Real-world utilization of molecular diagnostic testing and matched drug therapies in the treatment of metastatic cancers

Abstract

Aims: To assess the frequency of biopsies and molecular diagnostic testing (human DNA/RNA analysis), anti-cancer drug use (genomically-matched targeted therapy [GMTT], unmatched targeted therapy [UTT], endocrine therapy [ET], and chemotherapy [CT]), and medical service costs among adults with metastatic cancer.

Methods: Adults diagnosed with metastatic breast, non-small cell lung (NSCLC), colorectal, head and neck, ovarian, and uterine cancer (2010Q1–2015Q1) were identified in the OptumHealth Care Solutions claims database and followed from first metastatic diagnosis for ≥1 month and until the end of data availability. Utilization was assessed for each cancer cohort (all and patients aged ≥65 years); per-patient-per-month (PPPM) medical service costs were assessed for all patients. Testing frequency estimates were applied to Surveillance, Epidemiology, and End Results Program data to estimate the number of untested patients (2010–2014).

Results: Patients with metastatic cancer (n = 8,193; breast [n = 3,414], NSCLC [n = 2,231], colorectal [n = 1,611], head and neck [n = 511], ovarian [n = 275], and uterine [n = 151]) were 63 years old (mean), with 11.1–22.2 months of observation. Biopsy and molecular diagnostic testing frequencies ranged from 7% (uterine) to 73% (ovarian), and from 34% (head and neck) to 52% (breast), respectively. Few were treated with GMTT (breast, 11%; NSCLC, 9%; colorectal, 6%). Treatment with UTT ranged from 0.7% (uterine) to 21% (colorectal). Biopsy, diagnostic testing, and anti-cancer drug therapy were less frequent for those ≥65 years. Medical service costs (PPPM, mean) ranged from $6,618 (head and neck) to $9,940 (ovarian). The estimated number of untested new patients with metastatic cancer was 636,369 (all) and 341,397 (≥65).

Limitations: In addition to the limitations of claims analyses, diagnostic testing frequency may be under-estimated if patients underwent testing prior to study inclusion.

Conclusions: The low frequency of molecular diagnostic testing suggests there are opportunities to better inform management of patients with advanced cancer, particularly decisions to treat with GMTT.

Keywords:

• Claims analysis
• molecular diagnostic testing
• genomic profiling
• metastatic cancer
• real world
• targeted therapy
• utilization

Introduction

Cancer is the second leading cause of mortality in the US, accounting for 23% of all deaths in 2014¹. An estimated 600,920 people in the US will die from cancer in 2017, primarily attributable to metastatic lung, breast, prostate, and colorectal cancer, although metastatic ovarian, uterine, and head and neck cancer are also significant contributors to cancer-related mortality¹. The estimated 5-year survival of patients with metastatic cancer depends on several factors, including the primary site of origin, stage and the extent of metastasis, characteristics of both the tumor and the patient, and the timeliness of and response to anti-cancer treatment. Despite a decades-long decline in cancer incidence and the introduction of innovative therapies, the 5-year survival of patients with stage IV cancers is low: 20–52% for head and neck^2–6, 17–35% for ovarian⁷, 22% for breast³, 15–17% for uterine⁶, 12% for colorectal⁸, and just 1% for non-small cell lung cancer (NSCLC)⁹.

Current treatment for metastatic cancer includes cytotoxic chemotherapy (CT) and targeted therapy, including genomically-matched targeted therapy (GMTT), unmatched targeted therapy (UTT), and endocrine therapy (ET). Cancer is not a monolithic disease, and tumor characteristics are highly heterogeneous in terms of molecular profile, mutational burden, and immune features¹⁰. The availability of genetic information related to cancer types has resulted in new fields of integrative medicine such as molecular pathological epidemiology, which can help inform pathogenesis based on diagnostics^11,12. As the molecular mechanisms involved in various types of cancers have become elucidated, the ability to specifically target these pathways has become a promising area of oncology research¹³. Targeted therapy has been shown to be effective in several metastatic cancers with specific gene mutations or molecular biomarkers, and sophisticated molecular diagnostics like next-generation sequencing now enable greater personalization of risk assessment and treatment selection to prevent treatment failure, avoid unnecessary treatment, and improve survival in certain types of cancer^14–24. Accordingly, over 600 molecular diagnostic tests are now included in oncology guidelines, with purposes spanning risk assessment and treatment selection²⁵. Clinical guidelines for several metastatic cancers now recommend molecular diagnostic testing for known molecular biomarkers and genomic alterations that can be matched with targeted therapies. For example, for the purposes of treatment selection, the National Comprehensive Cancer Network Clinical Practice Guidelines in Oncology (NCCN Guidelines) recommend testing for expression levels of HER2 in breast cancer²⁶; broad molecular profiling for various mutations in NSCLC (e.g. EGFR, KRAS, ALK, ROS1, PD-L1, BRAF V600E, MET, and HER2)²⁷; testing for KRAS, NRAS, and BRAF mutations in colon and rectal cancer^28–30; and testing for EBV/DNA for nasopharyngeal cancer^31,32.

