Association of Preoperative Functional Performance With Outcomes After Surgical Treatment of Head and Neck Cancer

This cohort study evaluates the association of the addition of adult patients’ functional performance to other important preoperative clinical characteristics with 30-day and 90-day hospital readmissions and overall survival after surgical treatment of head and neck cancer.

Key Points

Question

Is evaluation of preoperative functional performance, when added to other well-recognized clinical factors, associated with greater prognostic ability for outcomes following surgical treatment of head and neck cancer?

Findings

In this cohort study of 657 patients, evaluation of preoperative functional performance, comorbidity, weight loss, and TNM stage was associated with independent prognostic information for unplanned hospital readmission, complications, and overall survival. When these 4 variables were combined to create a composite clinical severity staging system, there was an association between greater severity stage and adverse outcomes.

Meaning

Poor preoperative functional performance is associated with negative outcomes for patients treated surgically for head and neck cancers, and it might be combined with other important clinically available factors to help predict adverse outcomes.

Abstract

Importance

Patients with head and neck cancers have comorbidities and other constitutional symptoms known to be associated with adverse postoperative outcomes, but the role of functional performance is not well studied.

Objective

To explore the addition of functional performance to other clinical factors for association with 3 patient outcomes: 30-day unplanned readmission (UR), 90-day medical complications, and overall survival (OS).

Design, Setting, and Participants

This retrospective cohort study was conducted in a single tertiary care center with patients surgically treated for squamous cell cancer of the lip, oral cavity, pharynx, or larynx from January 2012 to December 2016. All analysis took place between January 2018 and November 2018. Data from 2 registries were analyzed, supplemented with medical record review. Logistic regression analysis was used to explore association of preoperative functional performance with outcomes. Conjunctive consolidation was used to create a useful clinical severity staging system, which included functional performance (estimated from metabolic equivalent [MET] score: <4, light-intensity activities; ≥4 at least moderate-intensity activities); overall comorbidity severity; preoperative weight loss; and TNM tumor staging. Logistic regression was used to assess the prognostic accuracy of the clinical severity staging system for 30-day UR and 90-day complications, and Cox proportional hazard regression for OS.

Exposures

All patients underwent surgical treatment for head and neck cancer.

Main Outcomes and Measures

The primary outcomes were 30-day UR and 90-day complications; the secondary outcome was OS.

Results

For the 657 patients included, the mean (SD) age was 62.0 (11.3) years; 73% were men (n = 477), and 88% were white (n = 580). A total of 75 (11%) had a 30-day UR; 204 (31%) developed a 90-day complication; and 127 (19%) patients died during the observation period. Individually, poor functional performance (<4 METs), high comorbidity burden, preoperative weight loss, and advanced TNM stage were associated with all 3 outcomes; the increased risk for each outcome ranged from 1.5 to 3.0 times the reference range. Using these 4 variables in combination, the 4-category clinical severity staging system demonstrated a strong association between severity stage and all 3 adverse outcomes: 30-day UR (C statistic, 0.63), 90-day complications (C statistic, 0.63), and OS (C statistic, 0.68).

Conclusions and Relevance

Poor preoperative functional performance, high comorbidity burden, preoperative weight loss, and advanced tumor stage were all associated with worse patient outcomes after head and neck cancer surgery. The model incorporating all 4 of these factors developed in this study may facilitate patient-centered risk assessment and patient-physician shared preoperative decision making.

Introduction

Head and neck cancers (HNCs) account for 3% to 4% of new cancer diagnoses in the United States and affect nearly 65 000 patients a year. Of these cancers, 40% to 50% are detected in patients 65 years or older.¹ Primary surgical treatment of these cancers is associated with substantial morbidity and mortality.² Currently, prognostic research for patients with HNC undergoing curative surgical treatments largely focuses on tumor factors, such as morphological spread at the time of presentation, and few patient factors, such as comorbidity burden and age.³ Current risk assessment for curative surgical treatment in older patients may capture higher comorbidity burdens,⁴ but it does not capture functional performance.

