Association of Delayed Time to Treatment Initiation With Overall Survival and Recurrence Among Patients With Head and Neck Squamous Cell Carcinoma in an Underserved Urban Population

Key Points

Question

What are the consequences of delaying treatment initiation in an urban underserved population of patients with head and neck squamous cell carcinoma?

Findings

In this cohort study of 956 patients with head and neck squamous cell carcinoma, delayed treatment initiation beyond 60 days was associated with worse survival and greater risk of recurrence independent of other relevant factors. Associated patient factors and reasons for treatment delay were also identified.

Meaning

Delayed treatment initiation contributes to worsened outcomes, and risk factors and reasons for delay should be targeted with intervention.

This cohort study investigates the association of delayed treatment initiation with overall survival and recurrence among patients with head and neck squamous cell carcinoma among an underserved urban population.

Abstract

Importance

Delay in time to treatment initiation (TTI) can alter survival and oncologic outcomes. There is a need to characterize these consequences and identify risk factors and reasons for treatment delay, particularly in underserved urban populations.

Objectives

To investigate the association of delayed treatment initiation with outcomes of overall survival and recurrence among patients with head and neck squamous cell carcinoma (HNSCC), to analyze factors that are predictive of delayed treatment initiation, and to identify specific reasons for delayed treatment initiation.

Design, Setting, and Participants

Retrospective cohort study at an urban community-based academic center. Participants were 956 patients with primary HNSCC treated between February 8, 2005, and July 17, 2017, identified through the Montefiore Medical Center Cancer Registry.

Exposures

The primary exposure was TTI, defined as the duration between histopathological diagnosis and initial treatment. The threshold for delayed treatment initiation was determined by recursive partitioning analysis.

Main Outcomes and Measures

Overall survival, recurrence, and reasons for treatment delay.

Results

Among 956 patients with HNSCC (mean [SD] age, 60.8 [18.2] years; 72.6% male), the median TTI was 40 days (interquartile range, 28-56 days). The optimal TTI threshold to differentiate overall survival was greater than 60 days (20.8% [199 of 956] of patients in our cohort). Independent of other relevant factors, patients with HNSCC with TTI exceeding 60 days had poorer survival (hazard ratio, 1.69; 95% CI, 1.32-2.18). Similarly, TTI exceeding 60 days was associated with greater risk of recurrence (odds ratio, 1.77; 95% CI, 1.07-2.93). Predictors of delayed TTI included African American race/ethnicity, Medicaid insurance, body mass index less than 18.5, and initial diagnosis at a different institution. Commonly identified individual reasons for treatment delay were missed appointments (21.2% [14 of 66]), extensive pretreatment evaluation (21.2% [14 of 66]), and treatment refusal (13.6% [9 of 66]).

Conclusions and Relevance

Delaying TTI beyond 60 days was associated with decreased overall survival and increased HNSCC recurrence. Identification of predictive factors and reasons for treatment delay will help target at-risk patients and facilitate intervention in hospitals with underserved urban populations.

Introduction

Head and neck squamous cell carcinoma (HNSCC) can progress and upstage during prolonged time to treatment initiation (TTI), which may increase mortality¹ and the likelihood of recurrence.² Delayed treatment initiation can also result in increased costs due to the need for more aggressive treatment and evoke patient anxiety.³ Despite the apparent benefits of avoiding treatment delay, analysis of the National Cancer Database between 1998 and 2011 found that the median TTI in head and neck cancer had increased from 19 to 30 days.⁴ Murphy et al⁴ ascribed this trend to the progression toward greater complexity of multimodality therapy, increased pretreatment radiologic/pathological workup, and more transitions in care.

Among studies characterizing the association of TTI with survival in HNSCC, there is large variability in the TTI thresholds used to define delayed TTI. The threshold for delayed TTI has ranged from 20 to 120 days in studies using different methods for selecting these thresholds.⁵ In addition, while some studies^5,6,7,8,9 supported an association between delayed TTI and survival, other studies^10,11,12,13 have not. Some of these studies are from a national data registry, which may not hold external validity with regard to a local population.^11,14 This may be especially salient in a socioeconomically disadvantaged and mostly minority population like that in the Bronx, New York, where $19 721 is the mean per capita annual income.¹⁵ National registries also do not contain specific data related to recurrence rates. Finally, there is a need to identify factors associated with delayed treatment initiation that are specific to the local population to better risk stratify patients and intervene in a prospective fashion.

