Characteristics, treatment patterns, healthcare resource use, and costs among pediatric patients diagnosed with neurofibromatosis type 1 and plexiform neurofibromas: a retrospective database analysis of a medicaid population

Abstract

Objectives

The objectives of this study were to describe the characteristics and initial treatment patterns, healthcare resource use (HCRU), and costs of patients newly diagnosed with neurofibromatosis type 1 (NF1)-related plexiform neurofibromas (PN).

Methods

This was a retrospective study of individuals enrolled in the IBM MarketScan Multi-State Medicaid database from 1 October 2014 to 31 December 2017. Patients aged ≤18 years at the index date (first diagnosis of NF1 or PN, whichever occurred later) with at least 1 ICD-10-CM diagnosis code for both NF1 and PN were included. All-cause HCRU and the associated direct costs during the follow-up period were calculated per patient per year (PPPY) in 2018 USD.

Results

A total of 383 patients were included with a mean follow-up of 448 days. Most patients were diagnosed by a specialist (63.5%). During the follow-up period, pain medications were used by 58.5% of patients, 25.1% were treated with chemotherapy, 7.1% received surgery for PN, 1.6% received MEK inhibitors, and 0.8% received radiation. Mean PPPY inpatient, outpatient, ER, pharmacy, and other visits were 1.4, 17.3, 1.6, 13.6, and 25.8, respectively. Mean ± SD (median) total PPPY healthcare costs were $17,275 ± $61,903 ($2889), with total medical costs of $14,628 ± $56,203 ($2334) and pharmacy costs of $2646 ± $13,303 ($26).

Conclusions

This study showed that many pediatric patients newly diagnosed with NF1 and PN were initially treated with supportive care only, highlighting a substantial unmet medical need. This study also highlights the considerable economic burden among patients with NF1 and PN.

Keywords:

• Neurofibromatosis type 1
• plexiform neurofibromas
• treatment patterns
• healthcare resource use
• treatment costs
• Medicaid

Introduction

Neurofibromatosis type 1 (NF1) is a rare autosomal dominant genetic disorder caused by mutations in the NF1 tumor suppressor gene¹. It is characterized by progressive cutaneous, neurological, skeletal, and neoplastic manifestations ranging from minor freckling of the skin to mobility-limiting disfigurement and serious neurological deficits¹. NF1 significantly decreases lifespan^2–4 and has a global prevalence of 1 in 3000 to 1 in 6000 people^1,5,6.

The major clinical manifestations of NF1 become apparent during childhood and adolescence, and most patients are diagnosed by the age of 8⁷. The diagnostic criteria are well established and include café au lait macules on the skin, Lisch nodules in the eye, dysplasia in the bones, and externally visible or internal neurofibromas⁸. Approximately 20–50% of NF1 patients exhibit plexiform neurofibromas (PNs)^9–12, which are distinguished from other NF1-related tumors by their growth in the nerve sheaths of multiple large nerve bundles and their anatomical invasiveness^10,13. They are typically benign but can grow rapidly during childhood and adolescence^1,8 and give rise to malignant peripheral nerve sheath tumors^14,15.

Clinical management of patients with NF1-related PNs consists of surveillance of disease progression and palliative care for symptoms, including, when possible, surgical excision of tumors⁸. However, because of their invasive nature, complete removal of PNs is generally impossible^1,16,17, and tumors tend to regrow¹⁸. Ongoing clinical trials are examining the efficacy of targeted therapies, that is, drugs targeted to the specific cellular signaling pathways that are disrupted by the NF1 mutation^19,20, with the goal of slowing or stopping the growth of PNs. Only recently has one of these therapies, selumetinib, been approved for use in pediatric patients with symptomatic, inoperable PNs²¹.

Other than the information gleaned from the National Cancer Institute’s NF1 natural history cohort (NCT00924196)^22–24, little is known about the real-world experience of NF1 patients with PN. The most recent study of costs in NF1 was published 20 years ago²⁵, and treatment patterns and healthcare resource utilization (HCRU) of NF1 patients with PN are underreported in the literature. The objective of this study was to describe the demographic characteristics and initial treatment patterns, HCRU, and costs among patients newly diagnosed with NF1 and PN in the United States. This was done in a pediatric population of Medicaid-insured patients.

Methods

Study design

We retrospectively analyzed medical and pharmacy claims data. The study period was from 1 October 2014, to 31 December 2017. The index date was the date of the first diagnosis of NF1 or PN, whichever occurred later, on or after 1 October 2015. The baseline period was defined as the 12-month period before the index date. The follow-up period varied in length, ranging from the index date to the end of the study period or to the end of continuous enrollment in the health plan, whichever occurred first.