The association of functional performance with surgical outcomes is of considerable interest^5,6 with potential opportunities for quality improvement. Previous research shows that patients who require assistance with activities of daily living, even after controlling for comorbidity, have higher rates of surgical site infection and death.^7,8,9,10 In a large meta-analysis of population-based cohorts, patient-reported leisure-time physical activity levels had dose-dependent association with mortality; higher levels of leisure-time physical activity were associated with a 15% to 20% lower risk of cancer-related deaths, even after controlling for patient age.¹¹ Leisure-time physical activity, measured using estimated metabolic equivalents (METs)—a measure of reported physical activity—is often used to assess cardiac risk for patients undergoing noncardiac surgery.¹² However, the predictive capacity of subjective functional performance assessment, as well as its accuracy compared with objective assessment by exercise testing, is the subject of current debate.^13,14

The scientific literature on HNC outcomes inconsistently includes measures of performance status. Researchers have incorporated modified frailty indices and substituted measures of comorbidity for functional performance.¹⁵ Large cancer registries currently do not include functional performance measures. Inconsistent capture or complete absence of functional performance information impedes the scientific evaluation of patients with HNC and comparative treatment effectiveness.

The association of pretherapeutic functional performance with outcomes and treatment decision making seems intuitive—patients with better baseline functional performance prior to surgery might be expected to have improved survival compared with patients with worse functional performance. In this study of patients with HNC undergoing surgical treatment, we explore the addition of functional performance to other important clinical factors on patient outcomes. We sought to develop a model that would be useful for clinical decision making.

Methods

Study Design

This was a retrospective cohort study combining prospectively gathered patient information from Siteman Cancer Center at Barnes-Jewish Hospital and Washington University Oncology Data Services Cancer Registry with anesthesiology preoperative clinic data from the MetaVision Anesthesia Information Management System (iMDsoft) database. The cancer registry is an American College of Surgeons–accredited registry that includes follow-up information. In addition to obtaining standard data elements,¹⁶ trained cancer registrars capture information on individual cogent comorbid ailments at the time of presentation using the Adult Comorbidity Evaluation-27 (ACE-27)¹⁷ instrument for all newly diagnosed patients with cancer. The captured comorbid health information has been incorporated into many observational studies.^18,19,20 The Department of Anesthesiology at Washington University uses MetaVision to store functional performance measures captured by trained nurses, nurse practitioners, and anesthesiology resident physicians during preoperative visits that occur within 30 days of surgery. Patient-reported maximum physical activity over the most recent 2 weeks is compared with published estimates of activity-associated METs.²¹ Common examples of peak activity measuring 4 METs or more include the ability to climb 1 or 2 flights of stairs without stopping or brisk walking on level ground (at about 3.5 mph). Poor functional performance is classified as less than 4 METs, while good functional performance is classified as 4 METs or more. This study received institutional review board approval from Washington University School of Medicine on May 25, 2017, waiving the need for informed consent of study participants.

Patients

Patients were included if they (1) were 18 years or older at the time of surgery; (2) had a diagnosis of squamous cell carcinoma of the oral cavity, oropharynx, hypopharynx, or larynx; (3) were undergoing primary curative surgical treatment (index surgery) at Barnes-Jewish Hospital between January 2012 and December 2016; and (4) had preoperative evaluation at the Center for Preoperative Assessment and Planning. Patients who underwent debulking or palliative surgery for their tumor or curative chemoradiation treatment were excluded.

Outcomes and Follow-up Information

The 2 primary outcomes studied were (1) 30-day unplanned readmission (UR) to Barnes-Jewish Hospital and (2) 90-day complications (surgical site infection, pneumonia, stomal or tracheoesophageal puncture complications, pharyngocutaneous fistula, chyle fistula, or other serious complications related to the surgical treatment of the patient’s cancer, classified by the Dindo-Clavein²² system as severity grade 2 or higher—ie, requiring medical treatment, reoperation, or admission to the intensive care unit to treat the complication). The secondary outcome was overall survival (OS). The duration of time between surgery and date of last patient contact or death defined the length of follow-up. For 30-day UR and 90-day complications, information was obtained through a review of the medical record by one of us (S.S.).

Definition of Zero Time

For 30-day UR, zero time was defined as the day of discharge after index surgery. For 90-day complications and OS duration, zero time was defined as the date of index surgery. This zero time allowed (1) comparison of patients at a comparable time point in their clinical course²³ and (2) capture of complications during the index surgery hospitalization.

