 |
 |
|
1 January 1998 | Volume 128 Issue 1 | Pages 37-48
Purpose: To examine the cost-effectiveness of test-treatment strategies for patients suspected of having Lyme disease.
Data Sources: The medical literature was searched for information on outcomes and costs. Expert opinion was sought for information on utilities.
Study Selection: Articles that described patient population, diagnostic criteria, dose and duration of therapy, and criteria for assessment of outcomes.
Data Extraction: The decision analysis evaluated the following strategies: 1) no testing-no treatment; 2) testing with enzyme-linked immunosorbent assay [ELISA] followed by antibiotic treatment of patients with positive results; 3) two-step testing with ELISA followed by Western blot and antibiotic treatment for patients with positive results on either test; and 4) empirical antibiotic therapy. Three patient scenarios were considered: myalgic symptoms, rash resembling erythema migrans, and recurrent oligoarticular inflammatory arthritis. Results were calculated as costs per quality-adjusted life-year and were subjected to sensitivity analysis. Adjustment was made for the diagnostic value of common clinical features of Lyme disease.
Data Synthesis: For myalgic symptoms without other features suggestive of Lyme disease, the no testing-no treatment strategy was most economically attractive (that is, had the most favorable cost-effectiveness ratio). For rash, empirical antibiotic therapy was less costly and more effective than other strategies. For oligoarticular arthritis with a history of rash and tick bite, two-step testing was associated with the lowest cost-effectiveness ratio. Testing with ELISA and empirical antibiotic therapy cost an additional $880 000 and $34 000 per quality-adjusted life-year, respectively. For oligoarticular arthritis with one or no other features suggestive of Lyme disease, two-step testing was most economically attractive.
Conclusions: Neither testing nor antibiotic treatment is cost-effective if the pretest probability of Lyme disease is low. Empirical antibiotic therapy is recommended if the pretest probability is high, and two-step testing is recommended if the pretest probability is intermediate.
We performed a cost-effectiveness analysis to guide selection of test-treatment strategies for patients suspected of having Lyme disease. Incremental cost-effectiveness ratios were calculated for four possible strategies applied to three common patient scenarios. Sensitivity analyses assessed the robustness of the results to changes in the values or assumptions incorporated into the analysis.
We considered four test-treatment strategies: 1) no testing-no treatment, in which no testing was done and no antibiotic treatment was given for Lyme disease; 2) testing with enzyme-linked immunosorbent assay [ELISA] followed by treatment of patients with positive results; 3) two-step testing with ELISA followed by Western blot of all specimens with equivocal or positive ELISA results and treatment of patients with positive results on either test; and 4) antibiotic treatment for all patients suspected of having Lyme disease.
Patient Scenarios
Three patient scenarios were considered (Table 1): myalgic symptoms (scenario A), rash resembling erythema migrans (scenario B), and recurrent oligoarticular inflammatory arthritis (scenario C). These scenarios represent common presentations for patients with Lyme disease [9] or clinical syndromes that may be confused with Lyme disease. ACADEMIA AND CLINIC
Test-Treatment Strategies for Patients Suspected of Having Lyme Disease: A Cost-Effectiveness Analysis
Lyme disease, the most common tick-borne disease in North America [1], is often misdiagnosed or treated inappropriately [2]. Most patients referred for assessment of suspected Lyme disease had other major diagnoses [3]. In a community with a low prevalence of disease, 60% of serologic tests for Lyme disease were performed as part of a battery of tests to investigate vague symptoms [4]. Eighty-three percent of surveyed physicians reported that they would treat a patient suspected of having Lyme disease even without positive serologic test results [5]. Patients who have positive results on serologic tests without objective clinical manifestations have received parenteral antibiotic therapy for suspected chronic Lyme disease [5]. Such unnecessary testing and treatment incur substantial costs [6] and adverse effects [7, 8]. The development and use of guidelines for the management of Lyme disease are necessary to decrease health care costs and increase quality of life for patients who may have Lyme disease.
Methods
Top
Methods
Discussion
Author & Article Info
References
Test-Treatment Strategies
|
Decision Model
A simple decision tree was constructed to represent the natural history of patients with possible Lyme disease [10] (Figure 1 , Figure 3). The model considered the incidence of Lyme disease in the community, sensitivity and specificity of laboratory tests, patient adherence to prescribed treatment, effectiveness and adverse effects of therapy, true presence or absence of Lyme disease, and sequelae of Lyme disease.
|
|
Data and Assumptions
Data were obtained by searching the MEDLINE database for English-language articles published from 1992 to 1996. We used the subject headings ‘Borrelia burgdorferi,’ ‘Borrelia infections,’ ‘Lyme disease,’ and ‘tick-borne diseases’ (Table 2). Most of the identified articles described case series. These articles were supplemented by information from the bibliographies of two recent cost-effectiveness analyses of treatment strategies for Lyme disease [6, 19].
|
Incidence of Lyme Disease
Pretest probability is the probability that a diagnosis will be made before new information is acquired [10]. In our analysis, the pretest probability of Lyme disease before testing or treatment was the annual incidence of new infection within a given area. We used community incidence data rather than prevalence data for two reasons. First, early Lyme disease does not protect against reinfection with B. burgdorferi [31]. In addition, the stage of Lyme disease influences the sensitivity of the tests [32]. A high annual incidence of Lyme disease, 1%, was used for all baseline analyses.
