Insurance Approval for Proton Beam Therapy and its Impact on Delays in Treatment

Abstract

Purpose:

Prior authorization (PA) has been widely implemented for proton beam therapy (PBT). We sought to determine the association between PA determination and patient characteristics, practice guidelines, and potential treatment delays.

Methods and Materials:

A single-institution retrospective analysis was performed of all patients considered for PBT between 2015 and 2018 at a National Cancer Institute–designated Comprehensive Cancer Center. Differences in treatment start times and denial rates over time were compared, and multivariable logistic regression was used to identify predictors of initial denial.

Results:

A total of 444 patients were considered for PBT, including 396 adult and 48 pediatric patients. The American Society for Radiation Oncology model policy supported PBT coverage for 77% of the cohort. Of adult patients requiring PA, 64% were initially denied and 32% remained denied after appeal. In patients considered for reirradiation or randomized phase 3 trial enrollment, initial denial rates were 57% and 64%, respectively. Insurance coverage was not related to diagnosis, reirradiation, trial enrollment, or the American Society for Radiation Oncology model policy guidelines, but it was related to insurance category on multivariable analysis (P < .001). Over a 3-year timespan, initial denial rates increased from 55% to 74% (P = .034). PA delayed treatment start by an average of 3 weeks (and up to 4 months) for those requiring appeal (P < .001) and resulted in 19% of denied patients abandoning radiation treatment altogether. Of pediatric patients, 9% were initially denied, all of whom were approved after appeal, and PA requirement did not delay treatment start (P = .47).

Conclusions:

PA requirements in adults represent a significant burden in initiating PBT and cause significant delays in patient care. Insurance approval is arbitrary and has become more restrictive over time, discordant with national clinical practice guidelines. Payors and providers should seek to streamline coverage policies in alignment with established guidelines to ensure appropriate and timely patient care. © 2018 Elsevier Inc. All rights reserved.

Summary

This study is the first to comprehensively assess the state of prior authorization for proton beam therapy (PBT). The majority of adults who require prior authorization are initially denied and thus experience significant care delays. We found no clinical characteristics that predicted insurance approval for PBT. Insurers have not provided coverage for patients on randomized clinical trials to generate high-quality evidence for PBT and have not adopted clinical practice guidelines to guide their coverage decisions.

Introduction

Prior authorization (PA) is a cost-containment strategy employed by payors to allow coverage evaluation of medical services and thus control health care utilization and spending.¹ However, studies have found that PA can inappropriately restrict medical care, resulting in greater long-term morbidity and costs.^2,3 PA may also consume a significant expenditure of time and resources for health care providers. In a recent American Medical Association survey, 86% of physicians reported that PA places an increasingly high burden on their practice, with physicians and staff spending an average of 14.6 hours per week solely on PA requirements.⁴

PA has been widely implemented by insurers and their subsidiary radiation oncology benefit managers for proton beam therapy (PBT). PBT is an advanced radiation treatment (RT) modality that has become widely available over the past decade: The number of operational PBT facilities in the United States expanded from 3 in 2005 to 28 currently.⁵ The favorable physical properties of PBT allow highly conformal dose-escalated treatment with less integral dose to normal organs, theoretically leading to improved tumor control with reduced rates of short- and long-term toxicity and secondary malignancy.^6,7 However, PBT is also significantly more expensive than other types of RT in light of the immense costs of building and operating particle accelerators.⁸ Commercial payors are reluctant to share the large upfront costs of PBT and as such have enacted PA policies restricting the use of PBT.⁹ In contrast, patients with Medicare insurance are able to access PBT with clear provisions on best use governed by local coverage determinations; however, patients on managed care plans administered by commercial payors are still subject to PA policies.