Analytical Plan

Analysis was conducted on IBM SPSS Statistics for Windows, version 24.0 (IBM Corp) and rms library in R, version 3.3.2. Basic descriptive statistics, univariable logistic regression, multivariable logistic regression, and Cox proportional hazards analyses were used. Variables significantly associated with each outcome were entered into multivariable models. Conjunctive consolidation,^24,25 a method of intuitively clustering clinical data using cross-table analysis based on biological coherence and statistical isometry, was used post hoc to develop the composite clinical severity staging system. Variables that were significantly associated with each of the 3 outcomes were entered into the conjunctive consolidation model. Odds ratios (ORs) from logistic regression or hazard ratios (HRs) from Cox proportional hazard regression were used to assess strength of association between the new conjoined staging system and outcomes. Discriminative power of the models was assessed by C statistic. Internal validation of the clinical severity staging system was performed using a bootstrapping technique with resampling using 200 bootstrap samples to correct for optimism of the C statistic.²⁶

Results

Description of the Study Population

For the 657 patients included in the study, the mean (SD) age was 62.0 (11.3) years ; 477 were men (73%); 580 were white (88%); 520 were current or previous smokers (79%); 408 were without a partner (62%); and 295 had private insurance (57%). A minority of patients (21%, n = 141) had poor functional performance; 39% (n = 253) had moderate or severe overall comorbidity burden; and 18% (n = 117) had preoperative weight loss. A majority of patients (67%, n = 438) had TNM stage III or IV cancers. The association of each variable with the outcomes is detailed in the Table.

Table. Patient Characteristics and Associations With Outcomes.