Diagnostic Value of Clinical Features
Physician-diagnosed erythema migrans is considered pathognomonic for Lyme disease, although not everyone who has Lyme disease develops erythema migrans [33] (Figure 2). A patient with endemic exposure to Lyme disease and characteristic erythema migrans meets the clinical definition of Lyme disease [34] and therefore has a high probability of having Lyme disease. However, fewer than 50% of patients with this rash report a history of tick bite [9, 33, 35-37]. Patients in an endemic area who present during the summer with a large expanding border of erythema at one or more sites are usually considered to have Lyme disease and are treated accordingly, even if they do not recall a tick bite. Therefore, the rash described in scenario B was considered typical of erythema migrans diagnosed by a physician.
|
Several dermatologic conditions can be confused with erythema migrans [38]. Although atypical forms of erythema migrans occur, all have expanding borders of erythema. The nature of the rash should therefore be determined clinically.
Some patients who present with possible late manifestations of Lyme disease recall a rash resembling erythema migrans or a tick bite [39]. Whether a patient recalls these features helps discriminate between Lyme disease and unrelated diagnoses [40]. If a person recalls having a rash resembling erythema migrans, the probability of Lyme disease increases; lack of such a history decreases the probability of Lyme disease. The ability of this information to discriminate between patients with and without Lyme disease is expressed as a likelihood ratio [41].
A cohort study [40] described the diagnostic value of clinical features associated with Lyme disease. If patients reported a rash resembling erythema migrans, the likelihood ratio for Lyme disease was 21.3 (95% CI, 8.7 to 27.3). If patients did not report such a rash, the likelihood ratio for Lyme disease was 0 (CI, 0 to 0.6). The presence or absence of a history of tick bite had likelihood ratios for Lyme disease of 3.6 (CI, 1.7 to 4.7) and 0.3 (CI, 0.05 to 0.8), respectively.
We adjusted the pretest probability of Lyme disease (that is, the community incidence) by using a Bayesian analysis, as follows [10]. The pretest probability was converted to an odds, multiplied by the appropriate likelihood ratio to obtain the post-test odds, and converted to the posterior probability. After we considered other features suggestive of Lyme disease, we used this probability of disease as the pretest probability of disease for the test-treatment strategies.
Because no available data describe the likelihood of Lyme disease in a patient who presents with a history of both rash resembling erythema migrans and tick bite, we calculated the likelihood ratio by multiplying the likelihood ratio for tick bite by that for rash resembling erythema migrans [10].
Diagnostic Tests
The sensitivity and specificity of tests for Lyme disease were calculated on the basis of pooled estimates from two studies ([11]; Centers for Disease Control and Prevention. Unpublished data). These estimates were calculated with a random-effects model [42].
Treatment Costs
For our analysis, we took the perspective of the health care system. Hospital charges were used as a proxy for cost because true costs were unavailable. All cost estimates were expressed in 1993 U.S. dollars; as necessary, these estimates were adjusted for inflation by using the health care component of the consumer price index (Table 2). A discount rate of 5% accounted for future costs and effects.
Treatment of early Lyme disease (that is, rash resembling erythema migrans) requires doxycycline, 100 mg twice daily for 30 days [13], with a single follow-up visit to a physician [12]. Treatment of late Lyme disease was assigned the cost of outpatient intravenous ceftriaxone therapy lasting 3 weeks [6].
Treatment of fibromyalgia requires amitryptiline, 25 mg/d [13]. Treatment of osteoarthritis requires naproxen, 500 mg twice daily [13]. In addition, a patient with a chronic illness visits a physician every 3 months [12]. Costs for other health states were derived from the literature.
The costs of each chronic state were calculated as follows. Annual costs were discounted over the estimated life expectancy of a person with the relevant condition after the appropriate disutilities were taken into account (see the example in the section on utilities below). Net cost was then calculated by adding the costs of tests, antibiotic therapy, and treatment for the chronic health condition. The costs associated with lost work as a result of illness and death and costs of unrelated future illness were not included in the analysis according to standards proposed by the Panel on Cost-Effectiveness in Health and Medicine, sponsored by the U.S. Public Health Service [43].
Sequelae of Lyme Disease
The analysis adjusted for cardiac, late neurologic, and arthritic complications of Lyme disease (Table 2). Cardiac complications almost always occur early in the course of the illness [9, 15-18]; if they are not treated, they last for about 3 months and then resolve [15, 16]. Neurologic manifestations involve the central or peripheral nervous system and result in residual memory difficulty or paresthesias [9, 21, 22, 27]. Refractory or untreated Lyme arthritis is more disabling than osteoarthritis and requires arthroscopic synovectomy [28, 44]. Compared with patients who have early Lyme disease, patients who have late Lyme disease are less likely to develop rheumatologic or neurologic sequelae.
