A Cost-Effectiveness Comparison of Desipramine, Gabapentin, and Pregabalin for Treating Postherpetic Neuralgia

Alec B. O'Connor MD; Katia Noyes PhD, MPH; Robert G. Holloway MD, MPH

J Am Geriatr Soc. 2007;55(8):1176-1184. 

Abstract and Introduction

Abstract

Objectives: To compare the net health effects and costs resulting from treatment with different first-line postherpetic neuralgia (PHN) medications.
Design: Cost–utility analysis using published literature.
Participants: Hypothetical cohort of patients aged 60 to 80 with PHN.
Interventions: Desipramine 100 mg/d, gabapentin 1,800 mg/d, and pregabalin 450 mg/d.
Measurements: A decision model was designed to describe possible treatment outcomes, including different combinations of analgesia and side effects, during the first 3 months of therapy for moderate to severe PHN. The main outcome was cost per quality-adjusted life-year (QALY) gained. Costs were estimated using the perspective of a third-party payer. Multivariate, univariate, and probabilistic sensitivity analyses were performed, and the time frame of the model was varied to 1-month and 6-month horizons.
Results: Desipramine was more effective and less expensive than gabapentin or pregabalin (dominant) under all conditions tested. Gabapentin was more effective than pregabalin but at an incremental cost of $216,000/QALY. Below $140/month, gabapentin became more cost-effective than pregabalin at a threshold of $50,000/QALY, and below $115/month gabapentin dominated pregabalin.
Conclusion: Desipramine appears to be more effective and less expensive than gabapentin or pregabalin for the treatment of older patients with PHN in whom it is not contraindicated. After its price falls, generic gabapentin will likely be more cost-effective than pregabalin.

Introduction

Approximately one-fourth of all Americans will develop herpes zoster during their lifetimes, and 10% to 33% of these cases will result in postherpetic neuralgia (PHN).[1] PHN occurs primarily in older adults[2] and is associated with substantial reductions in quality of life and function.[3,4]

The management of PHN is expensive,[5] and patients often continue to have severe pain despite taking prescribed analgesics.[3,4] The best initial therapy for patients is unknown, because there have been no head-to-head randomized controlled trials (RCTs) of the first-line medications.[6,7] Recent systematic reviews of treatment options for PHN[6] and peripheral neuropathic pain[7] have concluded that tricyclic antidepressants are the most efficacious option, but when the potential for side effects is included in decision-making, tricyclic antidepressants, gabapentin, and pregabalin are all considered first-line treatment options.[7]

Gabapentin is currently the most commonly prescribed neuropathic pain medication for PHN and other neuropathic pain conditions[3,4,8,9] despite costing several times more than tricyclic antidepressants. Amitriptyline continues to be the most commonly prescribed tricyclic antidepressant for older patients with PHN,[3,4,8] even though its use is considered inappropriate in older people because of the risk of serious side effects.[10-15] Desipramine and nortriptyline have equivalent efficacy but fewer anticholinergic and adrenergic side effects;[16-22] the American Geriatrics Society has recommended them as an appropriate treatment option,[23] but they are underused in PHN.[3,4]

This analysis used existing RCT data and decision-analytic modeling to compare the costs and net treatment effects that result from treating moderate to severe PHN in older patients with desipramine, gabapentin, or pregabalin.

Methods

Target Population

This analysis estimated the costs and utilities of the first 3 months of therapy in a hypothetical cohort of patients aged 60 to 80 with moderate to severe PHN and no history of cardiovascular disease or significant electrocardiographic abnormalities. The model excluded patients with known ischemic heart disease or cardiac conduction abnormalities, because they are considered contraindications to the use of tricyclic antidepressants.[17,24]

Perspective of Analysis

The perspective used for cost estimates is that of a third-party payer. The scope is limited to consideration of direct medical expenses and changes in patients' health-related quality of life caused by pain severity and side effects. The main outcome measure used in this analysis is the cost per quality-adjusted life-year (QALY), which is the product of the utility associated with a health state and the time spent in the state, measured in years.