Previous studies have assessed the frequency of molecular diagnostic testing in actual clinical practice for specific cancer types and, despite the guideline recommendations, have generally identified low rates of testing. The most recent studies have focused on NSCLC and have found a range of testing frequencies, several of which are less than 50%. A retrospective study of EGFR mutation testing among patients newly-diagnosed with metastatic NSCLC (01/2013–06/2014) found evidence of testing within 6 months of diagnosis for 18% of patients; 5% of the overall cohort had claims indicating treatment with erlotinib had been initiated³³. In that study, among patients treated with a specific drug indicated for NSCLC, 42% and 37% of those treated with erlotinib and bevacizumab or pemetrexed, respectively, underwent EGFR testing³³. A retrospective study of patients with NSCLC who were treated with erlotinib or crizotinib (2009–2012) reported that 72.7% had undergone biopsies; however, just 13.7% underwent biomarker testing for EGFR and ALK mutations that would have informed treatment decisions³⁴. In a study of EGFR testing of patients diagnosed with NSCLC in 2010 to assess eligibility for treatment with erlotinib (n = 1,358), 16.8% of all patients and 22.6% of stage IV adenocarcinoma patients were tested³⁵. Among advanced non-squamous NSCLC patients (n = 1,168) who were treated in US community oncology practices and initiated second-line CT in 2007–2011, testing rates for EGFR, KRAS, and ALK among patients were found to be 11.0%, 3.4%, and 2.4%, respectively. Nevertheless, in this timeframe EGFR testing rates increased significantly in 2010 (15.2%) and in the first 6 months of 2011 (32%) compared with testing rates before 2010³⁶. Molecular diagnostic testing rates were greater than 50% in a retrospective administrative claims data study of patients identified with metastatic NSCLC (2010–2012) who were insured by Humana (n = 2,623), the majority of which were enrolled in Medicare Advantage (n = 2,501) vs commercial insurance (n = 122)³⁷. Evidence of molecular diagnostic testing was identified for 60.9% (n = 1,597); 733 of these patients also had claims for CT or targeted therapy, and 88.8% (95% CI = 86.1–90.9%) were tested before claims indicating treating initiation³⁷. There is limited real-world information on the direct and downstream impacts of molecular diagnostic testing on healthcare resource use and costs. This information may help determine the economic value of these tests in the management of metastatic cancer, and it can contribute to a greater understanding of the value of treatments in oncology, in general³⁸.

To provide a broader benchmark for current care patterns, the present study describes the frequencies of biopsy procedures and molecular diagnostic testing among commercially insured adults in the US with selected metastatic cancers, along with use of anti-cancer therapies and associated costs. The number of patients with metastatic cancer at a population level who were untested is estimated based on the testing frequencies observed in this study.

Methods

Data sources

The data source for this study was the OptumHealth Care Solutions, Inc. administrative claims database, which included complete medical and drug claims for beneficiaries, their dependents, and retirees (over 19 million covered lives) who were privately insured by 84 Fortune 500 companies located throughout all census areas of the US. The database contains the complete medical and drug claims for those younger than 65 years of age (over 17 million lives), and drug claims and claims for Medicare co-pays and deductibles for those 65 years of age and older (over 1.5 million lives). The data were de-identified and in compliance with the Health Insurance Portability and Accountability Act; thus, no institutional board review was required.

To estimate the total number of untested cases of metastatic cancer for selected tumor types of interest in the US population during each of the years spanning 2010–2014, Surveillance of Epidemiology and End Results (SEER) Program data on incidence of new cancers per year, and the percentage of new cancers by age group and by stage were used. The estimated number of new cases of cancer per year was obtained from published literature^39–44. The percentage of new cancer diagnoses by age group was obtained from SEER 18 (2010–2014) data using the SEER Cancer Stat Facts database³⁹, and data on the stage distribution of new cancers were obtained from the 2005–2014 series available through the interactive SEER*Explorer website⁴⁵.