A recent report by the Alliance for Proton Therapy Access found that 63% of PA requests for PBT were initially denied by commercial insurers; after appeal, 42% remained denied.⁹ Given the variability in PA requirements and the current lack of randomized level 1 evidence supporting PBT, the American Society for Radiation Oncology (ASTRO) released a model policy to guide the appropriate use and coverage of PBT.¹⁰ ASTRO also highlighted the burden of PA on radiation oncology practices as one of its 2018 legislative priorities for congressional reform.¹¹

To date, there has been no objective assessment of patient characteristics associated with insurance denial for PBT or the impact of the PA process on PBT delivery, particularly at a high-volume National Cancer Institute–designated Comprehensive Cancer Center. We therefore sought to evaluate the time course and outcomes of PA requests as they relate to clinical practice guidelines for PBT and the resulting delays in treatment caused by prolonged PA determinations.

Methods and Materials

Study cohort

All patients who were treated with or considered for PBT at our institution between May 1, 2015 and April 30, 2018 were included in this institutional review board–approved retrospective analysis. For those who did not receive PBT, inclusion criteria to qualify as being “considered for” PBT included administrative staff initiating PA for PBT and/or patients undergoing computed tomography (CT) simulation specifically for PBT planning. All patients who were considered for PBT were reviewed at a weekly departmental conference to determine collective appropriateness for PBT.

Baseline variables

We abstracted patients’ baseline data, including age, sex, treatment site, diagnosis, year of diagnosis, history of RT, clinical trial enrollment, and insurance category. Age was categorized as adult (>21 years old) or pediatric (≤21 years old). History of RT was only included if it overlapped with the site that was currently being considered for treatment (reirradiation setting). In addition to successfully enrolled patients, the clinical trial category included patients for whom enrollment was attempted but insurance denial for PBT precluded eligibility. In adults, only clinical trials that were randomized phase 3 trials of PBT versus photon RT were included in the clinical trial category.

Group category as described by ASTRO’s PBT model policy was also determined for each patient. Group 1 consists of sites that frequently support the use of PBT based on published clinical data and medical necessity and includes pediatric cancers, central nervous system cancers, head and neck cancers, primary liver cancers (hepatocellular carcinoma and cholangiocarcinoma), retroperitoneal sarcomas, and reirradiation cases.¹⁰ Group 2 covers all other disease sites that are recommended for PBT coverage when treated as part of a clinical trial or multi-institutional patient registry.¹⁰ Cases that did not fall into Group 1 or 2 were classified as “other.”

Outcome variables

Outcome variables included treatment modality, presence of PA requirement, PA determination, and time to authorization and treatment start. Treatment modality included PBT, photon RT, or no RT (patient either received surgery or chemotherapy or deferred all treatment). If PA was required, PA determination was categorized as approved upfront or initially denied. Those who were initially denied were further categorized as approved after appeal or ultimately denied, depending on the outcome of the appeals process. The total number of appeals for each patient was also recorded. Time to authorization was calculated as the number of days between the start date of the authorization process and receipt date of the final PA determination, to capture the entire time interval of the PA process. Time to start was calculated as the number of days between CT simulation date and treatment start date (whether PBT or photon RT); however, if the PA process delayed scheduling of CT simulation, number of days between authorization start date and treatment start date was used to calculate time to start.

Statistical analysis

Descriptive statistics were computed for the entire study cohort and for PBT and non-PBT groups separately. We used χ² testing was used to examine differences in receipt of PBT by baseline characteristics and in PA determination by year. Differences in time to treatment start were compared using 2-sample uneven variance t tests or Kruskal-Wallis testing. Post hoc Dunn’s test with Bonferroni correction was used to determine significant pairwise differences. Univariable and multivariable logistic regression analyses were conducted to evaluate predictors of initial denial. A significance level of P < .05 was used. All analyses were performed using STATA 14.0 (StataCorp, College Station, TX).

Results

Baseline characteristics

Table 1 presents baseline characteristics of those treated and not treated with PBT. A total of 444 patients were considered for PBT, of whom 300 (68%) were treated with PBT, 114 (26%) were treated with photon RT, and 30 (7%) were not treated with any RT. Adult patients represented 89% of the study cohort (396 patients) and had a mean age of 56.9 years (range, 22–92 years); 254 (64%) were treated with PBT. Pediatric patients represented 11% of the study cohort (48 patients) and had a mean age of 13.2 years (range, 2–21 years); 46 (96%) were treated with PBT. Thirty-eight patients, including 36 adult and 2 pediatric patients, were treated with photon RT because of its favorable dosimetry rather than insurance noncoverage (33% of photon RT patients).