Characteristic	30-Day Unplanned Readmission			90-Day Complication			Overall Survival, HR (95% CI)
	No	Yes	OR (95% CI)	No	Yes	OR (95% CI)
Patients, No.	582	75	NA	450	207	NA	657
Age, median (range), y	60.7 (18.8-93.9)	64.0 (32.8-85.3)	1.02 (1.00-1.04)	60.9 (22.1-93.9)	61.6 (18.8-90.4)	1.01 (0.99-1.02)	1.03 (1.02-1.05)
Sex
Male	421 (72)	56 (75)	1.13 (0.65-1.96)	331 (73)	146 (71)	0.86 (0.60-1.24)	0.71 (0.49-1.02)
Female	161 (28)	19 (25)	1 [Reference]	119 (26)	601(29)	1 [Reference]	1 [Reference]
Race
White	516 (89)	64 (85)	1 [Reference]	402 (89)	175 (86)	1 [Reference]	1 [Reference]
Other	66 (11)	11 (15)	1.34 (0.68-2.68)	48 (11)	29 (14)	1.36 (0.83-2.24)	1.08 (0.65-1.80)
Ever smoked
No	128 (22)	9 (12)	1 [Reference]	105 (23)	32 (16)	1 [Reference]	1 [Reference]
Yes	451 (78)	66 (88)	2.07 (1.00-4.26)	345 (77)	175 (84)	1.66 (1.08-2.57)	1.18 (0.76-1.83)
Drug abuse
No	531 (93)	68 (92)	1 [Reference]	410 (93)	189 (92)	1 [Reference]	1 [Reference]
Yes	42 (7)	6 (8)	1.12 (0.46-2.72)	32 (7)	16 (8)	1.09 (0.58-2.03)	1.03 (0.52-2.03)
Marital status
Partner	220 (38)	28 (37)	1 [Reference]	161 (36)	88 (43)	1 [Reference]	1 [Reference]
No partner	359 (62)	47 (63)	1.03 (0.63-1.69)	289 (64)	119 (57)	0.75 (0.54-1.05)	0.68 (0.47-0.96)
Insurance status
Private	272 (48)	23 (31)	1 [Reference]	223 (50)	72 (36)	1 [Reference]	1 [Reference]
Medicare	213 (37)	42 (56)	2.33 (1.35-4.00)	163 (37)	92 (45)	1.75 (1.21-2.53)	1.98 (1.32-2.97)
Medicaid/other	88 (15)	10 (13)	1.34 (0.61-2.93)	60 (13)	38 (19)	1.96 (1.21-3.19)	2.14 (1.28-3.55)
TNM stage
Stage 1 or 2	203 (35)	16 (21)	1 [Reference]	169 (38)	50 (24)	1 [Reference]	1 [Reference]
Stage 3 or 4	379 (65)	59 (79)	1.98 (1.11-3.52)	281 (62)	157 (76)	1.89 (1.30-2.74)	1.67 (1.11-2.51)
Cancer site
Lip/oral cavity	214 (37)	33 (44)	1 [Reference]	143 (32)	104 (51)	1 [Reference]	1 [Reference]
Pharynx	245 (43)	27 (36)	0.72 (0.42-1.23)	213 (48)	58 (29)	0.38 (0.26-0.56)	0.38 (0.25-0.58)
Larynx	115 (30)	15 (20)	0.85 (0.44-1.62)	89 (20	41 (20)	0.63 (0.41-0.99)	0.76 (0.49-1.18)
Grade
Grade 1 or 2	269 (46)	43 (57)	1 [Reference]	205 (46)	107 (52)	1 [Reference]	1 [Reference]
Grade 3 or 4	129 (22)	15 (20)	0.73 (0.39-1.36)	81 (18)	63 (30)	1.49 (1.00-2.23)	1.81 (1.25-2.62)
Unknown	184 (32)	17 (23)	0.58 (0.32-1.05)	164 (36)	37 (18)	0.43 (0.28-0.66)	0.24 (0.13-0.45)
Comorbidity
None/mild	361 (63)	37 (49)	1 [Reference]	293 (65)	105 (51)	1 [Reference]	1 [Reference]
Moderate/severe	215 (37)	38 (51)	1.72 (1.06-2.80)	154 (35)	99 (49)	1.74 (1.28-2.51)	2.05 (1.44-2.92)
BMI
Underweight	40 (7)	5 (7)	1.02 (0.36-2.83)	30 (7)	15 (7)	0.85 (0.43-1.67)	1.54 (0.86-2.75)
Normal weight	187 (32)	23 (31)	1 [Reference]	132 (29)	78 (38)	1 [Reference]	1 [Reference]
Overweight	192 (33)	20 (27)	0.85 (0.75-1.59)	157 (35)	55 (26)	0.59 (0.39-0.90)	0.68 (0.44-1.04)
Obese	163 (28)	27 (36)	1.35 (0.74-2.44)	131 (29)	59 (28)	0.76 (0.50-1.16)	0.42 (0.25-0.69)
Weight loss
No	473 (83)	54 (73)	1 [Reference]	378 (86)	149 (73)	1 [Reference]	1 [Reference]
Yes	97 (17)	20 (27)	1.81 (1.03-3.15)	61 (14)	56 (27)	2.33 (1.55-3.51)	2.70 (1.85-3.95)
Functional performance
≥4 METs	465 (80)	51 (68)	1 [Reference]	365 (81)	151 (73)	1 [Reference]	1 [Reference]
<4 METs	117 (20)	24 (32)	1.87 (1.11-3.16)	85 (19)	56 (27)	1.59 (1.08-2.35)	2.16 (1.50-3.12)
Nursing needs
Independent	522 (96)	60 (95)	1 [Reference]	407 (96)	172 (96)	1 [Reference]	1 [Reference]
Dependent	24 (4)	3 (5)	1.09 (0.32-3.72)	19 (4)	8 (4)	0.98 (0.42-2.28)	2.28 (1.15-4.51)
Radiation therapy
No	288 (49)	33 (44)	1 [Reference]	231 (51)	90 (44)	1 [Reference]	1 [Reference]
Yes	294 (51)	42 (56)	1.25 (0.77-2.02)	219 (49)	117 (56)	1.37 (0.99-1.31)	0.98 (0.69-1.39)
Chemotherapy
No	431 (74)	54 (72)	1 [Reference]	334 (74)	151 (73)	1 [Reference]	1 [Reference]
Yes	151 (26)	21 (28)	1.11 (0.65-1.90)	116 (26)	56 (27)	1.07 (0.74-1.55)	1.28 (0.76-2.16)

Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); OR, odds ratio; HR, hazard ratio; MET, metabolic equivalent; NA, not applicable.

There were 75 (11%) patients with 30-day UR and 204 (31%) patients with 90-day complications. Among patients with 90-day complications, 106 (51%) received medical intervention; 68 (33%) required surgical treatment; 18 (9%) had intensive care unit admission for a treatment-related complication; and 15 (7%) required interventional radiology procedures. The median follow-up time in the study was approximately 21.2 months (range, 0.26-64.9 months), and there were 125 (19%) deaths in the cohort.