Patient selection

This retrospective claims analysis included US adults with six selected metastatic cancers (breast, NSCLC, colorectal, head and neck, ovarian, and uterine). These cancers were selected to comprise frequently encountered solid tumors with multiple approved targeted therapies and guideline recommendations for molecular diagnostic testing. Patients were eligible for inclusion in the study if they had: (1) claims with diagnosis codes for breast, NSCLC, colorectal, head and neck, ovarian, or uterine cancer on two different dates; and (2) claims with diagnosis codes for metastasis on two different dates between January 1, 2010 and March 31, 2015 and following their first observed claim for the primary cancer. The first such claim for metastatic disease was defined as the index date. Diagnoses were identified using International Classification of Disease-9th revision, Clinical Modification (ICD-9-CM) codes, listed in Supplementary Appendix A. Head and neck cancer included cancers of the oral cavity, pharynx, larynx, paranasal sinuses and nasal cavity, and salivary glands; cancers of the brain, esophagus, eye, parathyroid, or thyroid were not included. Eligible patients were also required to be ≥18 years old as of the index date, and to have continuous eligibility for at least 1 month after the index date. Patients were excluded if they had diagnoses of other cancers, such as sarcoma, melanoma, brain cancer, thyroid cancer, renal cancer, or cancer of unknown primary origin.

Analyses

Selected patients were stratified by cancer type on the index date; the study observation period was at least 1 month and until the end of data or end of the observation period (March 31, 2015), whichever came first. Patient characteristics as of the index date (age, sex, US geographic region, and health plan type) were summarized as counts and percentages, stratified by cancer type. The mean length of the post-index observation period in months, along with its standard deviation (SD), was calculated for each cancer-specific cohort.

The proportions of patients in each cancer-specific cohort who underwent biopsies and diagnostic tests were assessed during the post-index observation period (codes used to identify biopsies and diagnostic tests are listed in Supplementary Appendix B). Mean costs associated with selected tests and procedures, anti-cancer drugs (GMTT, UTT, ET, and CT; codes used to identify treatments are listed in Supplementary Appendix C), and all other medical services (including hospice or palliative care, emergency room visits, hospitalizations, outpatient visits, and other visits) were also assessed on a per-patient-per-month (PPPM) basis following the index date in each cancer-specific cohort. Costs were calculated in 2015 US dollars using the Consumer Price Index, medical component⁴⁶; means and SDs were calculated by cohort. A sub-group analysis was conducted for patients aged 65 years or older.

To assess the implications of frequencies of molecular diagnostic testing, SEER data were used to estimate the number of patients with metastatic cancer for the selected cancers of interest in the US who were untested. For each year, the number of new cases was summed, and the percentage of new cases by age group and by stage were averaged for each of the selected cancers. The percentage of new cases occurring in patients 65 years of age and older was based on the reported percentage of new cases by age group for those 65–74, 75–84, and over 84 years of age. The number of new cases of cancer each year that were metastatic was calculated by applying the percentage of new distant-stage cancers to the number of new cases in each year. For this analysis, the percentage of new cases that were metastatic and the percentage of new cases occurring in patients ≥65 years of age were assumed to be constant over the period 2010–2014. The total number of tested cases for each year was estimated by applying the observed frequency of molecular testing estimated in this study to the number of newly-diagnosed distant-stage cancers. The number of untested cases for each year was then calculated as the difference between the estimated number of new cases of distant-stage cancer and the total number of tested cases for each year. To align the SEER data as closely as possible with the cancer diagnoses for colorectal, head and neck, and uterine cancer used in the analyses of administrative claims data, SEER cancer site data were aggregated: colon and rectum cancers (colorectal); oral cavity and pharynx and larynx cancers (head and neck); and cervix uteri and corpus and uterus cancers (uterine).

Results

The study sample included 8,193 patients with metastatic cancers of interest who met all criteria for inclusion (Figure 1). The numbers of patients in each cancer-specific cohort and lengths of observation (months, mean [SD]) were: breast, 3,414 (22.2 [16.0]); NSCLC, 2,231 (11.1 [10.9]); colorectal cancer, 1,611 (17.1 [13.8]); head and neck cancer, 511 (21.1 [16.0]); ovarian cancer, 275 (17.5 [14.6]); and uterine cancer, 151 (17.5 [14.5]). Of the overall study population, 38.8% were at least 65 years of age, including 895 patients with breast cancer, 1,284 patients with NSCLC, 676 patients with colorectal cancer, 190 with head and neck cancer, 82 with ovarian cancer, and 55 with uterine cancer.

Figure 1. Sample selection: adult patients with selected metastatic cancers. Abbreviations. n, number; NSCLC, non-small cell lung cancer.