In the present study, we sought to identify a threshold for delayed TTI based on survival in patients with HNSCC from a large urban health center in New York City, and we analyzed the association of delayed TTI with overall survival and recurrence. In addition, we investigated factors predictive of delayed TTI and reasons for treatment delay in patients with HNSCC.

Methods

This retrospective cohort study at an urban community-based academic center included 956 patients treated for primary HNSCC at Montefiore Medical Center (MMC) in Bronx, New York, between February 8, 2005, and July 17, 2017. The study protocol was approved by the Institutional Review Board at the Albert Einstein College of Medicine/MMC with a waiver of informed consent as a retrospective study with no risk of harm to patients. All patient identifying information was handled in a compliant manner. Patients with squamous cell carcinoma of the head and neck region (lip/oral cavity, oropharynx, hypopharynx, larynx, nasopharynx, and nose/sinus) were identified through the Montefiore Medical Center Cancer Registry. Patients were treated with radiotherapy (RT), chemoradiotherapy (CRT), surgery alone, or surgery with adjuvant RT/CRT. We excluded patients who had stage IVC cancer/metastatic disease, patients who underwent noncurative palliative treatment, patients who received chemotherapy only, and patients who had a TTI of 0 (reflecting a coding error) (eFigure 1 in the Supplement). Characteristics of the excluded patients with a TTI of 0 and sensitivity analysis of excluding these patients can be found in eTable 1 and eTable 2 in the Supplement.

Time to treatment initiation was defined as the number of days between histopathological diagnosis and first treatment. An optimal TTI threshold was identified using recursive partitioning analysis to determine the greatest differences in overall survival.¹⁶ The threshold was validated by iterative log-rank tests for overall survival at increasing TTI thresholds of 5 days, and the TTI with the most significant P value was selected.

The primary outcomes of this study were overall survival and recurrence (locoregional recurrence, distant recurrence, or never disease free). Time to recurrence was not available for all patients, which precluded recurrence-based survival analysis. Patient characteristics analyzed included the following demographic factors: age, sex, race/ethnicity, language spoken, insurance type, body mass index (BMI), marital status, and socioeconomic status (SES) score. The SES score was stratified into 4 quartiles determined by the patient’s address, which was geocoded to a census block, with a neighborhood SES score calculated using the method outlined by Diez Roux et al.¹⁷ The SES measurement reflects the number of standard deviations from the mean SES score of New York State census blocks. Also analyzed were initial diagnosis at a different institution, Charlson-Deyo comorbidity score (0, 1, or ≥2), tobacco use, primary tumor site, p16 status for oropharyngeal cancers (based on immunohistochemistry), TNM stage, primary treatment modality (surgery, RT/CRT, or surgery plus RT/CRT), and treatment completion. A cohort building tool (Looking Glass Analytics; Streamline Health) was used to assist in the acquisition of many of these variables.

Differences in patient characteristics based on the optimal TTI threshold were assessed by the χ² test to identify factors associated with treatment delay. To evaluate associations between TTI delay and survival, we generated Kaplan-Meier plots and constructed Cox proportional hazards regression models, adjusting for strong relevant indicators (P < .05) and potential confounders. The potential for confounding was examined for nonsignificant covariates using a change point estimate criterion.¹⁸ The proportional hazards assumptions were tested for using log-log plots and Schoenfeld residuals, which revealed nonproportionality at 2 months. To address this and achieve proportional hazards, we assessed time-varying interactions and stratified analyses, excluding patients with less than 2 months of follow-up time. Sensitivity analyses were performed with and without exclusion of patients with less than 2 months of follow-up time to assess for bias.

Multivariable logistic regression models were used to identify (1) significant predictors of delayed TTI (based on the optimal threshold) and (2) significant predictors of disease recurrence. Covariates included in the multivariable logistic regression models were based on statistical significance (2-sided P < .05) in univariate analyses. Potential for confounding was also examined for the recurrence model using a change point estimate criterion for TTI. Differential associations for specific subgroups (eg, by primary tumor site, stage, and primary treatment modality) were tested for by including cross-product terms in the multivariable regression models. Adjusted hazard ratios (HRs) and odds ratios (ORs), along with 95% CIs, were generated from the multivariable regression models using SPSS (version 25.0; IBM Corp) and R (version 1.1.456; The R Foundation).