Early Lyme disease (for example, erythema migrans) is usually treated with oral antibiotics for at least 2 weeks [9]. Late Lyme disease is usually treated with intravenous antibiotics for 2 to 4 weeks [25, 26, 45]. Oral antibiotics can be used to treat oligoarticular arthritis, but this strategy is associated with a risk for inadequate treatment of unrecognized neurologic involvement [45].
We adjusted the analysis for costs or illness incurred by side effects of antibiotic therapy. Major side effects, including pancytopenia, renal impairment, fever and chills, and anaphylactoid reaction, occurred in 0.1% of patients and lasted for 1 month [6]. Minor side effects, including diarrhea, upper gastrointestinal symptoms, and rash, occurred in 3% of patients and lasted for 2 weeks [6].
We obtained estimates of the probability of adherence to treatment and effectiveness of therapy from the literature (Table 2). Although treatment with intravenous antibiotics can alleviate Lyme disease-related arthritic or neurologic symptoms in a patient with diffuse myalgic symptoms, it does not alleviate fibromyalgic symptoms [2, 46, 47].
Utilities
Three authors constructed scenarios that described the symptoms and functional abilities of hypothetical patients with myalgic symptoms (scenario A), rash resembling erythema migrans (scenario B), recurrent oligoarticular arthritis (scenario C), or late neurologic involvement (details available from the authors). The description of each patient was based on Torrance's Health Utility Index [48].
An expert panel independently completed a time-tradeoff task for each scenario. For intermittent health states (such as recurrent oligoarticular inflammatory arthritis), the scenario described a lifetime with intermittent symptoms. The utility for each chronic health state was calculated as the mean of the values provided by the expert panel.
To account for disutilities for short-term health states, such as treatment, side effects, or cardiac complications, we assumed that the utility was 0 for the duration of that short-term state [49]. For example, a hypothetical patient's life expectancy was 29.1 years without treatment or complications. The patient presented with erythema migrans, received antibiotic treatment for 3 weeks, and had cardiac complications for 3 months. The patient's net life expectancy was thus calculated as 28.8 years (29.1 years –[3 weeks and 3 months]).
Life Expectancy
Our base case considered testing or treating a 50-year-old person (Table 2). We modeled for 50 years of age because this was the approximate mean age of the U.S. population and we could not quantify the age distribution of patients with Lyme disease [30].
No empirical data describe the mortality rate for patients with Lyme disease. We assumed that neither Lyme arthritis nor fibromyalgic symptoms decrease life expectancy. Because patients frequently receive therapies to prevent neurologic Lyme disease that are expensive and associated with risk for substantial side effects, we assumed that neuroborreliosis increased mortality rates as much as partial paraplegia does [24].
We estimated net life expectancy for each health state by adding the age-specific and disease-specific mortality rates, inverting the result, and subtracting any disutilities associated with short-term health states [49]. We then made quality adjustments for the presence of any chronic symptoms by multiplying life expectancy by utilities obtained from the expert panel.
Baseline Cost-Effectiveness Analyses
Incremental cost-effectiveness ratios were calculated as costs per quality-adjusted life-year (QALY) for a patient with myalgic symptoms (scenario A), rash resembling erythema migrans (scenario B), or recurrent oligoarticular inflammatory arthritis (scenario C). The pretest probability of Lyme disease was adjusted by considering other features that were suggestive of Lyme disease, as described above. Therefore, incremental cost-effectiveness ratios were calculated for presentations with and without a history of tick bite or rash resembling erythema migrans for scenarios A and C and for presentations with typical or atypical rash resembling erythema migrans for scenario B.
Incremental cost-effectiveness ratios were calculated for each scenario as follows [10]. Strategies were ranked from lowest to highest total cost. Incremental cost was calculated by subtracting the cost of one strategy from that of the next most expensive strategy. Incremental effectiveness was calculated by subtracting the effectiveness of one strategy from that of the next most expensive strategy. The average cost-effectiveness ratio was calculated by dividing total cost by total effectiveness. The incremental cost-effectiveness ratio was then calculated by dividing incremental costs by incremental effectiveness. A strategy was considered more economically attractive than another if it was associated with an incremental cost-effectiveness ratio less than $50 000 per QALY. Conversely, a strategy was considered dominated by another if it cost more and was less effective.
Sensitivity Analyses
We assessed the robustness of the results to changes in the model by substituting a plausible range of values for each variable (Table 2). The value of clinical variables (such as complication rate, adherence to treatment regimen, and effectiveness) was varied across the range of published values.
We varied the life expectancy of the patient under consideration in two ways. First, the life expectancy of a 1-year-old or an 85-year-old person was substituted into the decision model [30]. In addition, the reduction in lifespan caused by neuroborreliosis was assigned a disease-specific mortality rate equal to zero or that of a stroke [24].
The sensitivity of the results to changes in quality of life was assessed as follows. For chronic health states, we substituted utilities of 0.5 to 1. For short-term health states (adverse effect of treatment or cardiac complications of Lyme disease), we substituted utilities of 0.95 (as opposed to baseline values of 0).
The diagnostic value of features suggestive of Lyme disease was varied by substituting the lower and upper 95% CIs of the likelihood ratio for each feature into the decision model.