Patient characteristics

Patient characteristics are presented in Table 1 (all patients) and Supplementary Appendix D (patients ≥65 years of age). Across cancer-specific cohorts, age (mean [SD]) ranged from 58.6 (12.9) years among patients with breast cancer to 68.0 (10.7) years among patients with NSCLC. Among patients ≥65 years, the average age of patients ranged from 74.1 (7.3) years to 77.0 (7.9) years in patients with ovarian and colorectal cancer, respectively. All patients in the ovarian and uterine cohorts were women; of the cancers that were not limited to women, 98.9% (breast), 43.7% (NSCLC), 39.8% (colorectal), and 17.2% (head and neck) were female. The majority of patients in each cancer-specific cohort resided in the Northeast (range among all patients: 26.5% [uterine] to 35.4% [head and neck]), Midwest (20.4% [head and neck] to 28.5% [uterine]), or South (31.1% [uterine] to 33.9% [head and neck]); less than 15% of all patients in any cohort were from the West. US geographic region was similarly distributed among patients ≥65 years old. The most common health plan type among all patients was a preferred provider organization (range = 43.9% [NSCLC] to 56.4% [breast]), followed by indemnity health insurance (26.2% [breast] to 41.2% [NSCLC]); less than 16% of all patients in any cohort received healthcare services through a point of service health plan. Among patients ≥65 years old, the most common health plan type was indemnity health insurance (range = 58.5% [ovarian] to 69.1% [uterine]).

Table 1. Baseline characteristics by cancer-specific cohort (metastatic cancer)^a.

	Breast	NSCLC	Colorectal	Head and neck	Ovarian	Uterine
Characteristic	(n = 3,414)	(n = 2,231)	(n = 1,611)	(n = 511)	(n = 275)	(n = 151)
Age (years) at the index date, mean ± SD	58.6 ± 12.9	68.0 ± 10.7	63.9 ± 13.8	62.9 ± 12.3	61.4 ± 11.2	63.6 ± 11.8
Female, n (%)	3,378 (98.9)	976 (43.7)	641 (39.8)	88 (17.2)	275 (100.0)	151 (100.0)
US geographic region, n (%)
Northeast	954 (27.9)	717 (32.1)	523 (32.5)	181 (35.4)	82 (29.8)	40 (26.5)
Midwest	824 (24.1)	595 (26.7)	374 (23.2)	104 (20.4)	64 (23.3)	43 (28.5)
South	1,098 (32.2)	699 (31.3)	523 (32.5)	173 (33.9)	88 (32.0)	47 (31.1)
West	507 (14.9)	180 (8.1)	174 (10.8)	49 (9.6)	34 (12.4)	19 (12.6)
Unknown	31 (0.9)	40 (1.8)	17 (1.1)	4 (0.8)	7 (2.5)	2 (1.3)
Health plan type, n (%)
Exclusive provider organization	—	1 (<0.1)	—	—	—	—
Indemnity health insurance	896 (26.2)	920 (41.2)	556 (34.5)	168 (32.9)	88 (32.0)	54 (35.8)
Point of service	511 (15.0)	305 (13.7)	212 (13.2)	74 (14.5)	37 (13.5)	14 (9.3)
Preferred provider organization	1,924 (56.4)	980 (43.9)	825 (51.2)	257 (50.3)	146 (53.1)	82 (54.3)
Other or unknown	83 (2.4)	25 (1.1)	18 (1.1)	12 (2.3)	4 (1.5)	1 (0.7)

n, number; NSCLC, non-small cell lung cancer; SD, standard deviation; US, United States.

^aBaseline characteristics were assessed at the index date.

Utilization of diagnostic tests, biopsies, and anti-cancer therapy

For each cancer-specific cohort, the proportion of patients who underwent any biopsy procedures and any molecular diagnostic testing during the observation period is presented in Figure 2; additional data on utilization are presented in Supplementary Appendix E (all patients) and Supplementary Appendix F (≥65 years old). The observed frequencies for biopsy procedures among all patients ranged from 7% for uterine cancer to 73% for ovarian. For breast, NSCLC, colorectal, and head and neck cancer, the observed frequencies of biopsy procedures were 23%, 31%, 31%, and 33%, respectively. Among patients ≥65 years old, the observed frequencies for biopsy procedures ranged from 9–81%. Across cancer-specific cohorts, the observed frequencies of molecular diagnostic tests among all patients were 52% for breast, 42% for NSCLC, 37% for colorectal, 34% for head and neck, 41% for ovarian, and 42% for uterine cancer; among patients ≥65 years old, the observed frequencies for diagnostic tests were less than 50%—42%, 41%, 31%, 30%, 38%, and 36%, respectively.

Figure 2. Utilization of biopsy procedures and molecular diagnostic tests, by cancer-specific cohort (metastatic cancer). Abbreviation. NSCLC, non-small cell lung cancer.