Table 1.

Baseline characteristics of patients considered for proton beam therapy

Characteristic	Patients treated with PBT, n (%)	Patients not treated with PBT, n (%)	Total number of patients, n (%)	P value
Total	300	144	444
Age, y				<.001*
Adult (>21)	254 (84.7)	142 (98.6)	396 (89.2)
Pediatric (≤21)	46 (15.3)	2 (1.4)	48 (10.8)
Sex				.32
Male	163 (54.3)	71 (49.3)	234 (52.7)
Female	137 (45.7)	73 (50.7)	210 (47.3)
Treatment site				.13
Thorax	100 (33.3)	51 (35.4)	151 (34.0)
Central nervous system	67 (22.3)	25 (17.4)	92 (20.7)
Head and neck	54 (18.0)	17 (11.8)	71 (16.0)
Pelvis	45 (15.0)	25 (17.4)	70 (15.8)
Abdomen	34 (11.3)	26 (18.1)	60 (13.5)
Diagnosis				.034*
Other	84 (28.0)	43 (29.9)	127 (28.6)
Breast cancer	44 (14.7)	30 (20.8)	74 (16.7)
Low- or high-grade glioma	31 (10.3)	16 (11.1)	47 (10.6)
Benign histology	33 (11.0)	10 (6.9)	43 (9.7)
Sarcoma	31 (10.3)	7 (4.9)	38 (8.6)
Prostate cancer	22 (7.3)	10 (6.9)	32 (7.2)
Primary liver cancer	14 (4.7)	16 (11.1)	30 (6.8)
Lymphoma	24 (8.0)	5 (3.5)	29 (6.5)
Non-small cell lung cancer	17 (5.7)	7 (4.9)	24 (5.4)
ASTRO PBT Model Policy				.002*
Group 1	202 (67.3)	83 (57.6)	285 (64.2)
Group 2	27 (9.0)	30 (20.8)	57 (12.8)
Other	71 (23.7)	31 (21.5)	102 (23.0)
Reirradiation Setting				.98
Yes	69 (23.0)	33 (22.9)	102 (23.0)
No	231 (77.0)	111 (77.1)	342 (77.0)
Clinical trial enrollment				.18
Yes	22 (7.3)	16 (11.1)	38 (8.6)
No	278 (92.7)	128 (88.9)	406 (91.4)
Insurance category				.020*
Commercial/managed care	139 (46.3)	86 (59.7)	225 (50.7)
Medicare	93 (31.0)	28 (19.4)	121 (27.3)
Medicaid	51 (17.0)	27 (18.8)	78 (17.6)
Military	4 (1.3)	0 (0.0)	4 (0.9)
Charity care	13 (4.3)	3 (2.1)	16 (3.6)

Abbreviations: ASTRO = American Society for Radiation Oncology; PBT = proton beam therapy.

^*Statistically significant.

The most common diagnoses to be considered for PBT were breast cancer (17%), low- or high-grade glioma (11%), and benign histology (93% of which were in the central nervous system) (10%). Sixty-four percent met ASTRO group 1 criteria and 13% met group 2 criteria, for a total of 77% for whom the model policy would support PBT coverage. One hundred and two patients (23%) and 38 patients (9%) were considered for reirradiation and clinical trial enrollment, respectively. The most common insurance carriers were commercial payors (including managed care) (51%), followed by Medicare (27%) and Medicaid (18%).

Prior authorization outcomes

PA was required in 220 adult patients, including 81%, 92%, and 75% of patients with commercial, Medicaid, and military insurance, respectively. PA was required in 35 pediatric patients, including 69% and 100% of patients with commercial and Medicaid insurance, respectively. Patients with Medicare insurance or charity care did not require PA.