Data source

Claims data were from the IBM MarketScanⁱ Multi-State Medicaid database²⁶. This national database, maintained by IBM Watson Health, belongs to a family of administrative claims databases that compile data on approximately 40 million patients per year in the United States²⁷. The databases contain patient-level data on healthcare expenditures and outpatient prescription claims from inpatient and outpatient medical claims as well as clinical utilization records. The Multi-State Medicaid database contains the same de-identified, individual-specific details as the other IBM databases, but contains additional details on race and reasons for eligibility (disability, financial need, etc.).

Study population

Patients diagnosed with both NF1 and PN, aged ≤18 on the index date, and continuously enrolled for ≥12 months before the index date were included in the study population (Figure 1). Continuous enrollment was defined as no lapse in insurance coverage longer than 45 days. NF1 and PN were identified using International Classification of Diseases, 10th revision, Clinical Modification (ICD-10-CM) codes of Q85.01 for NF1; and D36.1x, D31.6x, and D33.3 for PN.

Figure 1. Patient selection flow chart. Abbreviations. NF1, neurofibromatosis type 1; PN, plexiform neurofibroma.

Study outcomes

The study outcomes were demographic characteristics in the baseline period (or on the index date); and all-cause HCRU, all-cause costs, imaging services, and treatments in the follow-up period. Patient demographics included age, gender, race, and type of health insurance. Information about the index diagnosis was also collected, including the place of service codes and provider type category on the diagnosis claim. The specialist provider type included provider codes 100–799 (excluding codes 204 and 240), which in practice meant primarily neurologists, ophthalmologists, pediatricians, radiologists, pediatric hematologists/oncologists, and geneticists.

Treatments were broadly classified as surgery for PN, pain medication, chemotherapy, radiotherapy, and targeted therapies. Surgery for PN was identified by Current Procedural Terminology codes 64790 and 64792²⁸. Pain medications were identified by National Drug Codes using previously published methods^22,23,29. Chemotherapy included all medications and procedures coded as “chemotherapy” in the Cancer Research Network’s 2017 chemotherapy look-up tables³⁰. Radiotherapy was identified by codes listed in the Cancer Research Network’s radiation therapy look-up tables³⁰. Targeted therapies included MEK inhibitors (binimetinib, cobimetinib, and trametinib dimethyl sulfoxide), tyrosine kinase inhibitors (bosutinib, dasatinib, imatinib, nilotinib, ponatinib, and sunitinib), and pegylated interferon alpha-2b.

All-cause HCRU included inpatient, outpatient, emergency room (ER), pharmacy, and other encounters. “Other” encounters encompassed home visits, ambulance rides, and visits to independent laboratories, federally qualified health centers, rural health clinics, and community mental health centers. Claims for imaging services (CT, MRI, and PET) were identified by the Healthcare Common Procedure Coding System (HCPCS) and ICD PROC codes (list available upon request). Costs associated with inpatient, outpatient, ER, and other encounters were classified as medical costs, separate from the costs associated with pharmacy claims.

Statistical methods

All statistical analyses were descriptive and were carried out using SAS version 9.4. Patient characteristics from the baseline period or index date were presented as numbers and percentages for categorical variables and as means and standard deviations (SDs) for continuous variables. Treatments during the follow-up period were presented as the number and proportion of patients with each service or treatment type. Follow-up HCRU was tabulated as the number and percentage of patients with each type of resource use and also as the mean (SD) number of encounters per patient per year (PPPY), which was calculated as the total number of claims for each type of resource use divided by the total length of follow-up in years among patients with each type of resource use. Costs during follow-up were presented as means (SD) or medians (interquartile range [IQR]). Healthcare costs PPPY were calculated as the total cost divided by the total number of days of enrollment in years, where costs were weighted by each patient’s length of follow-up to avoid overestimation and annualized for patients observed for less than 1 year. All costs were adjusted to 2018 US dollars based on the medical care component of the Consumer Price Index³¹.

Results

Demographic characteristics

A total of 383 patients were included in the analysis, 199 (52.0%) of whom were male (Table 1). The mean (SD, range) age on the index date was 11.4 (4.4, 2.0–18.0) years. The racial composition of the population was 47.3% White, 30.0% Black, and 5.2% Hispanic. Most index diagnoses occurred in a campus-based outpatient hospital (68.4%) or doctor’s office (25.3%) and were provided by a specialist (63.5%). The mean (SD) and median (range) duration of follow-up were 448 (239) and 473 (5–823) days, respectively, for a total of 171,444 follow-up days (data not shown).