Last, the medical records of patients with TTI in the top 15th percentile (corresponding to delays of >68 days) were reviewed to identify reasons for prolonged TTI. Only medical records with readily available otolaryngology and radiation oncology notes were reviewed. Treatment delay reasons were categorized based on patterns identified in patient medical record review and summarized qualitatively.

Results

Population Characteristics

After applying exclusion criteria, a total of 956 patients with HNSCC (mean [SD] age, 60.8 [18.2] years; 72.6% male) were included, with a median follow-up of 32 months (interquartile range [IQR], 14-70 months). The median TTI was 40 days (IQR, 28-56 days). Recursive partitioning analysis and log-rank tests demonstrated that an optimal TTI threshold to differentiate overall survival was greater than 60 days. One-fifth of patients (199 of 956 [20.8%]) in our cohort had TTI exceeding 60 days and were classified as having delayed treatment.

Patient characteristics are summarized in Table 1. The most common primary tumor sites were larynx (332 of 956 [34.7%]) and oropharynx (289 of 956 [30.2%]). Most patients were diagnosed at American Joint Committee on Cancer (AJCC) stage IV (483 of 956 [50.5%]). Compared with patients with TTI of 60 days or less, a greater proportion of patients with TTI exceeding 60 days were African American (31.2% [62 of 199] vs 21.8% [165 of 757]) and Hispanic (22.6% [45 of 199] vs 17.8% [135 of 757]). Patients with TTI exceeding 60 were also more likely to have Medicaid insurance (38.2% [76 of 199] vs 24.6% [186 of 757]), BMI (calculated as weight in kilograms divided by height in meters squared) less than 18.5 (14.6% [29 of 199] vs 8.3% [63 of 757]), an SES score in the bottom quartile (30.2% [60 of 199] vs 21.9% [166 of 757]), and initial diagnosis at a different institution (42.2% [84 of 199] vs 30.6% [232 of 757]).

Table 1. Patient Characteristics Based on TTI Threshold of 60 Days.