CT was the most frequently used treatment, ranging from 21% (uterine) to 52% (colorectal) across all cancer-specific cohorts, and from 18% (head and neck) to 49% (ovarian) in those ≥65 years of age (Figure 3 and Supplementary Appendix F). GMTT was incorporated in treatment regimens among 11% of all patients with breast cancer, 9% of all patients with NSCLC, and 6% of all patients with colorectal cancer; among patients ≥65 years old, the observed frequencies for GMTT were 6% for breast cancer, 7% for NSCLC, and 6% for colorectal cancer, respectively. No patients in the head and neck, ovarian, or uterine cancer cohorts received treatment with matched targeted therapy. UTT was used by at least one patient in each cancer-specific cohort, ranging from 0.7% (uterine) to 21% (colorectal) among all patients, and 2% (uterine) to 15% (colorectal) among patients ≥65 years old. ET was used within the breast (60% of all patients, 50% of patients ≥65 years old), ovarian (14% of all patients, 17% of patients ≥65 years old), and uterine (16% of all patients, 15% of patients ≥65 years old) cancer cohorts.

Figure 3. Utilization of anti-cancer drugs, by tumor type (metastatic cancer). Genomically-matched targeted therapies are intended to be genomically-matched per the prescribing information. Abbreviation. NSCLC, non-small cell lung cancer.

Healthcare costs

The PPPM costs (mean [SD]) associated with biopsies, molecular diagnostic tests, anti-cancer drugs, and medical services are presented in Table 2. Costs associated with biopsy procedures and molecular diagnostic testing were much lower than those associated with anti-cancer drug therapy and all medical services. PPPM costs for biopsy procedures ranged from $3 ($31) for uterine cancer to $42 ($84) for ovarian cancer. Molecular diagnostic testing costs were much lower than either drug costs or costs of medical services on a PPPM basis; PPPM testing costs (mean [SD]) ranged from $7 ($24) for head and neck cancer to $106 ($780) for ovarian cancer.

Table 2. Healthcare costs by cancer-specific cohort (metastatic cancer).

	Breast	NSCLC	Colorectal	Head and neck	Ovarian	Uterine
Costs (2015 USD), PPPM, mean ± SD	(n = 3,414)	(n = 2,231)	(n = 1,611)	(n = 511)	(n = 275)	(n = 151)
Biopsy procedures	14 ± 180	37 ± 531	12 ± 142	16 ± 80	42 ± 84	3 ± 31
Molecular diagnostic tests	40 ± 127	35 ± 132	29 ± 155	7 ± 24	106 ± 780	32 ± 135
Targeted therapy
Genomically matched	349 ± 1,464	255 ± 1,152	164 ± 1,005	—	—	—
Endocrine	60 ± 238	—	—	—	4 ± 31	3 ± 14
Unmatched	84 ± 697	240 ± 1,420	545 ± 1,685	77 ± 690	176 ± 845	—
Cytotoxic chemotherapy	293 ± 823	425 ± 1,528	701 ± 1,709	45 ± 264	300 ± 881	61 ± 299
Total medical costs, excluding anti-cancer drug costs	6,667 ± 11,011	8,405 ± 16,642	8,521 ± 14,503	6,618 ± 9,969	9,940 ± 15,043	7,823 ± 15,176
Hospice or palliative care services	17 ± 152	114 ± 605	61 ± 351	49 ± 403	80 ± 412	35 ± 304
Emergency room visits	93 ± 359	244 ± 715	161 ± 472	91 ± 288	223 ± 559	153 ± 375
Hospitalizations	1,484 ± 8,181	3,582 ± 12,965	3,842 ± 11,847	1,894 ± 5,407	5,306 ± 13,243	3,207 ± 12,930
Outpatient visits	4,946 ± 6,642	4,322 ± 7,831	4,172 ± 6,991	4,452 ± 7,113	3,990 ± 5,166	4,258 ± 7,087
Other visits	128 ± 375	142 ± 614	284 ± 1,322	132 ± 430	341 ± 1,356	170 ± 830

Abbreviations. PPPM, per patient per month; n, number; NSCLC, non-small cell lung cancer; SD, standard deviation; USD, United States dollars.

Among the anti-cancer drug treatments, CT was least costly compared with GMTT and UTT on a PPPM basis. The mean and (SD) PPPM costs of treatment with CT were $293 ($823) for breast, $425 ($1,528) for NSCLC, $701 ($1,709) for colorectal, $45 ($264) for head and neck, $300 ($881) for ovarian, and $61 ($299) for uterine cancer. Among the three cancer-specific cohorts in which GMTT use was observed, the mean and (SD) PPPM costs of treatment were $349 ($1,464) for breast, $255 ($1,152) for NSCLC, and $164 ($1,005) for colorectal cancer. Among the five cohorts with more than one patient who used UTT, the mean and (SD) PPPM costs of treatment were $84 ($697) for breast, $240 ($1,420) for NSCLC, $545 ($1,685) for colorectal, $77 ($690) for head and neck, and $176 ($845) for ovarian cancer. Among the three cohorts that used ET, the mean and (SD) PPPM costs of treatment were $60 ($238) for breast, $4 ($31) for ovarian, and $3 ($14) for uterine cancer.

