Impact of L-Methylfolate Combination Therapy Among Diabetic Peripheral Neuropathy Patients

Diabetic peripheral neuropathy (DPN) accounts for significant morbidity and predisposes the lower extremities to debilitating complications such as infection, ulceration, and eventually amputation.¹ Nearly 70% of adults with diabetes have some manifestation of neuropathy.² Diabetic peripheral neuropathy involves progressive deterioration of nerve fi bers and results in chronic pain in as many as 16% of patients with diabetes.³

The economic burden of DPN is substantial. Compared with patients without neuropathy, patients with DPN have a higher incidence of diabetes-related conditions such as coronary artery disease, back pain, limb amputations, depression, hypertension, valvular or peripheral vascular disease, and limb infections.⁴ The annual medical costs per patient with DPN were estimated to be $14,062 in 2003, compared with $6651 per patient with diabetes without DPN.⁴ The total US annual direct costs of DPN and its complications were estimated to range from $4.6 to $13.7 billion in 2001.⁵ Diabetic peripheral neuropathy is associated with prolonged hyperglycemia and abnormal glucose tolerance.⁶ The pathophysiologic process of DPN includes increased oxidative stress, which in conjunction with a hyperglycemic environment leads to the destruction of nerve fibers.⁶ A major feature of DPN is sensory loss, which may lead to foot ulceration following even minor trauma as the patient may be oblivious to the injury. Such ulcers in the lower extremities are common, with the annual incidence of lower extremity ulcers in the population with diabetes being approximately 7%.⁷ Of every 6 people with diabetes, 1 person will develop a foot ulcer at some point,⁸ with peripheral neuropathy being a contributing cause in as many as 60% to 70% of all diabetic foot ulcers.^9,10 One study identified peripheral neuropathy as a cause in 78% of cases, making it the most common factor leading to ulceration.¹¹

Several classes of pharmacologic agents have been used for symptomatic pain relief or DPN.¹² These include anticonvulsants (eg, gabapentin, pregabalin), some antidepressants (eg, tricyclics, selective serotonin reuptake inhibitors/serotonin-norepinephrine reuptake inhibitors), opioid analgesics, anxiolytics, and sedative hypnotics.¹³ These agents may provide pain relief, but do not address the underlying pathophysiology of DPN. Side effects and inadequate response to these agents encourage many patients with DPN to explore alternative treatments.¹⁴

A product classifi ed by the US Food and Drug Administration as a prescription medical food¹⁵ containing L-methylfolate, pyridoxal-5'-phosphate, and methylcobalamin (MPM; Metanx, Pamlab LLC, Covington, Louisiana) is available. MPM is indicated for the distinct nutritional requirements of patients with endothelial dysfunction who present with loss of protective sensation and neuropathic pain associated with DPN.¹⁶ This oral formulation has been studied in patients with DPN with sensation loss, neuropathic pain, and lower extremity ulcerations. Clinical studies have associated MPM with improved sensory perception, reduced neuropathic pain, and restored sensation in patients with DPN.^12,17 A recent placebo-controlled trial of 214 patients demonstrated a signifi cant improvement on the Neuropathy Total Symptom Score-6 at 16 and 24 weeks in patients with DPN. The rate of adverse events was comparable to that with placebo.¹⁸

Few studies report the impact of treatments for DPN on healthcare costs and hospitalization risk. Previous research concluded that use of duloxetine in the treatment of DPN was associated with lower total healthcare costs than standard-of-care medications.^19,20 A preliminary study examining the impact of MPM on prescription drug utilization and healthcare costs identifi ed a positive trend in healthcare costs, and recommended that further research be conducted to validate the fi ndings.²¹ The objective of this study was to assess the impact of MPM treatment on hospitalization risk and healthcare costs among patients with DPN in a real-world setting.

METHODSStudy Design

This retrospective cohort study used patient-level data obtained from the HealthCore Integrated Research Database. This database contains administrative claims data from 14 commercial health insurance plans geographically dispersed across the United States, representing approximately 43 million patients. All study data were de-identifi ed and complied with regulations set by the Health Insurance Portability and Accountability Act of 1996. Patient confi dentiality was preserved, and the anonymity of all patient data was safeguarded throughout the study. It was determined that a waiver of informed consent from an institutional review board was not required to conduct this study.