Multivariable Logistic Regression

For 30-day UR, the unadjusted OR of poor functional performance was 1.87 (95% CI, 1.11-3.16), and the adjusted OR was 1.35 (95% CI, 0.77-2.38); significant covariates that remained in the multivariable model were insurance status and TNM stage. For 90-day complications, the unadjusted OR of poor functional performance was 1.59 (95% CI, 1.08-2.35), and the adjusted OR was 1.35 (0.79-1.97); significant covariates that remained in the multivariable model were smoking status, TNM stage, cancer site, comorbidity burden, and preoperative weight loss. For OS, the unadjusted HR of poor functional performance was 2.16 (95% CI, 1.50-3.12), and the adjusted HR was 1.35 (95% CI, 0.89-2.04); significant covariates in the multivariable model were age at surgery, TNM stage, cancer site, comorbidity burden, BMI class (body mass index, calculated as weight in kilograms divided by height in meters squared), and preoperative weight loss (eTable 1 in the Supplement).

Conjunctive Consolidation

The first step of conjunctive consolidation was to examine the rates of UR and complications and the HRs for OS within the 4 conjoined categories of functional performance (poor vs not poor) and comorbidity burden (none/mild vs moderate/severe) (Figure 1). Within each of the 2 levels of comorbidity, there was a gradient for rates of the outcomes across functional performance. Likewise, within each of the 2 levels of functional performance, there was a gradient of outcomes across levels of comorbidity. This conjoint association with outcome is referred to as a double gradient or gradient within a gradient. Based on statistical isometry and biologic coherence, different conjoint cells of functional performance and comorbidity were consolidated to create a new 3-category severity staging system called the functional severity system. The outcome rates for each of the 3 outcome measures for each of the 3 functional severity system categories (good, fair, and poor) are shown in Figure 1.

Figure 1. Conjunction of MET Score and Comorbidity.

Data for 30-day unplanned readmission and 90-day complications are reported as number of patients/total possible number (percentage); overall survival is reported as hazard ratio (95% CIs). ACE-27 indicates Adult Comorbidity Evaluation 27; MET, metabolic equivalent.

The functional severity system was then conjoined with preoperative weight loss (yes or no), yielding 6 conjoined categories (Figure 2). Patients with similar rates of outcomes were combined to create the 3-category patient severity staging system (A, B, and C).

Figure 2. Conjunction of Functional Severity System and Preoperative Weight Loss.

Data for 30-day unplanned readmission and 90-day complications are reported as number of patients/total possible number (percentage); overall survival is reported as hazard ratio (95% CI).

Patients with different combinations of patient severity categories were then conjoined with patients with different tumor stage (Figure 3) to create the 4-category clinical severity staging system (CSS) (Alpha, Beta, Gamma, and Delta). The discriminative power of the CSS for each of the 3 outcomes was moderate, with a C statistic of 0.63 for 30-day UR, 0.63 for the 90-day complications, and 0.68 for OS. Internal validation based on 200 bootstrap samples displayed an average optimism of 0.001 around the C statistics. The absolute differences (95% CIs) in the rates of outcomes for patients in the best vs worst CSS categories were 17.0% (6.5%-27.0%) for 30-day UR and 40% (25% to 52%) for 90-day complications, Relative to the alpha category, the HR for OS based on CSS categories beta, gamma, and delta were, respectively, 2.79 (1.09-7.18), 6.62 (2.65-16.52), and 10.36 (4.02-26.67) (Figure 3). The new clinical severity staging system was significantly associated with each outcome after controlling for other significant confounders (eTable 2 in the Supplement). Kaplan-Meier survival curves based on the 4 clinical severity staging categories are shown in Figure 4. The clinical severity staging system developed by conjunctive consolidation can be accessed at http://jpiccirillo.com/EXTRAS/clinicalSeverity/.

Figure 3. Conjunction of Patient Severity and TNM Stage.

Data for 30-day unplanned readmission and 90-day complications are reported as number of patients/total possible number (percentage); overall survival is reported as hazard ratio (95% CI).

Figure 4. Kaplan-Meier Survival Curves of the Groups of the New Clinical Severity Staging System.