Excluding the cost of anti-cancer drugs, PPPM costs for all medical services were $6,667 ($11,011) for breast, $8,405 ($16,642) for NSCLC, $8,521 ($14,503) for colorectal, $6,618 ($9,969) for head and neck, $9,940 ($15, 043) for ovarian, and $7,823 ($15,176) for uterine cancer. These costs were primarily driven by the costs of hospitalization (mean [SD] PPPM range = $1,484 [$8,181] for breast to $5,306 [$13,243] for ovarian cancer) and outpatient visits ($3,990 [$5,166] for ovarian to $4,946 [$6,642] for breast cancer).

When observed frequencies of molecular diagnostic testing were applied to the SEER data on the number of new cases of metastatic cancer during the 2010–2014 time period, in the US a total of ∼552,871 and 341,397 cases of breast, NSCLC, colorectal, head and neck, ovarian, and uterine cancer among all ages and those ≥65 years, respectively, were estimated to be untested. Even if the frequency of diagnostic testing was doubled, the number untested would have been ∼169,013 and 99,541 for all ages and those ≥65 years of age, respectively (Table 3).

Table 3. Estimated number of untested cases of metastatic cancer for all patients and patients ≥65 years of age, 2010–2014

Cancer type	New cases, n^40–44	New cases, ≥65 years, %^39	New cases, metastatic, %^40	Untested, n
Breast
All	1,140,350	—	5.6	30,397
≥65 years	—	42.8	5.8	16,560
Lung and bronchus^a
All ages	1,122,210	—	53.2	346,866
≥65 years	—	69.4	50.7	233,361
Colorectal^b
All ages	706,890	—	19.6	91,305
≥65 years	—	57.0	17.6	49,073
Head and neck^b
All ages	262,720	—	17.9	29,414
≥65 years	—	45.8	16.4	13,886
Ovarian
All	110,370	—	58.7	38,419
≥65 years	—	45.1	66.9	20,713
Uterine^b
All	301,040	—	9.5	16,470
≥65 years	—	36.9	11.1	7,803
Total
All	3,643,580	—	25.8	552,871
≥65 years	—	53.5	28.7	341,397

n, number; NSCLC, non-small cell lung cancer.

^aTesting rates for NSCLC were applied to the SEER cancer data for lung and bronchus cancer.

^bThe following SEER cancer site data were aggregated: colon and rectum cancers (colorectal); oral cavity and pharynx and larynx cancers (head and neck); and cervix uteri and corpus and uterus cancers (uterine).

Discussion

Matching targeted anti-cancer treatment to the molecular characteristics of an individual’s tumor is potentially transformative in the treatment of patients with advanced cancer, and it may lead to better patient outcomes and lower rates of treatment failure. The current study is the first to assess the frequency of molecular diagnostic testing in the US among patients with an array of metastatic cancers and across cancer sites to assess molecular diagnostic testing costs relative to the anti-cancer drug and total medical service costs associated with managing patients with metastatic cancer. Moreover, this study extrapolated from observed molecular diagnostic testing frequencies to evaluate the implications of the frequency of testing by estimating the number of patients with the studied metastatic cancers who were most likely untested during a 5-year time period.

This retrospective claims analysis found that the frequency of molecular diagnostic testing was generally low, with the majority of patients in the overall sample not undergoing molecular diagnostic testing during the study period—despite the availability of targeted therapies with approved indications, listings in recognized compendia for tumor types other than those listed in the prescribing information for anti-cancer drugs, potential for enrollment in clinical trials of investigational therapy, and recommendations for testing for hereditary risk for cancer. Across tumor types, the frequency of testing was highest (52%) among patients with metastatic breast cancer and lowest among those with head and neck (34%) cancer. For breast, colorectal, ovarian, and uterine cancers the observed frequencies reflect testing for hereditary risk factors in addition to testing to inform treatment selection, suggesting that the frequency of molecular diagnostic testing for the purpose of potentially matching patients to targeted therapies was even lower. In contrast, the frequency of biopsy was highest for ovarian cancer (73%) and lowest for uterine cancer (7%). In comparison with the overall patient population, patients with metastatic cancer ≥65 years of age underwent biopsy procedures less frequently across all cancers except ovarian and uterine; patients in this age cohort underwent molecular diagnostic testing less frequently across all tumor types. This observed pattern may be due to the underlying health status of older patients with cancer^47,48, which may also be reflected in the lower utilization of CT (for all cancers except ovarian) among patients ≥65 years of age vs the overall sample. Patients ≥65 years of age also had lower use of GMTT for all three tumor types, as well as lower use of UTT for every tumor type but uterine cancer.