Variable	Total, No. (N = 956)	Patients by TTI, No. (%)		χ2 Statistic
		≤60 d (n = 757)	>60 d (n = 199)
Age, y
<50	101	79 (10.4)	22 (11.1)	1.89
50-59	279	219 (28.9)	60 (30.2)
60-69	294	228 (30.1)	66 (33.2)
≥70	282	231 (30.5)	51 (25.6)
Sex
Male	694	548 (72.4)	146 (73.4)	0.08
Female	262	209 (27.6)	53 (26.6)
Race/ethnicity
African American	227	165 (21.8)	62 (31.2)	15.01
Asian	11	8 (1.1)	3 (1.5)
White	277	236 (31.2)	41 (20.6)
Hispanic	180	135 (17.8)	45 (22.6)
Unknown/other	261	213 (28.1)	48 (24.1)
Language spoken
English	765	607 (80.2)	158 (79.4)	0.70
Non-English	160	124 (16.4)	36 (18.1)
Unknown	31	26 (3.4)	5 (2.5)
Insurance type
Medicare	338	274 (36.2)	64 (32.2)	25.13
Medicaid	262	186 (24.6)	76 (38.2)
Commercial	200	174 (23.0)	26 (13.1)
Other	104	88 (11.6)	16 (8.0)
None	22	15 (2.0)	7 (3.5)
Unknown	30	20 (2.6)	10 (5.0)
BMI
>30	178	147 (19.4)	31 (15.6)	8.58
18.5-30	592	476 (62.9)	116 (58.3)
<18.5	92	63 (8.3)	29 (14.6)
Unknown	94	71 (9.4)	23 (11.6)
Marital status
Married	309	257 (33.9)	52 (26.1)	7.77
Unmarried	616	472 (62.4)	144 (72.4)
Unknown	31	28 (3.7)	3 (1.5)
Socioeconomic status score
Less than −5.60, bottom quartile	226	166 (21.9)	60 (30.2)	10.45
−5.60 to −2.17	227	176 (23.2)	51 (25.6)
Greater than −2.17 to −0.57	227	183 (24.2)	44 (22.1)
Greater than −0.57	226	187 (24.7)	39 (19.6)
Unknown	50	45 (5.9)	5 (2.5)
Initial diagnosis at a different institution
No	640	525 (69.4)	115 (57.8)	9.52
Yes	316	232 (30.6)	84 (42.2)
Charlson-Deyo comorbidity score
0	829	654 (86.4)	175 (87.9)	0.74
1	16	12 (1.6)	4 (2.0)
≥2	111	91 (12.0)	20 (10.1)
Tobacco use
Current	373	291 (38.4)	82 (41.2)	4.43
Previous	193	163 (21.5)	30 (15.1)
Never	368	287 (37.9)	81 (40.7)
Unknown	22	16 (2.1)	6 (3.0)
Primary tumor site				4.64
Lip/oral cavity	202	162 (21.4)	40 (20.1)
Oropharynx	289	219 (28.9)	70 (35.2)	2.09a
p16+	116	90	26
p16−	57	39	18
Unknown	116	90	26
Hypopharynx	63	49 (6.5)	14 (7.0)
Larynx	332	272 (35.9)	60 (30.2)
Nasopharynx	40	33 (4.4)	7 (3.5)
Nose/sinus	30	22 (2.9)	8 (4.0)
AJCC stage
I	139	113 (14.9)	26 (13.1)	0.78
II	127	99 (13.1)	28 (14.1)
III	186	149 (19.7)	37 (18.6)
IV	483	379 (50.1)	104 (52.3)
Unknown	21	17 (2.2)	4 (2.0)
T
T1	221	176 (23.2)	45 (22.6)	1.53
T2	288	230 (30.4)	58 (29.1)
T3	202	158 (20.9)	44 (22.1)
T4	209	165 (21.8)	44 (22.1)
Unknown	36	28 (3.7)	8 (4.0)
N
N0	408	330 (43.6)	78 (39.2)	4.87
N1	131	104 (13.7)	27 (13.6)
N2	368	288 (38.0)	80 (40.2)
N3	29	20 (2.6)	9 (4.5)
NX	11	7 (0.9)	4 (2.0)
Unknown	9	8 (1.1)	1 (0.5)
Primary treatment modality
Surgery alone	218	179 (23.6)	39 (19.6)	2.54
Surgery plus RT/CRT	213	172 (22.7)	41 (20.6)
RT/CRT alone	525	406 (53.6)	119 (59.8)
Treatment completion
Yes	691	556 (73.4)	135 (67.8)	2.54
No	44	34 (4.5)	10 (5.0)
Unknown	221	167 (22.1)	54 (27.1)

Abbreviations: AJCC, American Joint Committee on Cancer; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); CRT, chemoradiotherapy; RT, radiotherapy; TTI, time to treatment initiation.

^aBased on χ² test of p16 status among patients with oropharyngeal cancer.

Estimated TTI thresholds varied somewhat between primary treatment modality subgroups. Cutoff thresholds for patients with RT/CRT alone and surgery alone were 56 days and 62 days, respectively, whereas the threshold for patients receiving surgery plus RT/CRT was 71 days. The median TTI was 42 days (IQR, 31-58 days) for patients who underwent RT/CRT alone, 37 days (IQR, 26-53 days) for patients who underwent surgery alone, and 35 days (IQR, 23-53 days) for patients who underwent surgery plus RT/CRT. Comparisons of the TTI thresholds and medians across treatment modality subgroups are shown in eFigure 2 in the Supplement.

Association of TTI With Overall Survival and Recurrence

The 5-year overall survival was 60.9% (653 of 956) for all patients, 47.0% (115 of 199) for patients with TTI exceeding 60 days, and 64.5% (538 of 757) for patients with TTI of 60 days or less. Kaplan-Meier survival analysis demonstrated a difference in the survival distributions between patients with delayed TTI exceeding 60 days and TTI of 60 days or less (Figure), with differences emerging between 18 and 24 months (log-rank P < .001). A strong association remained between overall survival and delayed TTI (>60 vs ≤60 days) after adjusting for all relevant factors (HR, 1.58; 95% CI, 1.20-2.00).

Figure. Kaplan-Meyer Survival Estimates Based on TTI Threshold of 60 Days.

