exceptionally promising opportunity to refine the principles and practices that will serve as the foundation for precision medicine for cancer management [32].
Compliance with Ethical Standards
Conflict of interest
The authors declare that they have no conflict of interest.
Footnotes
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Bessell A, Glenny AM, Furness S, et al. Interventions for the treatment of oral and oropharyngeal cancers: surgical treatment. Cochrane Database Syst Rev. 2011;9(9):CD006205. doi: 10.1002/14651858.CD006205.pub3. [DOI] [PubMed] [Google Scholar]
- 2.http://seer.cancer.gov/statfacts/html/oralcav.html. Accessed 19 Sept 2016
- 3.Ang KK, Trotti A, Brown BW, et al. Randomized trial addressing risk features and time factors of surgery plus radiotherapy in advanced head-and-neck cancer. Int J Radiat Oncol Biol Phys. 2001;51(3):571–578. doi: 10.1016/S0360-3016(01)01690-X. [DOI] [PubMed] [Google Scholar]
- 4.Gross ND, Patel SG, Carvalho AL, et al. Nomogram for deciding adjuvant treatment after surgery for oral cavity squamous cell carcinoma. Head Neck. 2008;30(10):1352–1360. doi: 10.1002/hed.20879. [DOI] [PubMed] [Google Scholar]
- 5.Montero PH, Yu C, Palmer FL, Patel PD, Ganly I, Shah JP, Shaha AR, Boyle JO, Kraus DH, Singh B, Wong RJ, Morris LG, Kattan MW, Patel SG. Nomograms for preoperative prediction of prognosis in patients with oral cavity squamous cell carcinoma. Cancer. 2014;120(2):214–221. doi: 10.1002/cncr.28407. [DOI] [PubMed] [Google Scholar]
- 6.Kattan MW. Nomograms. Introduction. Semin Urol Oncol. 2002;20:79–81. [PubMed] [Google Scholar]
- 7.Stephenson AJ, Scardino PT, Kattan MW, et al. Predicting the outcome of salvage radiation therapy for recurrent prostate cancer after radical prostatectomy. J Clin Oncol. 2007;25(15):2035–2041. doi: 10.1200/JCO.2006.08.9607. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Stephenson AJ, Scardino PT, Eastham JA, et al. Postoperative nomogram predicting the 10-year probability of prostate cancer recurrence after radical prostatectomy. J Clin Oncol. 2005;23(28):7005–7012. doi: 10.1200/JCO.2005.01.867. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Stephenson AJ, Scardino PT, Eastham JA, et al. Preoperative nomogram predicting the 10-year probability of prostate cancer recurrence after radical prostatectomy. J Natl Cancer Inst. 2006;98(10):715–717. doi: 10.1093/jnci/djj190. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Kattan MW, Zelefsky MJ, Kupelian PA, Scardino PT, Fuks Z, Leibel SA. Pretreatment nomogram for predicting the outcome of three-dimensional conformal radiotherapy in prostate cancer. J Clin Oncol. 2000;18(19):3352–3359. doi: 10.1200/JCO.2000.18.19.3352. [DOI] [PubMed] [Google Scholar]
- 11.Kattan MW. Nomograms are superior to staging and risk grouping systems for identifying high-risk patients: preoperative application in prostate cancer. Curr Opin Urol. 2003;13(2):111–116. doi: 10.1097/00042307-200303000-00005. [DOI] [PubMed] [Google Scholar]
- 12.Rouzier R, Pusztai L, Delaloge S, et al. Nomograms to predict pathologic complete response and metastasis-free survival after preoperative chemotherapy for breast cancer. J Clin Oncol. 2005;23(33):8331–8339. doi: 10.1200/JCO.2005.01.2898. [DOI] [PubMed] [Google Scholar]
- 13.Ravdin PM. A computer based program to assist in adjuvant therapy decisions for individual breast cancer patients. Bull Cancer. 1995;82(suppl 5):561s–564s. [PubMed] [Google Scholar]
- 14.Ravdin PM. A computer program to assist in making breast cancer adjuvant therapy decisions. Semin Oncol. 1996;23(1, suppl 2):43–50. [PubMed] [Google Scholar]
- 15.Ravdin PM, Siminoff LA, Davis GJ, et al. Computer program to assist in making decisions about adjuvant therapy for women with early breast cancer. J Clin Oncol. 2001;19(4):980–991. doi: 10.1200/JCO.2001.19.4.980. [DOI] [PubMed] [Google Scholar]
- 16.Olivotto IA, Bajdik CD, Ravdin PM, et al. Population-based validation of the prognostic model ADJUVANT! for early breast cancer. J Clin Oncol. 2005;23(12):2716–2725. doi: 10.1200/JCO.2005.06.178. [DOI] [PubMed] [Google Scholar]
- 17.Brennan MF, Kattan MW, Klimstra D, Conlon K. Prognostic nomogram for patients undergoing resection for adenocarcinoma of the pancreas. Ann Surg. 2004;240(2):293–298. doi: 10.1097/01.sla.0000133125.85489.07. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Kattan MW, Karpeh MS, Mazumdar M, Brennan MF. Postoperative nomogram for disease-specific survival after an R0 resection for gastric carcinoma. J Clin Oncol. 2003;21(19):3647–3650. doi: 10.1200/JCO.2003.01.240. [DOI] [PubMed] [Google Scholar]