The present study found comparatively lower frequencies of biopsy procedures and higher frequencies of testing in the NSCLC patient population that was studied. The differences in these findings may be due to study design and the specification of the observation period over which frequency of testing and biopsy procedures were calculated. The present study estimated the overall frequency of testing in the studied populations—how often testing is conducted, regardless of the ultimate treatment decisions—rather than the frequency of testing prior to treatment initiation with a particular drug, which has been observed to be as high as 89% in NSCLC³⁷. The higher observed frequencies of molecular diagnostic testing compared with those of biopsy (with the exception of ovarian cancer) may indicate that many biopsy samples used for testing were banked prior to the development of metastatic disease or progression on initial therapy for metastatic disease. In addition, some patients may decline biopsy or treatment, and molecular diagnostic testing may be less common among frail patients, including the elderly and patients with multiple comorbid conditions⁴⁹. For ovarian cancer, biopsies are not usually performed until the tumor (and usually the entire ovary) is surgically removed, due to the risk that early biopsy may spread the cancer⁵⁰. Therefore, although the present results may under-estimate the total frequency of biopsy for non-ovarian cancers, they reflect what is observed in later stages of cancer progression.

The current study identified a range of total costs for medical services (mean PPPM) among patients with metastatic cancer ranging from $6,618 (head and neck) to $9,940 (ovarian). These costs were primarily driven by hospitalization and outpatient visits, which were highest among patients with ovarian ($5,306) and breast cancer ($4,946), respectively. Among anti-cancer drugs, CT was the most commonly used treatment in all cancer-specific cohorts, ranging from 21% (uterine) to 52% (colorectal), with corresponding PPPM (mean) costs ranging from $45 (head and neck) to $701 (colorectal). In only three of the six cohorts were patients treated with GMTT—breast (11%), NSCLC (9%), and colorectal cancer (6%)—although the costs of treatment were low ($349, $255, and $164, respectively, PPPM). Costs associated with UTT ranged from $77 PPPM (head and neck) to $545 PPPM (colorectal), and ET PPPM costs ranged from $3 (uterine) to $60 (breast). Conversely, the cost contribution of diagnostic tests relative to overall medical costs as well as anti-cancer drug costs was small, under $110 (mean PPPM) in all cancer-specific cohorts, and as low as $7 (head and neck cancer).

Molecular diagnostic testing has been found to be cost effective. Germline BRCA mutation testing among women with ovarian cancer has been found to be a cost-effective option for the UK (threshold of £20–30,000/quality-adjusted life year (QALY)⁵¹) compared with no testing, with an incremental cost-effectiveness ratio (ICER) of £4,339/QALY⁵². Molecular diagnostic testing for EGFR mutations and ALK over-expression in NSCLC prior to initiation of therapy yields an ICER of $136,000 per QALY gained, compared with no testing and treatment with CT alone⁵³. Among the tumor-testing strategies available for patients with colorectal cancer, BRAF testing had an ICER of $36,200/LY gained and microsatellite instability plus BRAF testing had an ICER of $108,000/LY gained⁵⁴. As the number of alterations or biomarkers for which GMTT are available increases, it is likely that comprehensive genomic profiling will enable the realization that greater efficiencies give the prospect of a single molecular diagnostic test vs testing for multiple alterations, potentially sequentially.

The relatively low frequency of molecular diagnostic testing observed in this study has broad implications. The estimated number of untested patients with newly-diagnosed metastatic cancer (2010–2014) in the US among the cancer sites selected for this study was over half a million. When patients undergo diagnostic testing, diagnostic tests impact between 60–70% of all clinical decisions, while accounting for just 4–5% of healthcare costs^55–57. If patients do not undergo diagnostic tests that assess risk or the suitability of potential treatments, significant costs are incurred, broadly. For patients, there is potentially a missed opportunity to be treated with GMTT, in clinical practice or in a genomic clinical trial. Given that older patients are generally most at risk for cancer-related morbidity and mortality^47,58, particularly in the context of metastatic cancer, the low frequency of testing, biopsy, and use of potentially less toxic targeted therapies among patients aged ≥65 years of age may indicate an unmet need in this age group. For the health system in the US, an estimated $604 billion could be saved annually if genomic tests were performed to determine suitability of anti-EGFR therapy among patients with metastatic colorectal cancer^57,59. Furthermore, prior studies have reported 30–50% reductions in direct hospital and outpatient charges through the use of diagnostic tests that identify key alterations in health status and facilitate modifications in therapeutic interventions to improve patient outcomes^55,57.