Log-rank (Mantel-Cox) χ² = 17.11; P < .001. The unadjusted hazard ratio for TTI exceeding 60 days is 1.632 (95% CI, 1.290-2.066). The adjusted hazard ratio for TTI exceeding 60 days is 1.693 (95% CI, 1.318-2.176). TTI indicates time to treatment initiation.

Compared with the rest of the cohort, the subset of 19 patients with less than 2 months of follow-up contained a larger portion that was 70 years or older (57.9% [11 of 19] vs 28.9% [271 of 937]) and underwent surgery alone as treatment (57.9% [11 of 19] vs 22.1% [207 of 937]). Sensitivity analysis was performed that excluded these patients to address proportional hazards, which revealed an even stronger association (HR, 1.69; 95% CI, 1.32-2.18) (Table 2 and eTable 4 in the Supplement).

Table 2. Multivariable Cox Proportional Hazards Regression Model for Overall Survival and Multivariable Logistic Regression Model for Recurrence.

Variable	Cox Proportional Hazards Regression Model for Overall Survival, HR (95% CI)a	Logistic Regression Model for Recurrence, OR (95% CI)b
TTI>60 d	1.693 (1.318-2.176)	1.772 (1.071-2.932)
Age, yc
<50	1 [Reference]	NA
50-59	0.999 (0.623-1.602)	NA
60-69	1.515 (0.952-2.411)	NA
≥70	2.060 (1.296-3.276)	NA
Race/ethnicity
African American	1 [Reference]	1 [Reference]
Asian	1.735 (0.666-4.523)	0.786 (0.128-4.819)
White	1.337 (0.983-1.818)	0.828 (0.474-1.447)
Hispanic	0.615 (0.399-0.947)	0.969 (0.514-1.826)
Unknown/other	1.269 (0.921-1.748)	0.761 (0.421-1.378)
Language spoken
Non-English	1 [Reference]	1 [Reference]
English	0.679 (0.497-0.928)	0.552 (0.308-0.990)
Unknown	0.933 (0.549-1.585)	1.735 (0.428-7.031)
Insurance typec
Commercial	NA	1 [Reference]
Medicare	NA	1.007 (0.570-1.778)
Medicaid	NA	1.074 (0.584-1.975)
Other	NA	1.331 (0.624-2.840)
None	NA	3.931 (0.727-21.255)
Unknown	NA	1.281 (0.328-5.009)
Charlson-Deyo comorbidity scorec
0	1 [Reference]	NA
1	1.610 (0.783-3.309)	NA
≥2	1.700 (1.235-2.340)	NA
Tobacco usec
Current	1 [Reference]	NA
Previous	0.838 (0.656-1.071)	NA
Never	0.535 (0.367-0.780)	NA
Unknown	1.442 (0.802-2.593)	NA
Primary tumor site
Larynx	1 [Reference]	1 [Reference]
Lip/oral cavity	0.852 (0.607-1.196)	1.059 (0.603-1.858)
Oropharynx	0.650 (0.478-0.885)	0.604 (0.352-1.038)
Hypopharynx	1.241 (0.821-1.877)	0.740 (0.301-1.815)
Nasopharynx	1.165 (0.645-2.103)	4.485 (1.038-19.383)
Nose/sinus	1.527 (0.816-2.859)	0.972 (0.339-2.785)
Final stage
I	1 [Reference]	1 [Reference]
II	1.413 (0.874-2.284)	1.594 (0.719-3.533)
III	1.414 (0.905-2.210)	2.278 (1.061-4.893)
IV	2.620 (1.726-3.976)	7.591 (3.654-15.773)
Unknown	3.780 (1.539-9.284)	5.369 (1.364-21.123)
Primary treatment modality
RT/CRT alone	1 [Reference]	1 [Reference]
Surgery alone	0.688 (0.483-0.979)	0.139 (0.078-0.245)
Surgery plus RT/CRT	0.741 (0.551-0.995)	0.124 (0.073-0.211)
Treatment completion
No	1 [Reference]	1 [Reference]
Yes	0.384 (0.242-0.609)	0.451 (0.158-1.290)
Unknown	0.809 (0.500-1.311)	0.705 (0.233-2.133)

Abbreviations: AJCC, American Joint Committee on Cancer; CRT, chemoradiotherapy; HR, hazard ratio; NA, not applicable; OR, odds ratio; RT, radiotherapy; TTI, time to treatment initiation.