Interventions

The following dosages were selected to maximize the cost-effectiveness of each of the interventions, balancing efficacy, tolerability, and cost: desipramine 100 mg administered orally once daily, gabapentin 600 mg administered orally three times daily (1,800 mg/d), and pregabalin 225 mg administered orally twice daily (450 mg/d). The rationale for these dosages, along with additional description of the methods and the rationale for the model and parameter assumptions (beyond that provided below) is available in the on-line Appendix.

Decision Model Structure

Decision-analysis software (TreeAge Pro 2005 Suite, TreeAge Software, Inc., Williamstown, MA) was used to construct a decision model analogous to one previously used for neuropathic pain,[25] with several possible outcomes resulting from therapy (Figure 1). Side effects were considered to be minor or intolerable, the latter resulting in discontinuation of the medication within the first month of therapy and no pain relief. Intolerable side effects typically resulted in cessation of medication and additional clinic visits, but could rarely result in "serious side effects," which resulted in hospitalization, additional disutility, and the possibility of death. To ensure that the relative benefits of desipramine were not overestimated, it was assumed that desipramine was the only medication that could result in serious side effects. In patients without intolerable side effects, medications could also produce good pain relief or poor pain relief.

Figure 1.

 

The decision model. Only the desipramine arm is shown for simplicity. The possibility of myocardial infarction (desipramine only) is considered to represent serious side effects in the model. The gabapentin and pregabalin arms are identical, but the probability of serious side effects is set at 0. *After 6 weeks of treatment.

It was assumed that all health states resulting from treatment were present for the duration of the time frame. The possibility that patients might continue to take medication throughout the time frame even if they did not achieve good pain relief was allowed for, because it is likely that this reflects clinical practice.[3,4]

The minimum number of clinic visits was assumed to apply to patients with good pain relief and no side effects; this baseline number of clinic visits was ignored in the model, because it was common to all arms of the tree. Patients with minor side effects but good pain relief required one additional clinic visit during the study time frame. It was assumed that patients with intolerable side effects or poor pain relief generated an extra clinic visit per month; this estimate may be excessive when added to the baseline number of visits, especially as the time frame of the analysis lengthens, but the costs associated with the additional visits could also be considered to pay for additional medications that might be prescribed for pain relief or to treat side effects. A screening electrocardiogram was performed in the model before the initiation of desipramine, because an electrocardiogram has been recommended before initiation of a tricyclic antidepressant in patients aged 40 and older.[17]

Clinical Probability Estimates

Published RCTs evaluating the study medications in patients with PHN were used to generate all of the clinical probability estimates used in the model except the probability of serious side effects and the probability of continuing medication in the absence of pain relief. One of the authors (ABO) abstracted all data in a consistent manner from the published results of the RCTs. The RCTs are summarized in .

Table 1.  Comparison of Relevant Postherpetic Neuralgia (PHN) Randomized Clinical Trials

The proportion of patients in PHN RCTs who achieved moderate or greater pain relief (or much improvement or greater on the patient global impression of change scale) was used to represent the probability in the model of achieving good pain relief from a medication. All of the available RCT data were incorporated into the probability estimates, except that data from subjects who received 150 mg/d of pregabalin, which appears to be a less-efficacious dose, were excluded. In some instances, the estimates differ from the published results because a strict intention-to-treat analysis was applied, incorporating the total number of subjects who received medication, to standardize efficacy estimates. The proportion of patients withdrawing from RCTs due to adverse events (after subtracting from the denominator the non-adverse event dropouts) was used to represent the probability of developing an intolerable side effect from a medication. The probabilities of achieving good pain relief and side effects for the different medications are presented in according to RCT. The weighted averages of the probabilities for each medication were used in the model and are summarized in .

Table 2.  Base-Case Clinical Probability, Utility, and Cost Estimates

Table 3.  Results of the Cost-Effectiveness Analysis of the Base Case and the 1-Month and 6-Month Models

It was not possible to identify literature describing the probability that patients would discontinue PHN (or other neuropathic pain) treatment because of failure to achieve pain relief. For this reason, it was estimated that, by 6 weeks, 75% of patients would stop their PHN treatment ( ). Because of the uncertainty of this estimate, this probability was allowed to range from 0 to 100% in the sensitivity analyses to ensure that the estimate did not substantially alter the results of the analysis.