Table 1. Baseline demographic characteristics among pediatric patients diagnosed with NF1 and PN^A.

	All patients (n = 383)
Gender^B
Male	199 (52.0)
Female	184 (48.0)
Age at index date, mean (SD)	11.4 (4.4)
Race^B
White	181 (47.3)
Black	115 (30.0)
Other	52 (13.6)
Hispanic	20 (5.2)
Missing^C	15 (3.9)
Health plan type^B
HMO	229 (59.8)
Comprehensive	152 (39.7)
PPO + EPO	2 (0.5)
Place of service at index date for index diagnosis^D
Outpatient hospital on campus	262 (68.4)
Doctor’s office	97 (25.3)
Inpatient hospital	11 (2.9)
Emergency room	6 (1.6)
Other	41 (10.7)
Provider type at index date for index diagnosis^E
Specialists	243 (63.5)
Facility^F	153 (40.0)
Non-physician professionals	36 (9.4)
Primary care physician	2 (0.5)
Missing^C	49 (12.8)

Abbreviations. EPO, Exclusive Provider Organization; HMO, Health Maintenance Organization; PPO, Preferred Provider Organization; SD, standard deviation.

^AData are presented as n (%) unless otherwise indicated.

^BValues were captured at the index date or the date nearest to the index date in the baseline period (if there was no record at index date).

^C“Missing” represents the total number of patients with no records for a particular category during the relevant time period.

^DThe total is >100% as there may be more than one place of service for each patient.

^EThe provider type represents the unique claims by each sub-category and not by each STDPROV (a variable in the database which provides data on “provider type”) value.

^F“Facility” corresponds to an unspecified provider at one of the following: acute care hospital, laboratory, other facility (not elsewhere classified), ambulatory surgery centers, home health organization/agency, case manager, mental health facilities, transportation, or supply center.

Treatment patterns

In analyses of treatment patterns, 27 patients (7.1%) had surgery for PN (Table 2). Over half (58.5%) of the patients had used pain medications, while 25.1% had received chemotherapy, 0.8% had received radiotherapy, and 1.6% had received targeted therapy. Neuropathic pain medications commonly used by NF1/PN patients included gabapentin (11.2%), amitriptyline (3.9%), pregabalin (1.0%), and duloxetine (0.8%; data not shown). The most common chemotherapy was dexamethasone (17.2%). The only targeted therapy used in this population was the MEK (MAPK/ERK) inhibitor trametinib dimethyl sulfoxide.

Table 2. Treatment received during the follow-up period by pediatric patients diagnosed with NF1 and PN.

Treatment	All patients, n (%)
PN surgery	27 (7.1)
Pain treatments^A	224 (58.5)
Analgesics/NSAIDs	128 (33.4)
Opioids, opioid-like agents, and combination opioid/analgesics	119 (31.1)
Corticosteroids (tablets, solutions and injections)	91 (23.8)
Anticonvulsant	64 (16.7)
SSRIs	24 (6.3)
Tricyclic antidepressants	18 (4.7)
Topical products	13 (3.4)
Other antidepressants	8 (2.1)
Muscle relaxants	8 (2.1)
Antirheumatics/immunologics	1 (0.3)
Chemotherapy^B^,C	96 (25.1)
Dexamethasone^D	66 (17.2)
Chemotherapy^E	15 (3.9)
Tretinoin^D	9 (2.4)
Ketoconazole	8 (2.1)
Radiotherapy	3 (0.8)
Targeted therapy	6 (1.6)
MEK inhibitors (trametinib dimethyl sulfoxide)	6 (1.6)

Abbreviations. MEK, MAPK/ERK kinase; NSAIDs, nonsteroidal anti-inflammatory drugs; PN, plexiform neurofibroma; SSRIs, selective serotonin reuptake inhibitors.

^AList of pain medications was derived from Rasu et al. [29], Wolters et al. [23], and Gross et al. [22].

^BThe 5 most common chemotherapies are shown. Chemotherapy agents were selected from the Cancer Research Network’s Cancer Therapy Look-up Tables.[30].

^CPatients who took multiple medications from the list were counted more than once.

^DDexamethasone and tretinoin may be used as non-chemotherapies in conditions other than cancers.

^EIdentified only as “chemotherapy” in the procedure code.