^aThe final Cox proportional hazards regression model for overall survival was adjusted for age, race/ethnicity, language spoken, Charlson-Deyo comorbidity score, tobacco use, primary tumor site, AJCC stage, and treatment completion.

^bThe final logistic regression model for recurrence was adjusted for race/ethnicity, insurance type, marital status, socioeconomic status score, and initial diagnosis at a different institution.

^cCovariate of insurance type was only in the Cox proportional hazards regression model for overall survival. Covariates of age, Charlson-Deyo comorbidity score, and tobacco use were only in the logistic regression model for recurrence.

The association between delayed TTI and overall survival within categories of primary tumor site, final stage, and primary treatment modality was examined (eTable 5 in the Supplement). The association was clinically meaningful for all 3 primary treatment modalities, with large effect sizes. For surgery alone, the HR was 2.15 (95% CI, 1.13-4.07). For surgery plus RT/CRT, the HR was 1.83 (95% CI, 0.95-3.54). For RT/CRT alone, the HR was 1.38 (95% CI, 0.99-1.92). In addition, delayed TTI was associated with clinically meaningful associations with survival for patients with laryngeal cancer (HR, 2.20; 95% CI, 1.39-3.49) and patients with oropharyngeal cancer (HR, 1.97; 95% CI, 1.18-3.28), as well as for both local-stage cancers (HR, 1.94; 95% CI, 1.08-3.50) and advanced-stage cancers (HR, 1.47; 95% CI, 1.12-1.96).

Because overall survival was not cancer specific, we also analyzed the association of delayed TTI with recurrence (Table 2). Time to recurrence was not available for all patients, so binary logistic regression was used to identify predictive factors. A total of 344 patients whose recurrence status was unknown were excluded from this analysis. Adjusting for other relevant factors, the likelihood of recurrence was higher among patients with delayed TTI (>60 vs ≤60 days; OR, 1.77; 95% CI, 1.07-2.93). Other important predictors included language spoken, primary tumor site, final stage, and primary treatment modality. No interactions between TTI and other predictors were identified for recurrence.

Reasons for Delayed TTI

We explored factors predictive of delayed TTI exceeding 60 days by multivariable logistic regression (Table 3); the full regression model is summarized in eTable 6 in the Supplement). Among the strongest predictors was initial diagnosis at the same institution as treatment, which decreased the odds of TTI delay by almost half (OR, 0.53; 95% CI, 0.37-0.76). Compared with commercial insurance, Medicaid insurance increased the odds of delay more than 2-fold (OR, 2.17; 95% CI, 1.28-3.66). Patients of white race/ethnicity (vs African American) had a lower likelihood of delayed TTI by an OR of 0.53 (95% CI, 0.33-0.85). Compared with BMI of 18.5 to 30, BMI less than 18.5 was associated with an increased odds of delayed TTI (OR, 1.74; 95% CI, 1.05-2.89).

Table 3. Multivariable Logistic Regression Model to Examine Factors That Can Potentially Predict Delayed TTI Based on the Optimal Threshold of 60 Days^a.

Variable	OR (95% CI)
Not initially diagnosed at a different institution vs initially diagnosed at a different institution	0.529 (0.369-0.757)
Medicaid insurance vs commercial insurance	2.168 (1.283-3.663)
White race/ethnicity vs African American	0.525 (0.326-0.845)
BMI<18.5 vs 18.5-30	1.743 (1.051-2.889)

Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); OR, odds ratio; TTI, time to treatment initiation.

^aOnly statistically significant predictors are listed. The full regression model is summarized in eTable 6 in the Supplement. The final model was adjusted for race/ethnicity, insurance type, marital status, socioeconomic status score, and initial diagnosis at a different institution.

To identify reasons behind delayed TTI, medical records of patients in the top 15th percentile of TTI were reviewed (Table 4). Of the 66 reviewed patients, 17 (25.8%) had no identified reasons for treatment delay. Among the remainder, the 2 most common reasons were missed appointments leading up to the initial treatment (14 of 66 [21.2%]) and extensive pretreatment evaluation (14 of 66 [21.2%]), which included additional visits to obtain medical clearance for treatment, numerous dental visits and tooth extractions before treatment, and workup of an unknown primary neck mass. Also common was treatment refusal (9 of 66 [13.6%]). Other identified reasons for treatment delay included unrelated medical reason (n = 6), seeking a second opinion (n = 3), a cancer-related medical reason (eg, airway obstruction due to the cancer) (n = 2), and lack of insurance (n = 1).