Inclusion and Exclusion Criteria

The study population consisted of patients who had at least 2 medical claims for type 2 diabetes (International Classifi cation of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 250.x0 or 250.x2) on different dates or 1 pharmacy claim for antidiabetic agents (Generic Product Identifier code 27 excluding 2730) during the study period of January 1, 2004, through July 31, 2010. Patients were also required to have at least 1 medical claim for peripheral neuropathy (ICD-9-CM codes 250.x6, 337.1x, 355.7x, or 357.2x) during the study period. The active treatment population consisted of patients with a minimum treatment

course of at least 2 pharmacy claims for MPM; other patients were available for matching as controls.

The first pharmacy claim date for MPM was defi ned as the index date for MPM users. The index date for the control population was randomly assigned within the same time period of study claims for MPM. All patients were aged 18 to 64 years as of the index date and had a minimum of 12 months preindex and postindex continuous eligibility within the study period for baseline and follow-up evaluations. A 12-month washout period of no pharmacy claims for any prescription folate—containing product was used to select patients for whom MPM was newly initiated.

Patients with severe lower limb morbidity in the 12-month preindex period were excluded. Severe lower limb morbidity included decubitus ulcer (ICD-9-CM code 707.0x), gangrene (ICD-9-CM code 785.4 or 440.24), and amputation (ICD-9-CM code 84.1x or 997.6x, or Current Procedural Terminology codes 28800-28825, 27880-27889, or 27590-27598). Patients with claims for unbranded forms of MPM (National Drug Code 42192-0317-90, 13925-0144-90, or 13925-0144-50) during the study period were excluded.

The control population was matched 1:1 to patients in the MPM treatment group using a greedy match propensity score algorithm.²² Propensity scores were generated using multivariate logistic regression predicting probability of MPM use. All available demographic characteristics and variables associated with disease severity such as diabetic comorbidities, preindex medication burden, and the Deyo-Charlson Comorbidity Index (DCI) were included as covariates in the propensity score model. Matched cases (MPM group) and controls (non- MPM group) with the highest digit on the propensity score were selected for analysis.

Figure 1

shows the study cohort construction.

Study Measures

The primary outcomes of interest in this study were hospitalization risk (all cause and disease related) and direct healthcare costs (all cause and disease related) during the 12-month postindex period. All-cause hospitalization was defined as any inpatient admission, whereas disease-related hospitalization was defi ned as an inpatient admission containing at least 1 billing code indicating diabetic neuropathy or severe lower extremity morbidity (see the

eAppendix

at www.ajpblive.com). Hospitalization risk was defined as the probability of 1 or more hospitalizations during the postindex period.

The baseline (preindex period) was defined as the 12 months prior to the index date; the follow-up (postindex period) was defined as the 12-month period after the index date, including the index date. Age, sex, region of residence, insurance plan type, and clinical characteristics were assessed for study patients. Patients’ comorbid conditions as well as their DCI score were determined during the baseline period.²³ Medications (nonnarcotic analgesics, nonsteroidal anti-infl ammatory drugs [NSAIDs], opioid analgesics, antidepressants, anticonvulsants, anxiolytics or sedative hypnotics, antibiotics, and insulin) were assessed during the 12-month preindex period for the MPM and control groups.

Total all-cause costs were defi ned as the 12-month postindex sum of all medical and pharmacy claim costs. Disease-related costs were defined as those associated with treatment of diabetic neuropathy or severe lower extremity morbidity in all treatment settings, as well as prescription costs associated with the treatment of DPN (eAppendix). Costs were calculated using the plan-paid amount and were adjusted to the 2010 mid-year Consumer Price Index.

Statistical Analyses

For the matched MPM and control populations, differences on matched variables used in propensity scores were evaluated using the Wilcoxon signed rank test for continuous data (age, DCI score, number of unique medications received) and the McNemar test for categorical data (sex, region, insurance plan type, presence of specific comorbid conditions). Descriptive statistics such as means (± standard deviation) and proportions (frequencies) were used to characterize continuous and categorical study variables, respectively. The t test was used to compare the differences between the MPM and control groups for continuous variables, the χ² test was used for categorical variables, and the Mann-Whitney test was used for cost data.

Multivariable logistic regression was conducted to compare hospitalization risk between the MPM and matched control groups, controlling for differences in preindex utilization of pain medications. Generalized linear models with gamma distribution were used to compare all-cause and disease-related costs between the 2 groups, adjusting for baseline demographics and clinical characteristics. All descriptive and multivariate analyses were conducted with SAS version 9.1 (SAS Institute, Cary, North Carolina). Statistical signifi cance was set at P <.05.