The present study results indicate low frequencies of diagnostic test use among patients with the selected metastatic cancers, suggesting potential misalignment with current treatment guidelines for management of metastatic cancer. In the case of NSCLC, prior to 2016, EGFR mutation testing was recommended for NSCLC patients with a high probability of EGFR mutation in order to tailor tyrosine kinase inhibitor therapy to patients who would receive the most benefit from these drugs after the first line⁶⁰. However, beginning in 2016, the NCCN guidelines recommend that all patients with non-squamous NSCLC undergo testing for EGFR mutations (even if previous small biopsies found other NSCLC histology) as part of broad molecular profiling, due to the benefit in progression-free survival for first-line matched targeted therapy vs CT for EGFR mutation-positive NSCLC^27,60. With this update to the NCCN guidelines for NSCLC, it will be important going forward to assess and promote guideline adherence, given that the currently observed real world frequencies of testing are so low. In the case of metastatic colorectal cancer, the NCCN recommends that all patients with suspected or proven adenocarcinoma receive KRAS/BRAF testing²⁸, which also was not reflected in the current study results. The low frequency of molecular diagnostic testing may explain why GMTT use was also low: 11% for breast, 9% for NSCLC, 6% colorectal cancer, and 0% for all others. This finding is concerning as it suggests that some patients, particularly those with NSCLC, may not be receiving optimal treatment tailored to their disease profile.

In light of prior research on the potential positive impact of molecular diagnostic testing in the management of metastatic cancer, future research on methods to increase the adherence to oncology treatment guidelines and a more widespread adoption of personalized medicine may enable patients with metastatic cancer to benefit from opportunities for improved clinical outcomes. Past research has noted that efforts to increase physician compliance with treatment and supportive care best practices can result in substantial cost savings to healthcare systems⁶¹. Additionally, with widening indications for some drugs according to the presence of a genomic alteration, as with the indication of pembrolizumab for unresectable or metastatic solid tumors positive for microsatellite instability-high (MSI-H)⁶², it becomes even more important to ensure that patients are diagnosed with appropriately comprehensive molecular testing.

Limitations

Administrative claims data provide a large sample of real-world practice patterns, but are subject to important limitations. The data are collected for reimbursement purposes, rather than clinical care or primary research, and may be subject to coding errors or data omissions. The drug utilization measured in the present study based on prescription fills may differ from actual utilization to the extent that some patients do not take their medications. The current study included only adult patients covered by commercial insurance though large employers. The results may not generalize to other populations, such as Medicaid enrollees, Medicare enrollees, or the uninsured. In addition, the database included limited cost information among patients ≥65 years old; since the data on costs were not complete among these patients, healthcare costs among the sub-set of patients ≥65 years old were not analyzed separately. Due to the time lag associated with administrative claims data, use of recently approved therapies, including immuno-oncology therapies, could not be assessed. Diagnostic tests could not be linked definitively to prescribed therapies in the claims data. In this analysis, the frequency of biopsy procedures and of molecular testing were assessed separately. A proportion of the molecular diagnostic tests may have been conducted following biopsy procedures; however, molecular diagnostic tests may have been conducted using archived samples of the primary tumor. Because we analyzed the utilization observed in the claims data after the first identified diagnosis of metastatic cancer, it was not possible to determine how many tests were conducted following a biopsy of a primary tumor prior to the first observed metastatic diagnosis. Some patients may have undergone earlier molecular diagnostic testing prior to inclusion in this study. Finally, to estimate the implications of observed relatively low frequencies of testing, the current study applied the observed frequencies from the administrative claims data to US population estimates reported in SEER program data.

Conclusions

This retrospective analysis of real-world care patterns among patients with selected metastatic cancers found that less than half underwent molecular testing to inform treatment selection. In addition, although total medical costs were substantial, molecular diagnostic testing was a relatively small proportion of the costs of managing patients with metastatic cancer. These findings suggest that there are opportunities to better inform treatment strategies for patients with metastatic cancers through appropriate use of molecular diagnostic testing.

Transparency

Declaration of funding

This study was funded by Foundation Medicine, Inc. The sponsor was involved in all parts of the study and in manuscript preparation.

Declaration of financial/other relationships

JR and RA are employees of Foundation Medicine, Inc. and own stock/stock options. AC, MP, and JS are employees of Analysis Group, Inc., which has received consulting fees from Foundation Medicine, Inc. NL was an employee of Analysis Group, Inc. during the study’s conduct. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgments

Medical writing assistance was provided by Shelley Batts, PhD, an employee of Analysis Group, Inc.