Table 4.  Results of Univariate Sensitivity Analyses*

Serious side effects, beyond intolerance leading to discontinuation of medication, are rare and without a clear pattern in RCTs. There is conflicting evidence regarding a possible association between tricyclic antidepressants and myocardial infarctions and sudden cardiac death[34,40,41] at the doses used in PHN trials. There is little information regarding any potential association between gabapentin or pregabalin and serious side effects. Serious side effects, represented by a worst-case estimate of the risk of myocardial infarction with tricyclic antidepressants,[34] were built into the model, as has been done previously.[25] It was also assumed that no serious side effects could occur with gabapentin or pregabalin use.

Utilities

The utilities used in the model are summarized in . The utilities associated with different pain states were obtained from a cross sectional survey of patients with PHN (median age 77).[3] The utilities of the baseline and poor pain relief states in the model were both considered to be the weighted average of utilities derived from patients with moderate pain and severe pain, using the EuroQol-5 Dimensions (a generic health-related quality-of-life instrument). It was assumed that the decision-model outcome of good pain relief corresponded to improvement from the baseline state to a mild pain state (complete pain relief in neuropathic pain RCTs is relatively rare), with a correspondingly higher utility value.

Costs

All costs were converted into 2006 U.S. dollars using the Consumer Price Index, Medical Care component.[42] Costs that were common to all of the treatment arms were ignored. The average wholesale prices of medications were derived from the 2006 Red Book,[38] using the least-expensive cost per pill. All patients were assumed to have obtained at least 1 month of medication.

Sensitivity Analyses

Sensitivity analyses were conducted varying the parameters through the range of published values to take into account the uncertainty in the parameter estimates. Parameters for which a range of published values did not exist were varied in either direction through a range of clinically plausible values ( ). To determine which parameters influenced the results of the model to the greatest degree, a multivariable sensitivity ("tornado") analysis was performed. Univariate sensitivity analyses were then performed on all of the identified critical parameters and on all of the medication variables.

To evaluate the sensitivity of the results to the duration of the model time horizon, the 3-month base case was converted to 1-month and 6-month models. In all of the models, the base-case parameters and assumptions were maintained except as noted. In the 1-month model, it was assumed that all patients obtained 1 month of medication, and that patients with any side effect or poor pain relief generated one extra clinic visit during the 1-month horizon. In the 6-month model, patients who discontinued medication due to poor pain relief were assumed to have done so after having obtained 3 months of medication. The probability of myocardial infarction was adjusted to fit the duration of time frame.

We also performed a probabilistic sensitivity analysis to simultaneously evaluate the uncertainty associated with the parameter estimates and the range of outcomes that result from the treatment of a population of patients.[43] We assumed a triangular distribution for all parameters[25] and performed Monte Carlo simulations (n=1,000) on a hypothetical cohort of 1,000 patients for the base-case model and for the 1-month and 6-month models.

Results

Base-Case Analysis

presents the results of the incremental cost-effectiveness analysis under the base-case conditions. Desipramine was more effective and less expensive than (dominated) gabapentin and pregabalin. Gabapentin was slightly more effective than pregabalin, but the incremental cost-effectiveness ratio for gabapentin compared to pregabalin was $216,000/QALY using the 2006 Red Book price of gabapentin.

Sensitivity Analyses

Desipramine continued to dominate gabapentin and pregabalin when the time frame of the analysis was shortened to 1 month or lengthened to 6 months ( ), and when the least favorable combination of pain relief, minor side effects, and intolerable side effects[26] was used.