Figure 1A illustrates PA determination for adult versus pediatric patients. In the 197 adult patients for whom PA was completed, PA was approved upfront in 71 patients (36%) and initially denied in 126 patients (64%). Of these initially denied patients, 49% (62 patients) were approved after appeal and 51% (64 patients) were ultimately denied; thus, the ultimate denial rate was 32%. Of patients approved after appeal, 71%, 15%, 11%, and 3% were approved after first-, second-, third-, and fourth-level appeals, respectively. Of patients ultimately denied, in 27% no appeal was placed after the first denial (usually in the interest of time), and 36%, 28%, and 9% were repeatedly denied after first-, second-, and third-level appeals, respectively. Five patients progressed to external appeal through a state-mandated independent third-party reviewer, 3 of whom (60%) had their insurer’s denial overturned. Twelve patients who were ultimately denied did not receive any RT (19% of ultimate denials): 5 pursued other treatment (chemotherapy or surgery), 4 opted for no treatment after consideration of risks and benefits of nonproton RT, and 3 had disease progression during an extended appeals process.

Fig. 1.

Percentage of patients approved upfront versus initially denied by age category (A) and in adult patients by sex (B), treatment site (C), diagnosis (D), American Society for Radiation Oncology model policy group (E), reirradiation setting (F), clinical trial enrollment (G), and insurance category (H). Note: Medicare and charity care do not require prior authorization, and only 3 patients with military insurance required prior authorization. Thus, these categories were not included in (H).

Abbreviations: CNS = central nervous system; H&N = head and neck; NSCLC = non-small cell lung cancer.

In pediatric patients, PA was approved upfront in 32 patients (91%) and initially denied in 3 patients (9%) (18, 20, and 21 years old) and who were all subsequently approved after first- or second-level appeals. On univariable logistic regression, pediatric patients were 95% less likely to receive initial denial compared with adults (odds ratio [OR], 0.05; 95% confidence interval [CI], 0.02–0.18; P < .001).

Figure 1B to 1H illustrates PA determination by baseline characteristics in adults. The most common diagnoses to be initially denied were prostate cancer (80%), lymphoma (77%), breast cancer (69%), and primary liver cancer (69%). In patients with prostate cancer, 63% of those initially denied were considered for trial enrollment or reirradiation. In patients with lymphoma, 60% of those initially denied were ≤30 years old with mediastinal disease. Similarly, in patients with breast cancer, 66% of those initially denied had strong rationale for considering PBT, including trial enrollment, reirradiation, young age (≤40 years) and receipt of regional nodal irradiation, or new cardiac or pulmonary toxicity from chemotherapy. The ultimate denial rate for primary liver cancer was the second highest at 44% (after prostate cancer at 50%), despite it being supported as a group 1 designation by the ASTRO model policy.

ASTRO model policy group 1 patients had an initial denial rate of 60%, similar to other patients at 63%, but group 2 patients had a considerably higher rate of 81%. In patients considered for reirradiation or clinical trial enrollment, initial denial rates were 57% and 64%, respectively, and ultimate denial rates were 23% and 43%, respectively. In patients with commercial or Medicaid insurance, initial denial rates were 74% and 36%, respectively. Three of 4 patients with military insurance required PA, all of whom were approved upfront.

Table 2 presents odds of initial denial by baseline characteristics in adult patients. On univariable analysis, the only factors that predicted initial denial were ASTRO model policy group 2 versus group 1 (OR, 2.74; 95% CI, 1.05–7.16; P = .039) and Medicaid versus commercial insurance (OR, 0.20; 95% CI, 0.10–0.41; P < .001). On multivariable analysis, only insurance category remained a significant independent predictor, with Medicaid-insured adult patients 83% less likely to receive initial denial compared with commercially insured adult patients (OR, 0.17; 95% CI, 0.08–0.39; P < .001). There was no association between initial denial and sex, treatment site, diagnosis, reirradiation setting, or clinical trial enrollment.