Table 4. Reasons for Treatment Delay.

Variable	No. (%) of Patients
No.	66
Missed appointment	14 (21.2)
Extensive pretreatment evaluation	14 (21.2)
Treatment refusal	9 (13.6)
Unrelated medical reason	6 (9.1)
Second opinion	3 (4.5)
Cancer-related medical reason	2 (3.0)
Lack of insurance	1 (1.5)
Unknown	17 (25.8)

Discussion

We investigated the consequences of and reasons for delayed TTI in patients with HNSCC treated at a large urban academic institution among an underserved population in New York City. When controlling for other factors, TTI exceeding 60 days was found to be associated with decreased overall survival and increased recurrence. Using regression analysis, we identified predictors of delayed TTI, which included African American race/ethnicity, Medicaid insurance, being underweight (BMI<18.5), and initial diagnosis at a different institution. Patient medical records were then reviewed to identify likely reasons for treatment delay.

To date, most studies of delayed TTI for patients with head and neck cancer in the United States have used national cancer registries to illustrate the negative consequences of treatment delay on patient outcomes.⁵ Although national cancer registries provide large sample sizes for statistical power, they contain a heterogeneous patient population and lack information on recurrence and reasons for treatment delay. It is also possible that the findings from national cohorts do not correspond with the experience of individual institutions, as was the case in the Netherlands.^11,14 Therefore, we sought to validate results from national cancer registries in our unique population that included patients from an underserved area of New York City, as well as to identify predictors of and reasons for treatment delay to facilitate future intervention.

Based on our calculated threshold of 60 days, 199 (20.8%) of our patients experienced treatment delay. This is twice the rate seen nationally¹⁹ and may reflect the underserved or more vulnerable characteristics of our patient population compared with the rest of the country. Setting a single cutoff to define treatment delay is useful for establishing a benchmark for quality of care and for determining what constitutes a tolerable amount of delay. However, we found that optimal thresholds varied for different treatment modalities (surgery alone, surgery plus RT/CRT, and RT/CRT alone), which suggests that the safest amount of TTI may depend on the planned therapy. However, despite a range in threshold estimates of 56 days (RT/CRT alone) to 71 days (surgery plus RT/CRT), the median TTI across treatment modalities remained constant at between 35 and 42 days.

Some studies that have failed to show an association between TTI and survival examined only oral cavity cancers^2,10 or only locoregionally advanced tumors (AJCC stage III/IV).²⁰ We performed subgroup analyses in these particular subsets of patients with HNSCC to explore the findings further in our population. In subgroup analysis by primary tumor site, delayed TTI was found not to be associated with overall survival in patients with lip/oral cavity cancer. This contrasted with prior studies^21,22,23 that were based on Taiwanese populations and used much lower thresholds of 20 days. We found that delayed TTI was associated with overall survival in patients with oropharyngeal cancer and patients with laryngeal cancer, which is consistent with other TTI studies.^7,8,9,24,25 Further assessment herein of human papillomavirus–positive oropharyngeal cancers revealed a similar association with delayed TTI; however, the number of cases tested for p16 was small, and the association estimate was imprecise.

Although delayed TTI was a predictor of overall survival irrespective of tumor stage, the HR for delayed TTI was greater in local-stage cancer than in advanced-stage cancer (1.94 vs 1.47). This finding is consistent with national observations¹⁹ and may be due to stage migration¹ as a result of upstaging during wait times. Moreover, the negative consequences of delayed treatment in local-stage cancers could be exacerbated by longer surgical wait times due to the prioritization of treatment in advanced-stage cancers.²⁶ A pattern of longer wait times in local-stage vs advanced-stage cancers was not evident in our cohort based on the distribution of delayed TTI.