Healthcare resource utilization and costs

In analyses of HCRU during the follow-up period (Table 3), outpatient visits were the most common type of encounter (99.7%), followed by pharmacy visits (85.6%), “other” visits (76.0%), ER visits (45.2%), and inpatient visits (13.6%). Among resource users, “other” visits were the most numerous, with a mean (SD) of 25.8 (55.0) visits PPPY. Outpatient and pharmacy visits occurred slightly more than once per month (mean (SD) 17.3 (19.0) and 13.6 (17.9) PPPY, respectively), whereas ER and inpatient visits occurred about once per year (mean (SD) 1.6 (1.4) and 1.4 (1.4) PPPY, respectively). In results not shown, imaging services were used by 258 patients (67.4%); MRI was the most common imaging service (65.3% of patients), followed by CT scans (11.8%) and PET scans (3.7%).

Table 3. All-cause health care resource use during the follow-up period among pediatric patients diagnosed with NF1 and PN.

	Number (%) of patients	Mean (SD) visits, PPPY^A
Inpatient visits	52 (13.6)	1.4 (1.4)
Emergency room visits	173 (45.2)	1.6 (1.4)
Outpatient visits	382 (99.7)	17.3 (19.0)
Other visits^B	291 (76.0)	25.8 (55.0)
Pharmacy visits	328 (85.6)	13.6 (17.9)

Abbreviations. PPPY, per patient per year; SD, standard deviation.

^AMean value among patients with each type of HCRU. Patients without HCRU were not counted in the denominator.

^B“Other” encounters encompassed home visits, ambulance rides, and visits to independent laboratories, federally qualified health centers, rural health clinics, and community mental health centers.

Mean (SD) and median (IQR) total costs PPPY during follow-up were $17,275 ($61,903) and $2889 ($207–$9499), respectively (Table 4). The cost breakdown showed that medical costs constituted 85% of the total and were driven by inpatient costs (46% of medical costs).

Table 4. All -cause costs during the follow-up period among pediatric patients diagnosed with NF1 and PN^A.

	Mean (SD) costs, PPPY	Median (IQR) costs, PPPY
Overall costs (Medical + Pharmacy)	$17274.66 (61902.59)	$2889.20 (207.05–9498.52)
Total medical costs	$14628.31 (56203.47)	$2334.35 (131.16–7749.24)
Inpatient costs	$6739.17 (48978.52)	$0.00 (0.00–0.00)
Emergency room costs	$92.79 (240.68)	$0.00 (0.00-65.78)
Outpatient costs	$3507.66 (6040.69)	$1216.08 (0.00–4660.99)
Other costs^B	$4288.69 (22166.84)	$122.55 (0.00–1142.81)
Pharmacy costs	$2646.35 (13302.55)	$25.97 (0.00–749.93)

Abbreviations. IQR, interquartile range; PPPY, per patient per year; SD, standard deviation.

^ACosts were adjusted to 2018 USD.

^BOther costs arise from “other” types of healthcare resource use as defined above.

Discussion

This study showed that the HCRU and costs of Medicaid-enrolled pediatric patients in the first year after diagnosis with NF1 and PN were substantial. Treatment patterns consisted primarily of supportive care with medications. Our data include many completely novel findings from a real-world population that highlight the burden of disease in pediatric NF1 patients with PN.

The last evaluation of HCRU and cost in NF1 was published in the year 2000; it was a prospective analysis of 201 adult NF1 patients receiving care at a hospital-based referral center in France²⁵. The long time interval between this study and ours highlights the scarcity of information on the burden and cost of the disease in NF1 patients and underscores the novelty of our findings.

One additional study published in the intervening time provides some insight into the burden of disease in children with NF1 and PN. Kim et al. assessed the pre-enrollment characteristics of 59 children (median age 8 years) enrolled in NF1/PN trials at the National Institutes of Health in 1996–2007³². Very few of these patients had had prior chemotherapy (5%), while 47% had had debulking surgery for their PN. These numbers vary widely from the proportions observed in the Medicaid population (25% had chemotherapy and 7% had surgery). The differences may be attributed to differences in age and severity between the study populations, with older patients (as in the Medicaid population) more likely to consider using chemotherapy, and patients with surgically intractable PNs (as in the NIH population) more likely to enroll in a trial.

