Overview
Prognostic Research for Salivary Gland Cancer
Prognostic research for patients with salivary gland carcinoma relates specific prognostic factors to specific oncologic outcomes. Outcomes generally studied are “overall survival” (death from any cause as the central event); “disease specific survival” (death due to tumor – reflects the best obtainable treatment result of all possible treatment given); and “disease control” (freedom of tumor recurrence: local, regional, at distance, or a combination of these, reflecting treatment result following initial therapy). Even recently published research frequently remains limited to prognostic factor identification.
Specific Focus: Development, National and International External Validation, and Clinical Application of a Prognostic Scoring System for Disease Control
The 10-year disease specific survival (DSS) rates for patients with parotid carcinoma treated in major centers range from 47% to 69%. This range reflects the population specific distribution of stage, percentage of high-grade tumors, treatment period and regimens, patient inclusion criteria and follow-up quality. These “group” DSS figures are too general to counsel any specific patient. A unique patient features a unique set of prognostic factors (patient-, tumor- and treatment-related, clinical, pathologic, and increasingly molecular biologic factors) that jointly imply a better or worse outcome than the whole group’s prognosis. The focus of this short chapter is on recent research efforts that summarize important prognostic factors and their respective weights into a unique individual patient’s prognosis.
Research Echelons in Prognostic Research
Research echelons are (1) univariate and multivariate identification of prognostic factors; (2) summarizing the identified factors into a user-friendly score or nomogram, and (3) preferably repeated external validation (proving applicability outside the source population). Studies that are higher on the echelon ladder imply an increased research effort but also increased clinical usefulness. Several studies have now reached this third level.
Univariate and Multivariate Survival Analyses for Patients With Parotid Carcinoma
As opposed to univariate Kaplan–Meier analysis, Cox proportional hazards multivariate analysis corrects the impact of one prognostic factor for the effect of other factors, and increases clinical usability. Cox based models incorporate: (1) patient factors (age, gender, pain, and comorbidity); (2) tumor factors (histologic type, grade, stage, skin and soft tissue invasion, facial nerve involvement and perineural growth, molecular biologic factors); and (3) treatment factors (resection margins and adjuvant radiotherapy), and are presented as a table listing the factors, the accompanying hazard ratios (HR) that reflect their respective weight, and p -values/confidence intervals (CI). Sadly, most clinicians seeking to counsel their patient fail to intuitively amalgamate this tabulated information into a concrete patient-specific prognosis.
Towards Increased User-Friendliness
Summary Score or Nomogram
A prognostic score (1) or a nomogram (2) summarizes the tabulated information from multivariate analysis in a practical easy-to-use format and is the next step on the ladder of clinical user-friendliness. The final step is (repeated) external validations.
1
Development of a Prognostic Score for Recurrence Free Interval in Patients With Parotid Carcinoma.
Multivariate analysis of the cohort of the Netherlands Cancer Institute resulted in a set of prognosticators that best explained the observed variability in “recurrence free interval” at two important clinical moments: one set based on the diagnostic work-up information (preoperative score PS1), the second set incorporates the postoperative histopathology information (postoperative score PS2: see Box 46.1 ). Filling out the formulas produces a score that reflects the weighted joint effect of the prognostic factors and corresponds to an individualized risk of recurrence of one patient. PS1 assigns the preoperative patient to one of four prognostic groups with good (92%, 5-year recurrence free) somewhat worse (83%), intermediate (48%), and poor prognosis (23%). The postoperative PS2 score classifies the patient as having 5-year recurrence free rates, ranging from good (95%), somewhat worse (83%), intermediate (56%), to poor (42%). A comparable effort was done in Mexico.
Preoperative Prognostic Score:
PS1 = 0.024 A + 0.62 P + 0.44 T + 0.45 N + 0.63 S + 0.91 F
Postoperative Prognostic Score:
PS2 = 0.018 A + 0.39 T + 0.34 N + 0.70 S + 0.56 F + 0.78 PG + 0.65 PM
A , age in years; P , 1 (no pain), 2 (pain/numbness); T , T1-T4 (0,1,2,3); N , N0-N3 (0,1,2,3,4,5); S , 1 (no skin invasion, 2 (invasion); F , 1 (normal VIIth nerve function), 2 (paresis/paralysis); PG , 1(no perineural growth), 2 (PG); PM , 1 (negative margins), 2 (close or involved margins).
2
Creation of a Nomogram for “all Sites of Major SGC”.
Authors from the Memorial Sloan Kettering Cancer Center (MSKCC) summarized their multivariate analysis into a nomogram (containing age, grade, vascular/perineural invasion and nodal metastasis) predicting “5-year recurrence”. They next presented a nomogram predicting “10-year survival” (containing age, grade, cT4, perineural invasion and tumor dimension) and one predicting “5 years cause specific survival” (based on grade, perineural invasion, cT4, positive nodal status, and positive margins”).
External Validation
Assessing “external validity” of a score or a nomogram supports clinical usefulness outside the population that was used to create the system. This final step has been taken repeatedly for PS1 and PS2, and recently, also for the MSKCC nomograms.
External Validation of PS1 and PS2 – Clinically Usable Format
Clinical and statistical validation was assessed by the authors in both a national and an international validation, and also by a Brazilian and a Taiwanese group, by evaluating “discrimination” ( Fig. 46.1 ) and “accuracy” (calculating Harrell’s C-index). In our own international validation study, discrimination and accuracy were good: an expected fall in the C index from 0.80 to 0.74 for PS1 and from 0.78 to 0.74 for PS2 was observed. Similarly, the Taiwanese study, assessing PS2, observed a C index of 0.74, but the latest UICC stage grouping also performed well. The Brazilian group did not find the same performance, but their sample differed in several important aspects. Provided the same therapeutic and follow-up strategy is used, PS1 and PS2 have been proven transportable (geographically, in time, in methodology, and in follow-up) and generalizable to patients outside the source population. A downloadable fill-out form for use in the clinic is available at: http://www.uzleuven.be/parotid and is now also integrated in the University Hospitals Leuven electronic patient file ( Fig. 46.2 ). The clinician introduces a set of prognostic factors for a patient with a previously untreated parotid carcinoma (in the example in Fig. 46.2 : a 36-year-old male with a T1N0 acinic cell carcinoma without pain, skin invasion or facial nerve dysfunction), and gets a scientifically based, validated 5-year recurrence free percentage (in the example: 94%), answering the patient’s question: “How likely will the upcoming treatment beat the disease?”