RESULTS

A total of 81,898 patients (814 in the MPM group and 81,084 in the control group) met all study inclusion and exclusion criteria prior to propensity score matching (Figure 1). Of those, 814 MPM patients and 814 matched controls were selected for the final study cohort.

The demographic and clinical characteristics were similar between the matched treatment groups (

Table 1

). Patients’ mean age was 54 years and 55% were male. A majority of patients were enrolled in preferred provider organization insurance plans in both the MPM (71.7%) and control (66.7%) groups. Among the study patients, the most frequent comorbidities were hypertension, dyslipidemia, and back pain. The mean DCI score was not significantly different between the groups.

Baseline Medication Use

No difference was found between the MPM and control groups on the use of baseline nonnarcotic analgesics and NSAIDs (32.2% for MPM vs 30.7% for non-MPM, P = .52), opioid analgesics (51.1% for MPM vs 49.9% for non- MPM, P = .62), antidepressants overall (39.4% for MPM vs 36.7% for non-MPM, P = .26), anxiolytics and sedative hypnotics (25.4% for MPM vs 25.4% for non-MPM, P = .99), antibiotics (66.5% for MPM vs 66.8% for non-MPM, P = .88), and insulin (63.5% for MPM vs 64.5% for non- MPM, P = .68) (

Figure 2

).

At baseline, a greater proportion of patients in the MPM group used anticonvulsants (34.52% vs 25.18%, P <.001), the most common being gabapentin (21.38% vs 17.08%, P = .03) and pregabalin (12.41% vs 4.42%, P <.001). Although overall preindex antidepressant use was similar between the 2 study groups, patients in the MPM group were more likely to use duloxetine (10.69% vs 4.67%, P <.001).

Hospitalization Risk

During the 12-month preindex period, the percentage of patients with at least 1 all-cause hospitalization was similar between the 2 groups (20.52% vs 22.11%, P = .43). However, the postindex difference was signifi cant: the percentage was 21.50% for the MPM group and 25.92% for the control group (P = .04). The percentage of patients with disease-related hospitalizations was similar between the MPM and control groups during both the preindex (4.79% vs 6.27%, P = .19) and postindex (6.63% vs 7.37%, P = .56) periods.

After multivariate adjustment for differences in baseline pain-related medication use, the MPM group had signifi cantly

lower risk of postindex all-cause hospitalization (odds ratio [OR] 0.74, 95% confidence interval [CI] 0.58-0.94) compared with the control group. Disease-related postindex hospitalization risk was not statistically signifi cantly different between the 2 groups (

Figure 3

), although the average length of stay for MPM users was 10.65 (±19.03) days compared with 20.17 (±39.18) days for the control group.

Healthcare Costs

Costs increased from the preindex to the postindex period for both the MPM and control groups. All-cause 12-month costs increased by a mean of $4096 in the MPM treatment group from the preindex to the postindex period (P <.001), whereas in the control population, the costs had a larger mean increase of $6283 (P = .017). In both the MPM and control groups, the increase in mean pre-post diseaserelated costs was signifi cant, again with a greater increase in the control group than in the MPM group (

Table 2

).

Following multivariate analyses, the total adjusted allcause costs in the MPM group were approximately 9% lower than those in the control group during the followup period, although the difference was not signifi cant (Table 3). However, the adjusted disease-related postindex costs were found to be signifi cantly lower in the MPM group compared with the control group (

Table 3

).

DISCUSSION

We believe this is the first study to use administrative claims from a large US insured population to determine the hospitalization risk associated with and the economic impact of MPM treatment in the DPN population. When we applied a propensity score matching using patient demographics and disease severity measures as covariates and further controlled for remaining differences in preindex pain medication use between the 2 groups, MPM-treated patients were signifi cantly less likely to be hospitalized in the postindex period for any cause. The MPM-treated group also had signifi cantly lower diseaserelated costs in the postindex period compared with the control population.

In DPN, hyperglycemia causes oxidative stress and leads to decreased nitric oxide availability and vasoconstriction in the peripheral vasculature,²⁴ and has been linked to impaired wound healing.^25,26 Results of previous research have shown MPM increases nitric oxide production and reduces homocysteine levels among DPN patients, 12 and MPM has been shown to improve endothelial function in patients with type 2 diabetes.²⁷ One study has reported that MPM treatment was associated with improved epidermal nerve fiber density and reduced paresthesias in patients with type 2 diabetes and small-fiber neuropathy.¹⁷ In a study using an animal model for type 2 diabetes, MPM was associated with alleviations in hindlimb digital sensory nerve conduction deficits, induction of small sensory nerve fiber regeneration, and increases in intra-epidermal nerve fiber density. It is postulated that the mechanisms for these effects involve inhibition of oxidative-nitrosative stress.²⁸ These data suggest that MPM has positive effects on the underlying pathology of DPN and that these effects may improve patient outcomes beyond palliative relief.