Economic variables in the model were varied in two ways. We varied the discount rate from 0% to 10% to account for possible differences in time preference. We also examined the results of the analysis if all costs were doubled.
Finally, a structural sensitivity analysis evaluated the cost-effectiveness of the no testing-no treatment strategy or testing if a proportion of untested patients were treated empirically, because patients with a low pretest probability of Lyme disease frequently receive empirical treatment with antibiotics.
Unnecessary Treatments Avoided
For late presentations of possible Lyme disease, we calculated the number of unnecessary treatments avoided by using the preferred (that is, the most economically attractive) strategy rather than empirical antibiotic therapy.
Data Synthesis
|
|---|
We created four clinical presentations for patients with the lowest probability of Lyme disease, beginning with myalgia alone (scenario A) and adding one or more potentially relevant clinical findings (rash or tick bite) (Table 3). Similar clinical presentations were created for patients with an intermediate probability of Lyme disease (scenario C). Finally, we substituted the likelihood ratio for rash recognized by patients [40] into scenario B to assess test-treatment strategies in a patient with atypical rash that was not considered by a physician to be indicative of Lyme disease. For each scenario, we adjusted the pretest probability of Lyme disease.
|
For a patient who had myalgic symptoms without rash, the no testing-no treatment strategy dominated the other test-treatment strategies. For a patient who had myalgic symptoms and rash but no tick bite, the no testing-no treatment strategy was associated with the lowest cost-effectiveness ratio. Two-step testing cost an additional $7000 per QALY. Other strategies cost more and were less effective (that is, they were dominated.) For a patient who had myalgic symptoms and tick bite but no rash, the no testing-no treatment strategy dominated all other strategies. For a patient who had myalgic symptoms and a history of rash and tick bite, two-step testing was associated with the lowest cost-effectiveness ratio. In this scenario, other strategies were dominated by the two-step testing strategy or were associated with large incremental cost-effectiveness ratios. The difference between the effectiveness of two-step testing and the no testing-no treatment strategy was less than 5% for any analysis of patients with myalgic symptoms.
For a patient who had a typical or atypical rash resembling erythema migrans, empirical therapy with antibiotics was less costly and more effective than any other test-treatment strategy.
For a patient with oligoarticular arthritis and no or one other feature suggestive of Lyme disease, two-step testing dominated other strategies or was associated with an incremental cost of less than $10 000 per QALY. For a patient with oligoarticular arthritis and a history of rash and tick bite, the two-step testing strategy was associated with the lowest cost-effectiveness ratio; other strategies were dominated by two-step testing or were associated with large incremental cost-effectiveness ratios. The difference between the effectiveness of two-step testing and of no testing-no treatment was less than 7% for any analysis of patients with oligoarticular arthritis.
Sensitivity Analyses
For most of the scenarios, changes in the values of variables influenced the magnitude of the cost-effectiveness ratios but did not affect the ranking of the test-treatment strategies or their relative economic attractiveness. In particular, the results did not change when the analysis considered younger patients or lack of death associated with neuroborreliosis, when the utilities assigned to short-term or chronic health states were changed, or when costs were increased. For example, for young patients who had myalgic symptoms with or without other features suggestive of Lyme disease, empirical antibiotic therapy was not economically attractive even if the patient was 1 year of age and no mortality was associated with neuroborreliosis.
Extreme values for the sensitivity and specificity of diagnostic tests, incidence of neurologic sequelae, short life expectancy (such as in older patients), and discount rate influenced which strategy was most economically attractive for a few scenarios. However, no single variable affected the cost-effectiveness of the test-treatment strategies in all scenarios.
The relation between the community incidence of Lyme disease and the pretest probability of Lyme disease with various clinical presentations is summarized in Table 4. The effectiveness of the two-step testing and no testing-no treatment strategies at any incidence of Lyme disease differed little. However, the results of the cost-effectiveness analysis were partly sensitive to changes in the incidence of Lyme disease. For a patient with myalgic symptoms and rash, two-step testing cost an additional $50 000 per QALY compared with no testing-no treatment if the community incidence was less than 0.2%; this result contrasts with $7000 per QALY for the base-case incidence of 1% (Table 3). For a patient with oligoarticular arthritis, rash, and tick bite, testing by ELISA cost less than $50 000 per QALY compared with two-step testing if the incidence was greater than 3.5%; this result contrasts with $880 000 per QALY for the base case. This incidence is greater than the incidence in all but hyperendemic regions [1]. For the other scenarios, changes in incidence did not influence the results of the analysis.
|
The results of the analysis were sensitive to changes in the diagnostic value of some features suggestive of Lyme disease. For example, changes in the likelihood ratio for Lyme disease in the absence of rash influenced which strategy was most economically attractive. If the likelihood ratio for Lyme disease in the absence of tick bite was as much as 0.6, two-step testing for a patient who had myalgic symptoms and rash but no tick bite cost $23 000 per QALY more than no testing-no treatment; this result contrasts with two-step testing being dominated for the base-case likelihood ratio of 0. If the likelihood ratio for Lyme disease in the absence of tick bite was as much as 0.6, then two-step testing for a patient with recurrent oligoarticular arthritis and rash but no tick bite cost $1200 per QALY more than no testing-no treatment. For other scenarios, changes in the likelihood ratio for tick bite did not influence the results of the analysis.