Table 2.

Odds of receiving initial denial for proton beam therapy in adults

Characteristic	Unadjusted OR (95% CI)	P value	Adjusted OR (95% CI)	P value
Sex
Male	Reference		Reference
Female	1.64 (0.91–2.95)	.10	2.04 (0.91–4.57)	.09
Treatment site
Central nervous system	Reference		Excluded from model*
Head and neck	0.94 (0.38–2.34)	.90
Thorax	1.09 (0.50–2.37)	.83
Abdomen	1.42 (0.51–3.97)	.51
Pelvis	1.57 (0.51–4.85)	.43
Diagnosis
Low- or high-grade glioma	Reference		Reference
Non-small cell lung cancer	0.42 (0.08–2.20)	.30	0.45 (0.06–3.42)	.44
Breast cancer	1.54 (0.53–4.51)	.43	0.99 (0.20–4.98)	.99
Primary liver cancer	1.52 (0.39–5.91)	.54	3.12 (0.66–14.74)	.15
Prostate cancer	2.77 (0.47–16.21)	.26	2.17 (0.27–17.56)	.47
Sarcoma	0.46 (0.12–1.76)	.26	0.63 (0.14–2.87)	.55
Lymphoma	2.31 (0.49–10.82)	.29	3.04 (0.38–24.59)	.30
Benign histology	0.97 (0.30–3.14)	.96	0.96 (0.26–3.55)	.95
Other	1.58 (0.55–4.55)	.40	2.46 (0.67–9.12)	.18
ASTRO PBT Model Policy
Group 1	Reference		Reference
Group 2	2.74 (1.05–7.16)	.039‡	2.44 (0.43–13.84)	.31
Other	1.10 (0.53–2.28)	.81	1.04 (0.28–3.94)	.95
Reirradiation setting
Yes	Reference		Reference
No	1.45 (0.75–2.87)	.26	1.46 (0.52–4.14)	.47
Clinical trial enrollment
Yes	Reference		Reference
No	0.99 (0.32–3.06)	.98	2.81 (0.53–15.04)	.23
Insurance category
Commercial/managed care	Reference		Reference
Medicare	†		†
Medicaid	0.20 (0.10–0.41)	<.001‡	0.17 (0.08–0.39)	<.001‡
Military	0.05 (0.002–1.57)	.09	0.03 (0.0004–1.38)	.07
Charity care	†		†

Abbreviations: ASTRO = American Society for Radiation Oncology; CI = confidence interval; OR = odds ratio; PBT = proton beam therapy.

^*Treatment site variable excluded from multivariable analysis because of multicollinearity with diagnosis category.

^†Prior authorization not required.

^‡Statistically significant.

Figure 2 illustrates PA policies becoming significantly more restrictive over time in adults. From 2015 to 2017, upfront approval for PBT declined from 45% to 26%, and initial denial increased from 55% to 74% (P = .034).

Fig. 2.

Over the course of 3 years, prior authorization became significantly more restrictive in adult patients: Upfront approvals declined from 45% to 26%, and initial denials increased from 55% to 74% (P = .034), including increases in both approval after appeals (27% to 38%) and ultimate denials (29% to 35%).

Impact on treatment delays

The average time to authorization for adults requiring PA was 22.5 days. For patients approved upfront, approved after appeal, and ultimately denied, the average (range) was 13.4 days (0–50 days), 26.2 days (3–100 days), and 29.0 days (1–79 days), respectively. The average time to authorization for pediatric patients was 7.9 days, including 7.4 days (0–22 days) and 13.0 days (5–26 days) for patients approved upfront and after appeal, respectively.

Table 3 presents differences in time to treatment start as related to insurance characteristics. The average time to start for adults requiring PA was 33.8 days (range, 7–120 days), which was significantly longer than 20.6 days (range, 6–48 days) for those not requiring PA (P < .001). Start times were significantly different by PA determination and insurance category. On pairwise comparisons, average time to start was significantly longer for approved upfront, approved after appeal, and ultimately denied categories compared with no PA requirement (P < .001 for each comparison) and for patients with commercial compared with Medicare insurance (P < .001).