Delayed TTI was also found to be associated with greater risk of tumor recurrence, an outcome that has not been studied to date in national cancer registry cohorts. While delayed initiation of RT is associated with an increased local recurrence rate,²⁷ institutional studies^2,20 of treatment delay overall failed to find an association. However, delayed interval time between surgery and postoperative RT has been associated with worse recurrence-free survival in several studies.^2,27,28

There has been less research on the predictors of and reasons for treatment delay in head and neck cancer, especially among underserved and minority populations. We found that African American race/ethnicity, Medicaid insurance, being underweight, and initial diagnosis at a different institution were associated with delayed TTI. This provides perspective on the type of patients with HNSCC who are most at risk of treatment delay, potentially allowing for targeted intervention. Other studies^{4,6,8,12,13,29,30} have also linked TTI delay in HNSCC with race/ethnicity, Medicaid insurance, and initial diagnosis at a different institution. A possible culprit for patients with Medicaid insurance may be decreased physician participation, limiting access or creating enrollment-related delays before treatment.³⁰ Another potential driver of delayed TTI at academic institutions like ours may be from transition of care for outside referrals.⁴

Further exploring the reasons for treatment delay in the top 15th percentile of TTI in our cohort, we profiled the mechanisms underlying significant treatment delay. Among the most common reasons for delay were missed appointments, extensive pretreatment evaluation, and treatment refusal by the patient. Extensive pretreatment evaluation consisted primarily of dental appointments; other patient-related factors included workup of unknown primary tumors or procedures that were postponed by the requisite imaging and diagnostic techniques. Seeking a second opinion, medical reasons related and unrelated to the primary cancer, and lack of insurance were less common reasons for delay. It is likely that the drivers of treatment delay identified in our analyses were interwoven with the identified reasons for delay. For instance, African American patients with breast cancer have been found to experience treatment delay and discontinuation more frequently than white patients due to missed appointments.³¹ Better understanding these reasons for delay and implementing solutions warrant a quality improvement approach that involves establishing a timeline before treatment, as well as interviewing and educating patients and health care providers. Such approaches have been applied to reduce intervals from surgery to adjuvant RT in HNSCC.²

Limitations

Our study has some limitations. Because this was a retrospective study, it was limited by the accuracy of the cancer registry and the patient medical records, where miscoding errors are concerns. Given that this was a single-institution study, our sample size was smaller than that of the national studies.⁵ Another limitation was missing data, which we opted not to impute given required assumptions for missing at random. Most of the missing information (eg, p16 status) came from older records that predated systematic testing of oropharyngeal cancer. A subset of 19 patients was also excluded to satisfy the assumption of proportional hazards, which could have introduced some bias. However, sensitivity analysis showed that exclusion of this subset of patients did not alter our findings. Finally, absence of recurrence details on many patients precluded time-to-event analyses. Logistic regression was deemed appropriate because most recurrence occurs within 1 to 2 years. Ultimately, our findings will need to be validated prospectively. In addition, time from surgery to postoperative RT and treatment package time may need to be analyzed in conjunction with TTI because a recent study³² found that the 2 variables outweighed TTI in importance in determining overall survival.

Conclusions

This study demonstrates that delayed TTI at a threshold of 60 days was associated with decreased overall survival and increased HNSCC recurrence. Identifying predictive factors and reasons for treatment delay can assist in recognizing at-risk patients with HNSCC and pinpoint specific areas to reduce delay. These initiatives will be especially prudent at medical centers that care for a largely urban and underserved community in which most patients are covered by public insurance.

Supplement.

eFigure 1. Diagram of Cohort Creation With Exclusion Criteria

eFigure 2. Box and Whisker Plots of TTI for Different Treatment Modalities

eTable 1. Hazard Ratios for TTI>60 (vs TTI≤60) in Subgroup Analysis of Primary Site, Treatment Modality, and Final Stage

eTable 2. Full Multivariate Logistic Regression to Examine Significant Factors That Could Potentially Predict Delayed TTI Based on the Optimal Threshold of 60 Days

eTable 3. Patient Characteristics of Patients With TTI = 0 That Were Excluded Due to Miscoding Compared to Remaining Cohort

eTable 4. Sensitivity Analysis for Exclusion of TTI = 0

eTable 5. Sensitivity Analysis for Exclusion of Patients With Follow-up ≤2 Months

eTable 6. Full Multivariate Logistic Regression to Examine Significant Factors That Could Potentially Predict Delayed TTI Based on the Optimal Threshold of 60 Days