When the analysis included empirical treatment of a proportion of untested patients with myalgic symptoms who did or did not have other features associated with Lyme disease, no testing-no treatment or two-step testing was the preferred strategy, regardless of the proportion of untested patients who would be empirically treated.
In summary, the incremental cost-effectiveness ratio were relatively robust to one-way sensitivity analyses, with the exception of changes in the incidence of Lyme disease or in the likelihood of Lyme disease without rash resembling erythema migrans.
Unnecessary Treatments Avoided
For late presentations of possible Lyme disease, no testing-no treatment or two-step testing avoided 552 to 990 unnecessary treatments and unnecessary expenditures of $1.85 million to $3.33 million per 1000 patients compared with empirical treatment of all such patients (Table 5).
|
Discussion
|
|---|
|
TopMethodsDiscussionAuthor & Article InfoReferences |
|---|
For a patient with myalgic symptoms or oligoarticular arthritis who did not have other features suggestive of Lyme disease, two-step testing was not economically attractive compared with no testing-no treatment. Changes in the incidence of Lyme disease or the likelihood ratio for the absence of a history of rash influenced which test-treatment strategy was most economically attractive.
For a patient with rash resembling erythema migrans, empirical antibiotic therapy was still the preferred therapy when the less specific finding of patient-recognized rash was substituted for the more specific finding of physician-recognized rash.
On the basis of our analysis, we recommend the following approaches to the management of patients suspected of having Lyme disease. First, we recommend neither testing nor antibiotic treatment for a patient with low probability of disease (such as a patient with myalgic symptoms without other possible symptoms or sequelae). Although the differences in the costs and effectiveness of the strategies considered were small, no evidence suggests that treatment changes the outcome [2, 46, 47]. Second, we recommend two-step testing for a patient with an intermediate probability of disease (such as a patient with recurrent oligoarticular arthritis or other possible late sequelae). Although the differences among the strategies were small, treatment does change the outcome [45]. Finally, we recommend empirical antibiotic therapy without serologic testing for a patient with a high probability of disease (such as a patient with rash resembling erythema migrans or other symptoms strongly suggestive of early Lyme disease).
Our analysis represents an example of the threshold approach to clinical decision making [50]. The physician's estimate of the probability that a particular patient has Lyme disease determines whether withholding treatment, testing, or administering treatment is optimal (Table 4). If the probability of Lyme disease is low, the side effects accrued by treatment of many patients who do not have Lyme disease outweigh any benefits of treating the few patients who do. Conversely, if the probability of Lyme disease is high, the benefit of treating the many patients who have Lyme disease outweighs the side effects accrued by the few patients who do not. Otherwise, testing is the best choice.
For patients with nonspecific symptoms and a positive result on ELISA, Lightfoot and colleagues [6] found that empirical treatment with antibiotics is not cost-effective. Although a negative test result effectively rules out Lyme disease in a patient with a low pretest probability of disease, a positive result does not increase the probability of Lyme disease enough to justify treatment. Our analysis expands on this research in several ways. First, we considered strategies before testing rather than only in the presence of a positive test result. Second, we accounted for the diagnostic value of other clinical features suggestive of Lyme disease. Finally, values for sensitivity and specificity were pooled, whereas Lightfoot and colleagues assumed a sensitivity of 100%. Our analysis suggests that neither testing nor antibiotic treatment is cost-effective for patients with nonspecific symptoms.
For patients with tick bite and possible early manifestations of Lyme disease, a previous analysis [19] found that empirical treatment with antibiotics was cost-effective compared with no treatment or expectant management. Because most patients do not recall tick bites, we considered patients who present with other possible early manifestations of Lyme disease (such as rash). On the basis of our analysis, empirical treatment with antibiotics is recommended for such patients regardless of whether they have typical or atypical rash resembling erythema migrans or recall a tick bite.
Other researchers [51] have recommended paired acute- and convalescent-phase serologic testing if a patient with suspected early Lyme disease has negative results on serologic tests. We used clinical and seroepidemiologic criteria for the diagnosis of Lyme disease. We did not consider use of acute- and convalescent-phase serologic testing because data were not sufficient to evaluate this testing.
Shadick and colleagues [52] identified several longer-term musculoskeletal and neurologic complications of Lyme disease in a small case–control study. However, the relation between these complications and previous treatment is unclear. Without a larger study, it is premature to generalize their results to broad guidelines.
Two caveats accompany our recommendations. First, our analysis did not consider rare sequelae of Lyme disease, such as myositis [53] or neuroophthalmologic conditions [54, 55]. The incidence of these sequelae is unclear: For example, only 16 case reports of myositis associated with Lyme disease have been published [53, 56-63]. Until additional information is available, patients with rare sequelae should have two-step testing for Lyme disease.