Discussion

This study demonstrates that functional performance is associated with clinically important postoperative outcomes—30-day UR, 90-day complications, and OS—for patients with HNCs, when conjoined with other important risk factors. Functional performance was defined by MET scoring of patient peak activity according to well-defined criteria. A new staging system generated from combining functional performance status with comorbidity, preoperative weight loss, and tumor stage discriminated patients at higher risk for outcomes. These clinical and tumor factors are readily available, require no additional testing, and can be incorporated into clinical practice for real-time clinical use and research. Of note, discrimination for the prediction of 30-day UR was relatively modest, but consistent with a recent systematic review, highlighting the challenge of predicting UR, particularly given that approximately 20% of URs occur at a hospital outside of the index hospital.^27,28 Model discrimination for 90-day complications and overall survival reflected possibly helpful discrimination.²⁹

Assessment of functional performance in the preoperative clinic is common, likely related to recommended use for perioperative cardiac morbidity estimation within the American College of Cardiology guidelines.¹² The strength of evidence for and against initiating preoperative cardiac testing and optimization of cardiac comorbidities varies across patient exercise capacities and is affected by the relative rarity of ischemic events in the postoperative period.

Older and colleagues³⁰ examined the utility of cardiopulmonary exercise testing (CPET) and anaerobic testing as a screening test to grade the susceptibility to surgical stress and to determine postoperative management. The 548 patients in that study all underwent major surgery and were either older than 60 years (n = 476) or had previous diagnoses of myocardial ischemia or cardiac failure (n = 72). The authors concluded that anaerobic testing was able to predict mortality from cardiovascular causes. However, anaerobic testing is a logistically complex and lengthy procedure.

Wiklund et al³¹ prospectively enrolled 5939 inpatients admitted for elective, noncardiac surgery. Of the enrolled patients, 94 (1.6%) had cardiac complications, but MET was not independently predictive. The authors speculated that inconsistency in the determination of MET by the variety of physicians in the preanesthetic assessment phase may have contributed to poor predictive performance. Notably, patients rated as METs fewer than 4 contributed almost half the cardiac complications, despite representing less than 20% of the study population. Another study by Wijeysundera and colleagues¹³ did not demonstrate a relationship between subjectively assessed functional capacity and postoperative cardiac complications. In their multicenter, international study, 1401 adults 40 years or older scheduled for major abdominal, pelvic, or orthopedic procedures and deemed to have 1 or more risk factors for cardiac complications or coronary artery disease were assessed to examine the association between several different types of preoperative assessment, immediate postoperative death, and major complications. Patients’ functional capacity was subjectively categorized as poor (<4 METs), moderate (4-10 METs), or good (>10 METs).

Fitness was also assessed by the Duke Activity Status Index (DASI) questionnaire.³² Of 1401 patients in the study, 28 (2%) died or had a myocardial infarction within 30 days of surgery. Subjective assessment of functional capacity was not associated with any clinical end points. DASI scores were associated with the primary outcome (adjusted OR, 0.96; 95% CI, 0.83-0.99). Peak oxygen capacity measured during CPET was associated only with in-hospital moderate to severe complications (adjusted OR, 0.86; 95% CI, 0.78-0.97), but not with 30-day myocardial injury/infarction or death or 1-year death. Consequently, the authors concluded that subjectively assessed functional capacity should not be used for preoperative risk evaluation; instead, clinicians should consider the DASI questionnaire and measures of peak oxygen concentration through CPET and NT pro-BNP concentrations (amino-terminal pro–B-type natriuretic peptide). There are several limitations of this study. The aggregate findings, that both clinician-estimated capacities and objective CPET testing were not predictive of study outcomes, while patient-reported, questionnaire-based subjective capacities were predictive, are challenging to reconcile. Additionally, there is no indication that the study used training or a standardized method for the clinician-estimated functional capacity assessment. Variability in measuring functional capacity may have reduced the ability to detect the true association between clinician-estimated functional capacity and outcomes. Finally, the evidence for the utility of DASI in the preoperative setting is exceedingly limited—this is the only study, to our knowledge, that has evaluated the perioperative prognostic accuracy of DASI.