Multivariable sensitivity ("tornado") analysis determined that the model was most sensitive to the following variables, in descending order of influence: utility in severe pain, utility in mild pain, probability of pain relief with desipramine, and utility with minor side effects (Figure 2). displays the results of the univariate sensitivity analyses for these variables and for all of the medication-related variables, listing the threshold values beyond which the order of the cost-effectiveness rankings changed, assuming a willingness to pay of $50,000 per QALY. Pregabalin becomes cost-effective when the probability of pain relief for desipramine decreases to 40% (from 62%), the probability of pain relief for pregabalin increases to 68% (from 45%), the utility of minor side effects decreases to 0.76 (from 0.95), the probability of serious side effects due to desipramine increases to 3.7% (from 0.15%), or the price of desipramine increases to $256 per month (from $42.60). Gabapentin becomes cost-effective when the probability of pain relief with gabapentin increases to 74% (from 42%). For six of the medication-related variables, there was no value for which desipramine was not the most cost-effective strategy.

Desipramine continued to dominate the other medications even when the prices of gabapentin and pregabalin were reduced to 0. In the base comparison between gabapentin and twice-daily pregabalin, gabapentin became cost-effective when its price fell below $145/month, and it dominated pregabalin below $115/month. In comparison with three times per day pregabalin dosing, gabapentin became cost-effective when its price fell below $203/month.

In the base model probabilistic sensitivity analysis, pregabalin was cost-effective relative to desipramine in 12.3% of trials, and gabapentin was cost-effective relative to desipramine in 9.1% of trials. Even at high valuations of QALY, the chances of pregabalin or gabapentin being cost-effective are below 20% for the base-case, 1-month, and 6-month models (online Appendix Figure).

Discussion

Desipramine was more effective and less expensive than gabapentin and pregabalin in the base case and throughout the range of plausible values considered in the sensitivity analyses. These conclusions are especially noteworthy, because older patients with PHN are almost twice as likely to be taking gabapentin as the tricyclic antidepressant amitriptyline, and desipramine use is rare in PHN.[3,4] Furthermore, these results probably underestimate the true dominance of desipramine over gabapentin and pregabalin, because the model was designed to favor gabapentin and pregabalin where uncertainty existed. The model also does not account for the potential benefits that patients with chronic pain may experience because of the antidepressant effects of tricyclics.

These results are similar to those obtained in a recent cost-effectiveness study comparing amitriptyline with gabapentin for neuropathic pain, despite that study's use of a different patient population, a different tricyclic antidepressant, substantially different utilities, more favorable tricyclic side effect estimates, and a shorter time frame.[25] In both analyses, pain relief efficacy differences between the medications, rather than side effect differences, determined the outcome. The large effect that moderate or severe chronic pain has on patients' quality of life, as measured using patient-derived utility scores, can explain this.[3,4,8,9] A recent secondary analysis of one of the PHN RCTs[29] found that analgesic effects were highly correlated with improvements in quality of life and mood, whereas treatment-emergent side effects, such as sedation, constipation, and dry mouth, did not correlate with quality of life or mood changes.[44] The results of the current study support the conclusion that patients with PHN value pain relief more than they value being free of side effects, even in the face of an exaggerated estimate of the potential for side effects.

The reasons why gabapentin is prescribed more commonly than tricyclic antidepressants have not been fully explored. The most likely explanations are that prescribers believe that gabapentin is safer than tricyclics, more effective than tricyclics, or both. Intensive marketing of gabapentin and pharmaceutical company promotion of the concept of "inappropriate prescribing" of tricyclics for neuropathic pain[46] and PHN[13] have almost certainly contributed to these perceptions; in fact, the results of published RCTs indicate that tricyclics are more effective (have lower numbers needed to treat) and seem to be better tolerated (have higher numbers needed to harm) than gabapentin or pregabalin for PHN in older patients.[6]

Another factor that may have contributed to greater prescribing of gabapentin than of tricyclics for PHN is that clinicians may not differentiate between tricyclic antidepressants. Amitriptyline is known to be associated with potentially serious anticholinergic side effects, and its use is considered to be a marker of poor care in geriatric patients,[10-12,14,15] including for PHN.[13] Secondary amine tricyclics, including desipramine, are rarely used,[3,8,9] although they are recommended for PHN;[16,17,23] they appear to be as effective as amitriptyline[19,22] but produce fewer anticholinergic side effects, including less gait and cognitive impairment in older patients.[18,20]

Given that the efficacy and side-effect data used for desipramine, gabapentin, and pregabalin were similar to those previously reported for other types of peripheral neuropathic pain syndromes,[7] the results of this analysis may be similar to cost-effectiveness analyses that would be obtained for comparisons between tricyclic antidepressants and gabapentin or pregabalin for other types of peripheral neuropathic pain. That the results were insensitive to changes in the drug parameters across a realistic range of possibilities supports this hypothesis.