Second, the relation between the incidence and prevalence of Lyme disease in a community remains uncertain. National surveillance for Lyme disease in the United States has identified county incidences of less than 5 per 100 000 persons to more than 400 per 100 000 persons [1, 64, 65]. Apart from selected regions of northeastern, upper north-central, and northern California, county incidences were generally 5 per 100 000 persons or less. The annual incidence in most communities in the United States is less than 1%. Because acute Lyme disease lasts about 3 months, the point prevalence of acute infection in an area with an annual incidence of 1% is only 0.25%. Conversely, because the cumulative prevalence is higher than 1%, the proportion of persons with chronic, recurrent disease is greater than 1%. Therefore, in our analysis, the pretest probability of Lyme disease was overestimated in patients with possible early manifestations of disease and underestimated in patients with possible late manifestations.
In endemic areas, 5% to 15% of persons are seropositive for Lyme disease. If myalgic symptoms develop in these persons, a positive test result does not mean that the symptoms are caused by active Lyme disease. In our analysis, testing was only cost-effective in patients with myalgic symptoms and other features suggestive of Lyme disease when the community incidence was greater than that reported in all but hyperendemic areas. It may be reasonable to use serologic testing in patients with myalgic symptoms in endemic or hyperendemic areas because the incremental costs of testing are small and physicians are under considerable pressure to offer such patients antibiotic therapy. Although testing may avoid the use of expensive and toxic therapy in patients with myalgic symptoms, these patients should not receive empirical antibiotic therapy.
Our analysis has several limitations. First, the model considered a 50-year-old person, even though the age of patients with possible Lyme disease varies widely. Although many patients with possible Lyme disease are younger than 50 years of age, the results did not change when the model considered younger patients. Second, we assumed that neurologic Lyme disease increases mortality. Although this assumption biases the analysis in favor of testing or treatment, the results support a more restricted approach of no testing-no treatment unless the patient has at least an intermediate probability of having Lyme disease. Third, utilities for chronic health states were derived from an expert panel. Other researchers [66] recommend that patient or public preferences be used, but the intermittent nature of the symptoms would require repeated measurements, which is impractical. Fourth, there are several ways to estimate disutilities for short-term health states. We adopted the standard approach of assigning zero utility for the duration of any short-term state [49]. Sensitivity analyses showed that changes in utilities for chronic or short-term health states did not influence which strategy was most cost-effective. Finally, we did not consider indirect costs, such as productivity lost because of illness.
Our analyses are limited by their use of imperfect data and assumptions. Some of the incremental cost-effectiveness ratios do not differ greatly. We recommend ongoing prospective assessment of diagnostic tests for Lyme disease in a large multi-center cohort study. Test-treatment guidelines should be reevaluated as serodiagnostic tests are improved and newer tests, such as polymerase chain reaction, are developed.
Dr. Dennis: Centers for Disease Control and Prevention, Box 2087, Fort Collins, CO 80521.
Dr. Steere: Rheumatology/Immunology, New England Medical Center #406, 750 Washington Street, Boston, MA 02111.
Dr. Lightfoot: Division of Rheumatology, University of Kentucky, Kentucky Clinic J511, Lexington, KY 40536-0284.
Dr. Tugwell: Department of Medicine, Ottawa General Hospital, 501 Smyth Road, Ottawa, Ontario K1H 8L6, Canada.
Author and Article Information
|
|---|
|
TopMethodsDiscussionAuthor & Article InfoReferences |
|---|
References
|
|---|
|
TopMethodsDiscussionAuthor & Article InfoReferences |
|---|
1. Lyme disease-United States, 1995. MMWR Morb Mortal Wkly Rep. 1996; 45:481-4.
2. Steere AC, Taylor E, McHugh GL, Logigian EL. The overdiagnosis of Lyme disease. JAMA. 1993; 269:1812-6.
3. Burdge DR, O'Hanlon DP. Experience at a referral center for patients with suspected Lyme disease in an area of nonendemicity: first 65 patients. Clin Infect Dis. 1993; 16:558-60.
4. Ley C, Le C, Olshen EM, Reingold AL. The use of serologic tests for Lyme disease in a prepaid health plan in California. JAMA. 1994; 271:460-3.
5. Eppes SC, Klein JD, Caputo GM, Rose CD. Physicians beliefs, attitudes, and approaches toward Lyme disease in an endemic area. Clin Pediatr (Phila). 1994; 33:130-4.
6. Lightfoot RW Jr, Luft BJ, Rahn DW, Steere AC, Sigal LH, Zoschke DC, et al. Empiric parenteral antibiotic treatment of patients with fibromyalgia and fatigue and a positive serologic result for Lyme disease. A cost-effectiveness analysis. Ann Intern Med. 1993; 119:503-9.
7. Ettestad PJ, Campbell GL, Welbel SF, Genese CA, Spitalny KC, Marchetti CM, et al. Biliary complications in the treatment of unsubstantiated Lyme disease. J Infect Dis. 1995; 171:356-61.
8. Ceftriaxone-associated biliary complications of treatment of suspected disseminated Lyme disease-New Jersey, 1990-1992. MMWR Morb Mortal Wkly Rep. 1993; 42:39-42.