- 19.Rosenbaum Paul R, Rubin Donald B. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70(1):41–55. doi: 10.1093/biomet/70.1.41. [DOI] [Google Scholar]
- 20.Pugliano FA, Piccirillo JF, Zequeira MR, Fredrickson JM, Perez CA, Simpson JR. Symptoms as an index of biologic behavior in head and neck cancer. Otolaryngol Head Neck Surg. 1999;120:380–386. doi: 10.1016/S0194-5998(99)70279-2. [DOI] [PubMed] [Google Scholar]
- 21.Cooper JS, Pajak TF, Forastiere A, et al. Precisely defining high-risk operable head and neck tumors based on RTOG #85–03 and #88–24: targets for postoperative radiochemotherapy? Head Neck. 1998;20:588–594. doi: 10.1002/(SICI)1097-0347(199810)20:7<588::AID-HED2>3.0.CO;2-F. [DOI] [PubMed] [Google Scholar]
- 22.Ang KK, Trotti A, Brown BW, et al. Randomized trial addressing risk features and time factors of surgery plus radiotherapy in advanced head-and-neck cancer. Int J Radiat Oncol Biol Phys. 2001;51:571–578. doi: 10.1016/S0360-3016(01)01690-X. [DOI] [PubMed] [Google Scholar]
- 23.Langendijk JA, Slotman BJ, Van der Waal I, Doornaert P, Berkof J, Leemans CR. Risk-group definition by recursive partitioning analysis of patients with squamous cell head and neck carcinoma treated with surgery and postoperative radiotherapy. Cancer. 2005;104:1408–1417. doi: 10.1002/cncr.21340. [DOI] [PubMed] [Google Scholar]
- 24.de Baatenburg Jong RJ, Hermans J, Molenaar J, Briaire JJ, le Cessie S. Prediction of survival in patients with head and neck cancer. Head Neck. 2001;23:718–724. doi: 10.1002/hed.1102. [DOI] [PubMed] [Google Scholar]
- 25.van den Broek GB, Rasch CR, Pameijer FA, et al. Pretreatment probability model for predicting outcome after intraarterial chemoradiation for advanced head and neck carcinoma. Cancer. 2004;101:1809–1817. doi: 10.1002/cncr.20556. [DOI] [PubMed] [Google Scholar]
- 26.Lydiatt WM, Shah JP, Hoffman HT. AJCC stage group- ings for head and neck cancer: should we look at alter- natives? A report of the head and neck sites task force. Head Neck. 2001;23:607–612. doi: 10.1002/hed.1086. [DOI] [PubMed] [Google Scholar]
- 27.Qureshi MM, Shah BA, Mak KS, Salama A, Ezzat W, Giacalone NJ, Patel SA, Kamran SC, Kirke D, Jalisi S, Truong MT. Determinants of survival in oral cavity cancer patients: an analysis of the National Cancer Data Base (NCDB) Int J Radiat Oncol Biol Phys. 2016;96(2S):E381. doi: 10.1016/j.ijrobp.2016.06.1588. [DOI] [Google Scholar]
- 28.van der Schroeff MP, Derks W, Hordijk GJ, de Leeuw RJ. The effect of age on survival and quality of life in elderly head and neck cancer patients: a long-term prospective study. Eur Arch Otorhinolaryngol. 2007;264(4):415–422. doi: 10.1007/s00405-006-0203-y. [DOI] [PubMed] [Google Scholar]
- 29.Gleich LL, Collins CM, Gartside PS, Gluckman JL, Barrett WL, Wilson KM, Biddinger PW, Redmond KP. Therapeutic decision making in stages III and IV head and neck squamous cell carcinoma. Arch Otolaryngol Head Neck Surg. 2003;129(1):26–35. doi: 10.1001/archotol.129.1.26. [DOI] [PubMed] [Google Scholar]
- 30.Lassig AA, Lindgren BR, Fernandes P, Cooper S, Ardeshipour F, Schotzko C, Yueh B. The effect of young age on outcomes in head and neck cancer. Laryngoscope. 2013;123(8):1896–1902. doi: 10.1002/lary.23932. [DOI] [PubMed] [Google Scholar]
- 31.http://www.cancerresearchuk.org/health-professional/cancer statistics/survival#P7RHYmby3K4APjTG.99. Accessed 21 Oct 2016
- 32.Giardino A, Gupta S, Olson E, Sepulveda K, Lenchik L, Ivanidze J, Rakow-Penner R, Patel MJ, Subramaniam RM, Ganeshan D. Role of imaging in the era of precision medicine. Acad Radiol. 2017;24(5):639–649. doi: 10.1016/j.acra.2016.11.021. [DOI] [PubMed] [Google Scholar]
- 33.https://www.nccn.org/professionals/physician_gls/pdf/head-and-neck.pdf. Accessed 21 Oct 2016
- 34.Specht MC, Kattan MW, Gonen M, Fey J, Van Zee KJ. Predicting nonsentinel node status after positive sentinel lymph biopsy for breast cancer: clinicians versus nomogram. Ann Surg Oncol. 2005;12:654–659. doi: 10.1245/ASO.2005.06.037. [DOI] [PubMed] [Google Scholar]
- 35.Spatz ES, Krumholz HM, Moulton BW. The new era of informed consent getting to a reasonable-patient standard through shared decision making. JAMA. 2016;315(19):2063–2064. doi: 10.1001/jama.2016.3070. [DOI] [PMC free article] [PubMed] [Google Scholar]