In the base-case analysis, gabapentin was slightly more effective than pregabalin, but also more costly, with a cost per gained QALY higher than society is often willing to pay, although the price of gabapentin, which is now generic, is falling. A search of Internet-based U.S. pharmacies revealed an average price per month of $90 for gabapentin (pregabalin's price was close to our base estimate);[47,48] at this price, gabapentin dominates pregabalin.

A major limitation of this analysis is the comparison of data from different published RCTs. The methods and conduct of each RCT are unique, and results from one RCT (e.g., the proportion of patients experiencing pain relief or side effects) are not necessarily comparable with results from another trial. This study attempted to address this by pooling the results of multiple RCTs, although the study designs of the desipramine RCTs were different from the industry-sponsored gabapentin and pregabalin RCTs. Furthermore, the number of participants in desipramine RCTs was small, although the clinical probability estimates used in the model were similar to those derived for amitriptyline,[25] which has been studied in a much larger number of patients. It is also possible that publication bias led to overestimates of the efficacy or safety of gabapentin and pregabalin.[45,49] To address these concerns, results were used from all of the available RCTs, the sensitivity analyses was extended to the full range of values available for each drug, and a probabilistic sensitivity analysis was conducted.

Another limitation of this analysis is the generalization of short-term clinical trial outcomes to chronic, real-life use. Current practice patterns suggest that tricyclic antidepressants and gabapentin are underdosed in general practice compared with the dosages studied in clinical trials.[3,4] The medications are also likely to be titrated more slowly in practice than the forced titration used in the clinical trials. However, it is unlikely that these factors would qualitatively change the relationships between the individual medications and patient outcomes and costs. Finally, the decision model represents an oversimplification of clinical practice and the spectrum of possible responses to treatment of a population of individuals, though it is likely that the major clinical outcome possibilities were incorporated. A large head-to-head RCT measuring costs and QALYs would be more definitive.

In conclusion, current management of PHN is expensive[5] and frequently ineffective.[3,4] On the basis of the cost-effectiveness analysis, desipramine appears to be superior to gabapentin and to pregabalin in the treatment of older patients with PHN who do not have a history of cardiac disease, significant electrocardiographic abnormalities, or other contraindications to desipramine use. This conclusion appears robust to variation in key model parameters.

References

  1. Koplan JP, Harpaz R. Shingles vaccine: Effective and costly or cost-effective? Ann Intern Med 2006;145 :386–387

  2. Hall GC, Carroll D, Parry D et al. Epidemiology and treatment of neuropathic pain: The UK primary care perspective. Pain 2006;122 :156–162

  3. Oster G, Harding G, Dukes E et al. Pain, medication use, and health-related quality of life in older persons with postherpetic neuralgia: Results from a population-based survey. J Pain 2005;6 :356–363

  4. van Seventer R, Feister HA, Young JP et al. Efficacy and tolerability of twice-daily pregabalin for treating pain and related sleep interference in postherpetic neuralgia: A 13-week, randomized trial. Curr Med Res Opin 2006;22 :375–384

  5. Dworkin RH, White R, O'Connor AB et al. Healthcare expenditure burden of persisting herpes zoster pain. Pain Med in press

  6. Hempenstall K, Nurmikko TJ, Johnson RW et al. Analgesic therapy in postherpetic neuralgia: A quantitative systematic review. PLoS Med 2005;2 :e164

  7. Finnerup NB, Otto M, McQuay HJ et al. Algorithm for neuropathic pain treatment: An evidence based proposal. Pain 2005;118 :289–305