9. Steere AC. Lyme disease. N Engl J Med. 1989; 321:586-96.
10. Sox HC Jr, Blatt MA, Higgins MC, Marton KI. Medical Decision Making. Boston: Butterworths; 1988.
11. Dressler F, Whalen JA, Reinhardt BN, Steere AC. Western blotting in the serodiagnosis of Lyme disease. J Infect Dis. 1993; 167:392-400.
12. Center for Health Policy Research (American Medical Association). Physician Marketplace Statistics: Profiles for Detailed Specialties, Selected States and Practice Arrangements. Chicago: American Med Assoc; 1993.
13. Drug Topics Red Book. Montvale, NJ: Medical Economics; 1993.
14. Cockburn J, Gibberd RW, Reid AL, Sanson-Fisher RW. Determinants of non-compliance with short term antibiotic regimens. Br Med J (Clin Res Ed). 1987; 295:814-8.
15. McAlister HF, Klementowicz PT, Andrews C, Fisher JD, Feld M, Furman S. Lyme carditis: an important cause of reversible heart block. Ann Intern Med. 1989; 110:339-45.
16. Steere AC, Batsford WP, Weinberg M, Alexander J, Berger HJ, Wolfson S, et al. Lyme carditis: cardiac abnormalities of Lyme disease. Ann Intern Med. 1980; 93:8-16.
17. Melski JW, Reed KD, Mitchell PD, Barth GD. Primary and secondary erythema migrans in central Wisconsin. Arch Dermatol. 1993; 129:709-16.
18. Rubin DA, Sorbera C, Nikitin P, McAllister A, Wormser GP, Nadelman RB. Prospective evaluation of heart block complicating early Lyme disease. Pacing Clin Electrophysiol. 1992; 15:252-5.
19. Magid D, Schwartz B, Craft J, Schwartz JS. Prevention of Lyme disease after tick bites. A cost-effectiveness analysis. N Engl J Med. 1992; 327:534-41.
20. Steere AC, Malawista SE, Newman JH, Speiler PN, Bartenhagen NH. Antibiotic therapy in Lyme disease. Ann Intern Med. 1980; 93:1-8.
21. Reik L, Steere AC, Bartenhagen NH, Shope RE, Malawista SE. Neurologic abnormalities of Lyme disease. Medicine (Baltimore). 1979; 58:281-94.
22. Pachner AR, Steere AC. The triad of neurologic manifestations of Lyme disease: meningitis, cranial neuritis, and radiculoneuritis. Neurology. 1985; 35:47-53.
23. Giesler WO, Jousse AT, Wynne-Jones M. Survival in traumatic transverse myelitis. Paraplegia. 1977; 14:262-75.
24. Lew EA, Gajewski J, eds. Medical Risks: Trends in Mortality by Age and Time Elapsed: A Reference Volume. New York: Praeger; 1990.
25. Dattwyler R, Halperin J, Volkman D, Luft B. Treatment of late Lyme borreliosis-randomised comparison of ceftriaxone and penicillin. Lancet. 1988; 1:1191-4.
26. Karlsson M, Hammers-Berggren S, Stiernstedt G, Svenungsson B. Comparison of intravenous penicillin G and doxycycline for treatment of Lyme neuroborreliosis. Neurology. 1994; 44:1203-7.
27. Logigian EL, Steere AC. Clinical and electrophysiologic findings in chronic neuropathy of Lyme disease. Neurology. 1992; 42:303-11.
28. Steere AC, Schoen RT, Taylor E. The clinical evolution of Lyme arthritis. Ann Intern Med. 1987; 107:725-31.
29. Lui NY, Dinerman H, Levin RE. Randomized trial of doxycycline vs amoxicillin/probenecid for the treatment of Lyme arthritis: treatment of non-responders with IV penicillin or ceftriaxone. Arthritis Rheum. 1989; 32:S46.
30. Annual Summary of Births, Marriages, Divorces and Deaths. Washington, DC: National Center for Health Statistics; 1994.
31. Gustafson R, Svenungsson B, Forsgren M, Gardulf A, Granstrom M. Two-year survey of the incidence of Lyme borreliosis and tick-borne encephalitis in a high-risk population in Sweden. Eur J Clin Microbiol Infect Dis. 1992; 11:894-900.
32. Tugwell P, Dennis DT, Weinstein A, Wells G, Shea B, Nichol G, et al. Laboratory evaluation in the diagnosis of Lyme disease. Ann Intern Med. 1997; 127:1109-23.
33. Steere AC, Bartenhagen NH, Craft JE, Hutchinson GJ, Newman JH, Rahn DW, et al. The early clinical manifestations of Lyme disease. Ann Intern Med. 1983; 99:76-82.
34. Wharton M, Chorba TL, Vogt RL, Morse DL, Buehler JW. Case definitions for public health surveillance. MMWR Morb Mortal Wkly Rep. 1990; 39:1-43.
35. Burgdorfer WA, Barbour AG, Hayes SF. Lyme disease-a tick-borne spirochetosis? Science. 1982; 216:1317-9.
36. Berger BW. Dermatologic manifestations of Lyme disease. Rev Infect Dis. 1989; 11(Suppl 6):S1475-81.
37. Jung PI, Nahas JN, Strickland GT, McCarter R, Israel E. Maryland physicians' survey on Lyme disease. Md Med J. 1994; 43:447-50.