Table 3.

Association of insurance characteristics with time to treatment start

Characteristic	Adult			Pediatric
	No. of patients	Average days to start*	P value	No. of patients	Average days to start*	P value
Total	396	27.1		48	18.9
PA requirement			<.001†			.47
Required	220	33.8		35	19.7
Not required	176	20.6		13	16.8
PA determination			.002†			.89
Approved upfront	71	27.7		32	19.8
Approved after appeal	62	36.2		3	18.7
Ultimately denied	64	39.9		0	–
Insurance category			<.001†			.016†
Commercial/managed care	199	31.8		26	16.3
Medicare	121	20.5		0	–
Medicaid	61	25.5		17	22.2
Military	4	21.8		0	–
Charity care	11	25.8		5	21.0

Abbreviation: PA = prior authorization.

^*Average days to start includes patients who were treated with proton beam therapy or photon radiation treatment.

^†Statistically significant.

The average time to start for pediatric patients requiring PA was 19.7 days (range, 7–80 days), which was not significantly different than 16.8 days (range, 6–32 days) for those not requiring PA (P = .47). Start times were significantly different by insurance category; on pairwise comparison, average time to start was significantly shorter for patients with commercial compared with Medicaid insurance (P = .006).

Discussion

There are 4 principal findings from this single-institution study. First, PA resulted in two-thirds of adult patients being initially denied by insurers, and after a lengthy appeals process, one-third of adults remained ultimately denied. Conversely, only 1 in 10 pediatric patients were initially denied, and all were subsequently approved with considerably less effort. Second, PA policies have become significantly more restrictive over time, and there were no clinical characteristics that were related to PA approval, including commonly accepted indications of clinical trial enrollment or reirradiation. Indeed, there was no association between insurance approval and ASTRO’s model policy guidelines for PBT coverage. Third, patients with commercial insurance were significantly more likely to receive initial denial compared with those with Medicaid. Finally, PA delayed treatment start by an average of 3 weeks (and up to 4 months) for adults requiring appeal and unfortunately resulted in a substantial percentage of patients abandoning treatment altogether.

The overall denial rates we found are in agreement with previous findings. In the aforementioned report by the Alliance for Proton Therapy Access,⁹ it was found that in a sample of 1100 adult patients with commercial insurance, 63% were initially denied and 42% were ultimately denied, similar to our findings of 64% and 32%, respectively. However, the authors did not assess patients’ clinical characteristics or the eventual impact on treatment delays, and furthermore they did not examine or compare to patients without PA requirement (namely, patients with Medicare insurance). In a separate study looking only at pediatric patients,¹² it was found that 11% were initially denied, and only a single patient was ultimately denied; similarly, we found 9% were initially denied with no ultimate denials.

PA is a time-consuming and labor-intensive process for patients requiring advanced treatment interventions such as PBT. Even among patients who had insurance approval upfront, treatment start was delayed by an average of 1 week compared with those with no PA requirement, reflecting the cumbersome process of repeated requests for documentation, back-and-forth facsimile and postal mail communication, and multiple peer-to-peer phone calls (which are often missed and rescheduled or, regrettably, not initially scheduled with a qualified specialty-specific provider). This process is amplified for each level of appeal in denied cases, substantially increasing the administrative burden and distracting providers from patient care with peer-to-peer discussions that are longer on average than those for photon RT authorizations. Even more critically, the PA process can result in direct patient harm; for example, in some cases patients were forced to undergo suboptimal therapeutic interventions while waiting for the appeals process to be completed, eventually receiving approval months later.