This study found that MPM-treated patients were 30% less likely to be hospitalized for any cause in the postindex period compared with the matched controls. The postindex mean all-cause cost for the MPM group was lower than that for the control group by $1663 (P = .09), but this difference was not statistically signifi cant. The postindex hospitalization costs for the MPM group were lower than those for the control population, but this difference was offset by increased costs in the outpatient and pharmacy settings. Previous studies reported that outpatient costs account for a substantial proportion of total healthcare costs associated with DPN.^19,20,29

Disease-related postindex hospitalization risk was not statistically different between the 2 groups after controlling for baseline characteristics. However, among patients with a disease-related admission, the average length of stay for the control group was longer than that for the MPM group during the postindex period. In contrast to all-cause costs, disease-related costs were dominated by inpatient utilization, which accounted for more than 50% of total costs for both groups. The higher percentage of patients with at least 1 disease-related hospitalization and the longer average length of stay resulted in a mean adjusted postindex disease-related cost that was $2258 higher for the control group than for the MPM group (P <.001).

A cost analysis of the diabetic Medicare population found that medical expenditures associated with lower extremity ulcers were 3 times higher on average than costs for Medicare patients in general. Lower extremity ulcer—related costs accounted for 24% of total costs in these patients, and most of the ulcer-related costs occurred in the inpatient setting.⁷ The current study suggests that utilization of MPM in patients with DPN may decrease hospitalization risk for lower extremity causes and may reduce the length of stay should hospitalization occur. This impact appears to be the major factor contributing to the disease-related cost benefi t seen in the MPM group.

Similar to other studies, this study found that DPN patients had high utilization of palliative therapy such as opioid analgesics, antidepressants, anticonvulsants, nonnarcotic analgesics, and NSAIDs.^19,30 In this study, more than two-thirds of the MPM and control patients used antibiotics in the preindex period, a potential marker of lower extremity infection.

This study should be interpreted in light of certain limitations. Any retrospective study can only describe the association between groups, and cannot construe causality. Medical conditions defi ned in this study were identified using ICD-9-CM billing codes and were subject to miscoding; however, our requirement for multiple claims reduced the likelihood of false positives when identifying the DPN population. Despite propensity score matching, whether there was unknown or unmeasured selection bias in the MPM-treated group cannot be ascertained. Although the preindex healthcare resource use was similar between the MPM and control populations, there was no practical way to determine whether the actual severity of disease was similar between the groups. Although the requirement for at least 2 MPM claims ruled out patients who immediately discontinued treatment, the length of therapy and adherence to treatment with MPM in the postindex period were unknown. Given that this study investigated commercially insured patients, findings may not be generalizable to other populations. The cost estimates in this study included only direct plan-paid costs, so no data were reported on out-of-pocket costs to the patient. Claims data do not capture indirect costs incurred by patients, including loss of productivity and travel expenses, although these costs are relatively higher for chronic illnesses such as DPN.^2,4,5,31,32

Despite these limitations, administrative claims data remain a powerful data source and have been used in previous studies of the healthcare costs associated with DPN.^19,20,29 The current study used standard techniques to control for confounders.

CONCLUSIONS

This study found that commercially insured patients with DPN who had at least 2 prescription claims for MPM had a significantly lower risk of all-cause hospitalization in the 12 months following initiation of treatment compared with a control population. The mean 12-month postindex disease-related cost also was $2258 lower for MPM patients than for controls (P <.001). When used in patients with DPN, MPM may have a favorable clinical and cost benefit.

Despite some serious and valid concerns, the FDA concluded that if just 1% of Americans use the test, OraQuick could identify up to 45,000 previously undetected HIVpositive people and prevent close to 4000 new infections each year. So it basically comes down to a risk/benefi t for the individual versus the community as a whole—the risk for the 1 in 12 false negative individuals versus the societal benefits from the projected thousands of consumers who are identified as HIV positive and can obtain earlier treatment and prevent further transmission.¹⁴