  8. McDermott AM, Toelle TR, Rowbotham DJ et al. The burden of neuropathic pain: Results of a cross-sectional survey. Eur J Pain 2006;10 :127–135

  9. Tolle T, Xu X, Sadosky AB. Painful diabetic neuropathy: A cross-sectional survey of health state impairment and treatment patterns. J Diabetes Complications 2006;20 :26–33

  10. Beers MH. Explicit criteria for determining potentially inappropriate medication use by the elderly: An update. Arch Intern Med 1997;157 :1531–1536

  11. Curtis LH, Ostbye T, Sendersky V et al. Inappropriate prescribing for elderly Americans in a large outpatient population. Arch Intern Med 2004;164 :1621–1625

  12. Lau DT, Kasper JD, Potter DEB et al. Potentially inappropriate medication prescriptions among elderly nursing home residents: Their scope and associated resident and facility characteristics. Health Serv Res 2004;39 :1257–1276

  13. Oster G, Berger A, Dukes E et al. Use of potentially inappropriate pain-related medications in older adults with painful neuropathic disorders. Am J Geriatr Pharmacother 2004;2 :163–170

  14. van der Hooft CS, Jong GW, Dieleman JP et al. Inappropriate drug prescribing in older adults: The updated 2002 Beers criteria—a population-based cohort study. Br J Clin Pharmacol 2005;60 :137–144

  15. Rigler SK, Jachna CM, Perera S et al. Patterns of potentially inappropriate medication use across three cohorts of older Medicaid recipients. Ann Pharmacother 2005;39 :1175–1181

  16. Dubinsky RM, Kabbani H, El-Chami Z et al. Practice parameter: Treatment of postherpetic neuralgia. An evidence-based report of the Quality Standards Subcommittee of the American Academy of Neurology. Neurology 2004;63 :959–965

  17. Dworkin RH, Backonja M, Rowbotham MC et al. Advances in neuropathic pain: Diagnosis, mechanisms, and treatment recommendations. Arch Neurol 2003;60 :1524–1534

  18. Blackwell B, Stefopoulos A, Enders P et al. Anticholinergic activity of two tricyclic antidepressants. Am J Psychiatry 1978;135 :722–724

  19. Watson CP, Vernich L, Chipman M et al. Nortriptyline versus amitriptyline in postherpetic neuralgia: A randomized trial. Neurology 1998;51 :1166–1171

  20. Draganich LF, Zacny J, Klafta J et al. The effects of antidepressants on obstructed and unobstructed gait in healthy elderly people. J Gerontol A Biol Sci Med Sci 2001;56A :M36–M41

  21. Dworkin RH, Schmader KE. Treatment and prevention of postherpetic neuralgia. Clin Infec Dis 2003;36 :877–882

  22. Rowbotham MC, Reisner LA, Davies PS et al. Treatment response in antidepressant-naïve postherpetic neuralgia patients: Double-blind, randomized trial. J Pain 2005;6 :741–746

  23. American Geriatrics Society. The management of persistent pain in older persons: AGS panel on persistent pain in older persons. J Am Geriatr Soc 2002;50 :S205–S224

  24. Roose SP, Laghrissi-Thode F, Kennedy JS et al. Comparison of paroxetine and nortriptyline in depressed patients with ischemic heart disease. JAMA 1998;279 :287–291

  25. Cepeda MS, Farrar JT. Economic evaluation of oral treatments for neuropathic pain. J Pain 2006;7 :119–128

  26. Kishore-Kumar R, Max MB, Schafer SC et al. Desipramine relieves postherpetic neuralgia. Clin Pharmacol Ther 1990;47 :305–312

  27. Rowbotham M, Harden N, Stacey B et al. Gabapentin for the treatment of postherpetic neuralgia. JAMA 1998;280 :1837–1842

  28. Rice ASC, Maton S. Postherpetic Neuralgia Study Group. Gabapentin in postherpetic neuralgia: A randomized, double blind placebo controlled study. Pain 2001;94 :215–224