Reilly et al³³ were the first to describe the association between self-reported exercise tolerance and the risk of serious perioperative complications. In that study, 600 adults were scheduled for major noncardiac surgery, and poor exercise tolerance was defined as inability to walk at least 3 blocks or climb 2 flights of stairs. Poor exercise tolerance was an independent predictor of postoperative complications (adjusted OR, 1.94; 95% CI, 1.19-3.17). Moreover, the likelihood of a serious complication was inversely related to the number of blocks the participant could walk or the number of flights of stairs they could climb. In a study by Arozullah et al,^34,35 functional performance independently predicted both respiratory failure and pneumonia. Qaseem et al³⁶ developed guidelines based on a systematic review of the literature,³⁷ and they concluded that partial functional dependence nearly doubled the risk, and total dependence nearly tripled the risk for postoperative pulmonary complications after noncardiac surgery. Functional performance has also been found to be independently predictive of surgical site infections.³⁸

The present study shows that a low-cost, low-technology assessment of the functional performance of patients with HNC can provide valuable independent prognostic information. Decreased functional performance identified during the preoperative assessment provides an opportunity for patient-centered risk assessment, including modifying patient expectations and better anticipating postoperative management needs. In the future, identification of decreased functional performance may be a target for preoperative optimization. At Barnes-Jewish Hospital, patients with fewer than 4 METs often receive additional attention focused on identifying reversible comorbid conditions and delineating the relative risks and benefits of surgical treatment.

Strengths and Limitations

The preoperative clinic at the study institution is exceptional in that it is a highly standardized evaluation center that evaluates approximately 35 000 patients annually. Estimation of MET is part of the standard training of preoperative clinicians and is used to drive clinical protocols used on a daily basis, such as disease severity stratification, test ordering, and risk assessment. The center uses several methods for training and quality maintenance in MET assessment, including mentored training, written reference materials, reminders embedded in the medical record, and monthly quality-assessment conferences. We recognize that this clinic infrastructure may be unique in its size and the efforts applied to standardizing MET capacity and other assessments, and this may influence the generalizability of the results.

Nonetheless, this study adds valuable information to the field of HNC treatment. The rate of 30-day UR was comparable to that in prior reports.^{15,19,39,40,41,42,43,44} Among patients with HNC, factors known to be associated with UR after surgery include patient comorbidities,^{15,17,18,19,41,42,44,45} marital/partner status,⁴² insurance status,⁴¹ postoperative complications,^19,39,40,42 complexity of surgery,⁴⁴ and advanced tumor stage.^40,42 Higher comorbidity burden is consistently associated with high risk of postoperative complications.^18,46,47 Complication rates range from 21% to 36%, with different lengths of follow-up and different levels of complication severity. Overall survival among patients with HNC is associated with patient age, traditional performance measures, comorbidities, weight loss at presentation, cancer site, TNM staging, and adjuvant radiation therapy,^3,48,49 much like the findings in the present study.

This study has other important limitations. Readmissions to any hospital outside of the index hospital could not be captured. Because the study was restricted to patients with HNC who received surgical treatment, the results are unlikely to be generalizable to patients who receive other treatments, including chemoradiation therapy or palliative treatment or to adult patients with other solid tumors undergoing surgery. Additionally, most of the study patients were elderly, white men. Finally, the model has not been externally validated in an unrelated cohort of patients with HNC.

The additional strengths of this study include the use of 2 institutionally maintained registries, which prospectively capture validated functional performance and comorbidity information, thereby eliminating risks of bias associated with retrospective data review. In addition, the individuals who completed the functional performance and comorbidity assessment are explicitly trained in proper data collection, minimizing random and systematic errors. Furthermore, the collection of this information is mandated by the study hospital, and therefore failure to capture data is minimized. However, there are many factors including social and environmental factors that might affect a patient’s recovery from surgery, and further studies should explore the addition of such factors to the model presented herein. The parsimonious model developed in this study allows clinicians to offer prognostic information and risk-stratify patients on 3 important patient-centered outcomes at their initial presentation to a clinic for discussion of surgical treatment for HNC.

Conclusions

The findings of this study support the clinical significance of preoperative functional performance, comorbidity burden, weight loss, and morphological extent of tumor at the time of treatment for patients with HNC. The composite clinical severity staging system requires factors easily obtained at the time of the preoperative appointment. Future research should investigate the utility of pretherapeutic interventions such as improving functional performance, optimizing comorbid ailment conditions, and reversing weight loss to improve patient outcomes in cancer care.

Supplement.

eTable 1. Multivariable Models

eTable 2. Multivariable Models with new Clinical Severity Stage