Insurers regularly justify their denial decisions by pointing to a lack of randomized level 1 data supporting PBT; indeed, almost all stakeholders agree that clinical trials are needed to demonstrate that PBT provides clinically significant benefit compared with advanced photon techniques.^13,14 As such, national collaborative groups have launched multiple federally funded randomized phase 3 clinical trials of PBT versus photon RT,¹⁴ and ASTRO and the National Cancer Institute have called for payors to approve PBT for patients on these trials (Appendix E1; available online at https://doi.org/10.1016/j.ijrobp.2018.12.021).¹⁰ Yet our findings show that insurers are largely unwilling to provide coverage for patients intended for enrollment on these randomized phase 3 trials: 64% were initially denied, and of these patients, 67% were ultimately denied after appeal. Not surprisingly, national trial investigators have had significant difficulty in enrolling patients owing to insurance coverage,¹⁵ and the investigators of a smaller, institutional phase 2 randomized trial had to close their trial entirely because of nonaccrual.¹⁶ If payors do not provide coverage for patients on trial, how will the requisite evidence base be developed to prove or disprove the merits of PBT?

Similarly, insurers have not implemented ASTRO’s model policy guiding the appropriate coverage of PBT. Group 1 tumors are recommended for PBT coverage based on medical necessity and clinical data,¹⁰ yet initial denial rates were equivalent between group 1 patients and those without any group designation, for whom ASTRO has made no such recommendation. Reirradiation is a particular group 1 indication that has widespread support for PBT and often results in patients being referred to PBT facilities. A systematic review of PBT reirradiation for locoregional recurrences across a variety of disease sites demonstrated that PBT largely provides effective salvage of recurrent disease while maintaining low to acceptable rates of toxicity in patients who have previously underwent multiple lines of multimodality treatment.¹⁷ However, we found that among patients seeking reirradiation, 57% were initially denied, of whom 40% were ultimately denied after appeal. Many of the cases that went to external review involved the reirradiation setting, and the high reversal rate of 60% suggests payors are inappropriately denying PBT without full consideration of the individual clinical scenario.

Although PBT is an expensive treatment modality, studies have found that it can be cost-effective if used in appropriately selected patients by reducing long-term morbidity and therefore long-term costs.¹⁸ Additionally, PBT costs are expected to decrease over time as a result of advancements in technology,^8,19 further increasing the cost-effectiveness ratio. However, payors are not incentivized to account for long-term costs because of their high patient turnover—studies have shown that approximately 17% to 43% of patients can change insurance carriers in any given year.^20,21 Thus, unless fundamental system-wide changes are made, insurers will tend to focus on upfront costs rather than the long-term benefits and cost savings that PBT can provide for an individual patient with cancer.

Nevertheless, this study has several limitations. First, the retrospective nature of the analysis makes it prone to selection bias. There may be patients for whom PBT was recommended but not pursued because of physician anticipation of denial (based on insurance carrier or other factors). However, including any such patients would likely increase the denial rates already found. Second, there is often criticism that PBT may be overused when available, and a potential limitation of this study could be the inclusion of patients for whom PBT was not advantageous compared with photon RT. However, we found that in one-third of patients who received photon RT, PBT was covered by insurance but not pursued by the physician because of inferior dosimetry (as comparison dosimetry is conducted between proton and photon plans for all patients), demonstrating clinical equipoise in selecting the modality that best serves the patient. Third, our results may not be applicable to different practice settings, especially those that have negotiated coverage policies or reference pricing for PBT,²² although concordance with other studies suggests general applicability. Lastly, because of loss of follow-up, we are unable to assess the long-term clinical impact of delaying or forgoing RT, which for some patients may be significant, especially those with recurrent disease who have a clinical urgency to begin treatment and undergo suboptimal interventions or experience disease progression while waiting for resolution of an extended appeals process. Further research is warranted in this area.

Conclusions

We found that PA requirements in adults represent a significant burden in initiating PBT through extensive time and resources associated with approval and subsequent delays in patient care. Insurance approval has also become more restrictive over time, without any relationship to individual patient characteristics or clinical practice guidelines. Of particular importance is the fact that insurers have not readily provided coverage for patients on randomized clinical trials to generate high-quality evidence for PBT. Payors and providers should seek to streamline coverage policies in alignment with established guidelines to ensure appropriate and timely patient care.