  29. Gilron I, Bailey JM, Tu D et al. Morphine, gabapentin, or their combination for neuropathic pain. N Engl J Med 2005;352 :1324–1334

  30. Dworkin RH, Corbin AE, Young JP Jr et al. Pregabalin for the treatment of postherpetic neuralgia: A randomized, placebo-controlled trial. Neurology 2003;60 :1274–1283

  31. Sabatowski R, Galvez R, Cherry DA et al. Pregabalin reduces pain and improved sleep and mood disturbances in patients with postherpetic neuralgia: Results of a randomised, placebo-controlled clinical trial. Pain 2004;109 :26–35

  32. van Seventer R, Feister HA, Young JP et al. Efficacy and tolerability of twice-daily pregabalin for treating pain and related sleep interference in postherpetic neuralgia: A 13-week, randomized trial. Curr Med Res Opin 2006;22 :375–384

  33. Freynhagen R, Strojek K, Griesing T et al. Efficacy of pregabalin in neuropathic pain evaluated in a 12-week, randomised, double-blind, multicentre, placebo-controlled trial of flexible- and fixed-dose regimens. Pain 2005;115 :254–263

  34. Cohen HW, Gibson G, Alderman MH. Excess risk of myocardial infarction in patients treated with antidepressant medications: Association with use of tricyclic agents. Am J Med 2000;108 :2–8

  35. Ergin A, Muntner P, Sherwin R et al. Secular trends in cardiovascular disease mortality, incidence, and case fatality rates in adults in the United States. Am J Med 2004;117 :219–227

  36. Wilby J, Kainth A, Hawkins N et al. Clinical effectiveness, tolerability and cost-effectiveness of newer drugs for epilepsy in adults: A systematic review and economic evaluation. Health Technol Assess 2005;9 :804

  37. Nicholson T, McGuire A, Milne R. Cost-utility of enoxaparin compared with unfractionated heparin in unstable coronary artery disease. BMC Cardiovasc Disord 2001;1 :2

  38. Fleming T. Red book: Pharmacy's Fundamental Reference. Montvale, NJ: Thompson Healthcare, 2006

  39. Centers for Medicare and Medicaid Services. Medicare physician fee schedule 2006 [on-line]. Available at: http://www.cms.hhs.gov/apps/pfslookup/ Accessed July 14, 2006

  40. Hippisley-Cox J, Pringle M, Hammersley V et al. Antidepressants as risk factor for ischaemic heart disease: A case-control study in primary care. BMJ 2001;323 :666–669

  41. Ray WA, Meredith S, Thapa PB et al. Cyclic antidepressants and the risk of sudden cardiac death. Clin Pharmacol Ther 2004;75 :234–241

  42. U.S. Department of Labor. Bureau of Labor Statistics. Consumer Price Indexes [on-line]. Available at: http://data.bls.gov/cgi-bin/surveymost Accessed June 12, 2006

  43. Briggs AH. Handling uncertainty in cost-effectiveness models. Pharmacoeconomics 2000;17 :479–500

  44. Deshpande MA, Holden RR, Gilron I. The impact of therapy on quality of life and mood in neuropathic pain: What is the effect of pain reduction? Anesth Analg 2006;102 :1473–1479

  45. Steinman MA, Bero LA, Chren MM et al. Narrative review: The promotion of gabapentin: An analysis of internal industry documents. Ann Intern Med 2006;145 :284–293

  46. Berger A, Dukes EM, Edelsberg J et al. Use of tricyclic antidepressants in older patients with painful neuropathies. Eur J Clin Pharmacol 2006;62 :757–764

  47. Drugstore.com [on-line]. Available at: http://www.drugstore.com Accessed September 7, 2006

  48. RxUSA [on-line]. Available at http://rxusa.com Accessed September 7, 2006

  49. Melander H, Ahlqvist-Rastad J, Meijer G et al. Evidence-based medicine—selective reporting from studies sponsored by pharmaceutical industry: Review of studies in new drug applications. BMJ 2003;326 :1171–1175