The use of a Medicaid population as a data source offers a unique perspective on the burden of NF1 with PN. First, it includes race data that are not available in other MarketScan databases. In comparison to previously published observational studies in the United States, which drew from sources such as the National NF Registry/Lurie Children’s Hospital in Chicago^33,34, the National Cancer Institute’s natural history cohort^23,35, and the Cincinnati Children’s Hospital³⁶, our study population had a lower proportion of White patients and a higher proportion of Black patients. Thus, our results provide insight into a different cross-section of NF1 patients than might be seen with other data sources. Second, in comparison to a similar analysis performed on a commercially insured population³⁷, several interesting differences were noted. Medicaid patients were most frequently diagnosed at a campus-based hospital (68%), versus a doctor’s office for the commercial population (58%). A higher proportion of Medicaid patients used pain medications (59%), and more of them were observed to use targeted therapies (1.6%), compared to the commercially insured patients (44% and 1.0%, respectively). ER visits were more frequent in the Medicaid population (45%) than the commercial population (25%), as were “other” types of resource use (76% Medicaid, 51% commercial), and medical costs were driven by inpatient rather than outpatient services. These differences, though qualitative, suggest that resource use and costs are different in NF1 patients with different types of health insurance coverage, and this may inform future studies of treatment patterns, disease burden, and health outcomes in different populations.

The rate of use of pain relievers (59%) in the Medicaid population was higher than in some previous studies^23,38. However, in a population drawn from the National Cancer Institute’s natural history cohort (n = 41; median age 8 years), the rate of pain reliever use rose from 43% at baseline to 63% at the time point of maximum PN volume (a median time span of 6.5 years)²². A high rate of pain reliever use may suggest that patients in the Medicaid population have larger tumors and/or experience more pain than some other cohorts.

The substantial resource use and cost burden described here, along with the findings of others regarding the burden of disease, highlight the need for new and more effective pharmacotherapies. Targeted agents like the recently approved selumetinib²¹ represent the future of treatment for NF1 patients, especially those with inoperable PNs. The use of targeted agents will almost certainly lead to changes in treatment patterns, HCRU, and costs that can be the subject of future studies.

Limitations

Several limitations inherent to the use of administrative claims data apply to our study. First, since the IBM MarketScan Multi-State Medicaid database contains information on individuals enrolled in Medicaid, the results may not be generalizable to other populations, for example, the uninsured or those with employer-sponsored coverage. Second, the use of diagnosis codes and procedure codes to identify NF1 patients and PN-related procedures is imperfect because of variations in coding. Specifically, there is no single ICD code identifying PN; however, the diagnosis codes used in this study were reviewed for accuracy by a clinician specializing in NF1. Also, the specific reasons for certain procedures, such as radiotherapy, are not provided. Third, drugs administered during an inpatient stay do not appear in the database, so we were not able to capture individual drugs administered in the hospital. Fourth, although dexamethasone and tretinoin are included as chemotherapy in the Cancer Research Network’s 2017 chemotherapy look-up tables³⁰, they may be used as non-chemotherapies in conditions other than cancer. Fifth, claims data lack the granularity that would provide details on the source of pain (PN versus something else), so the specific reason for use of pain medication cannot be inferred. Similarly, details on the site and the size of PNs are lacking. Finally, NF1 is a chronic disease, and compared to a long-term horizon, the follow-up period in our study was relatively short (mean follow-up of 448 days, or 1.2 years). Study timing also affected the list of treatments observed, as many therapies were still undergoing investigation at the study endpoint in 2017.

Conclusions

This study characterized the initial treatment patterns, HCRU, and costs of Medicaid-enrolled pediatric patients with NF1 and PN in the United States. The results show that many patients were treated with supportive care only, highlighting a substantial unmet medical need. This study also highlights the considerable economic burden among patients newly diagnosed with NF1 and PN. Our data alleviate scarcity of information on the burden of NF1 with PN, which is needed to help improve the treatment and health outcomes of these patients.

Transparency

Declaration of funding

Funding was provided by Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA.

Declaration of financial/other relationships

XY, KD, ES, and SS are employees of Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA, and maybe stockholders of Merck & Co., Inc., Kenilworth, NJ, USA. NA, KM, and SC, are employees of Complete HEOR Solutions, North Wales, PA, USA, and received financial compensation for their participation in the study analysis. A reviewer on this manuscript has disclosed that they have been a medical adviser for Astra Zeneca on plexiform neurofibromas. Peer reviewers on this manuscript have no other relevant financial relationships or otherwise to disclose.

Author contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis, and interpretation, or in all these areas; took part in drafting, revising, or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Ethics approval and consent to participate

The Medicaid database contains de-identified information, and, therefore IRB approval and informed consent were not necessary.

Acknowledgements

The authors thank Melissa Stauffer, PhD in association with ScribCo, for medical writing assistance.

Data availability statement

The data that support the findings of this study are available from the Medicaid claims repository, https://www.ibm.com/products/marketscan-research-databases, but restrictions apply to the availability of these data, which were used under license for the current study and so are not publicly available.

Notes

i IBM MarketScan Research Databases is a registered trademark of IBM Watson Health, Ann Arbor, MI, USA.