38. Malane MS, Grant-Kels JM, Feder HM Jr, Luger SW. Diagnosis of Lyme disease based on dermatologic manifestations. Ann Intern Med. 1991; 114:490-8.
39. Steere AC, Taylor E, Wilson ML, Levine JF, Spielman A. Longitudinal assessment of the clinical and epidemiological features of Lyme disease in a defined population. J Infect Dis. 1986; 154:295-300.
40. Blaauw I, Nohlmans L, van den Bogaard T, van der Linden S. Diagnostic tools in Lyme borreliosis: clinical history compared with serology. J Clin Epidemiol. 1992; 45:1229-36.
41. Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical Epidemiology: A Basic Science for Clinical Medicine. 2d ed. Boston: Little, Brown; 1991.
42. DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled Clin Trials. 1986; 7:177-88.
43. Gold MR, Siegel JE, Russell LB, Weinstein MC, eds. Cost-Effectiveness in Health and Medicine. New York: Oxford Univ Pr; 1996.
44. Schoen RT, Aversa JM, Rahn DW, Steere AC. Treatment of refractory chronic Lyme arthritis with arthroscopic synovectomy. Arthritis Rheum. 1991; 34:1056-60.
45. Steere AC, Levin RE, Molloy PJ, Kalish RA, Abraham JH 3d, Liu NY, et al. Treatment of Lyme arthritis. Arthritis Rheum. 1994; 37:878-88.
46. Hsu VM, Patella SJ, Sigal LH. "Chronic Lyme disease" as the incorrect diagnosis in patients with fibromyalgia. Arthritis Rheum. 1993; 36:1493-500.
47. Dinerman H, Steere AC. Lyme disease associated with fibromyalgia. Ann Intern Med. 1992; 117:281-5.
48. Torrance G. Utility approach to measuring health-related quality of life. J Chronic Dis. 1987; 40:593-603.
49. Beck JR, Pauker SG, Gottlieb JE, Klein K, Kassirer JP. A convenient approximation of life expectancy (the "DEALE"). II. Use in medical decision-making. Am J Med. 1982; 73:889-97.
50. Pauker SG, Kassirer JP. The threshold approach to clinical decision making. N Engl J Med. 1980; 302:1109-17.
51. Recommendations for test performance and interpretation from the Second National Conference on Serologic Diagnosis of Lyme Disease. MMWR Morb Mortal Wkly Rep. 1995; 44:590-1.
52. Shadick NA, Phillips CB, Logigian EL, Steere AC, Kaplan RF, Berardi VP, et al. The long-term clinical outcomes of Lyme disease. A population-based retrospective cohort study. Ann Intern Med. 1994; 121:560-7.
53. Horowitz HW, Sanghera K, Goldberg N, Pachman D, Kamer R, Duray P, et al. Dermatomyositis associated with Lyme disease: a case report and review of Lyme myositis. Clin Infect Dis. 1994; 18:166-71.
54. Karma A, Seppala I, Mikkila H, Kaakkola S, Viljanen M, Tarkkanen A. Diagnosis and clinical characteristics of ocular Lyme borreliosis. Am J Ophthalmol. 1995; 119:127-35.
55. Smith JL, Winward KE, Nicholson DF, Albert DW. Retinal vasculitis in Lyme borreliosis. J Clin Neuroophthamol. 1991; 11:7-15.
56. Atlas E, Novak SN, Duray PH, Steere AC. Lyme myositis; muscle invasion by Borrelia burgdorferi infection. Ann Intern Med. 1988; 109:245-6.
57. Schoenen J, Sianard-Gainko J, Carpentier M, Reznik M. Myositis during Borrelia burgdorferi infection (Lyme disease). J Neurol Neurosurg Psychiatry. 1989; 52:1002-5.
58. Hoffman JC, Stichtenoth DO, Zeidler H, Follmann M, Brandis A, Stanek G, et al. Lyme disease in a 74-year-old forest owner with symptoms of dermatomyositis. Arthritis Rheum. 1995; 38:1157-60.
59. Schmutzhard E, Willeit J, Gerstenbrand F. Meningopolyneuritis Bannwarth with focal nodular myositis. A new aspect in Lyme borreliosis. Klin Wochenschrift. 1986; 64:1204-8.
60. Kengen RA, v.d. Linde M, Sprenger HG, Piers DA. Scintigraphic evidence of Lyme disease: gallium-67 imaging of Lyme myositis. Clin Nucl Med. 1989; 14:728-9.
61. Seidenberg KB, Leib ML. Orbital myositis with Lyme disease. Am J Ophthamol. 1990; 109:13-6.
62. Fraser DD, Kong LI, Miller FW. Molecular detection of persistent Borrelia burgdorferi in a man with dermatomyositis. Clin Exp Rheumatol. 1992; 10:387-90.
63. Reimers CD, de Koning J, Neubert U, Preac-Mursic V, Koster JG, Muller-Felber W, et al.Borrelia burgdorferi myositis: report of 8 patients. J Neurol. 1993; 240:278-83.
64. Lyme Disease-United States, 1993. MMWR Morb Mortal Wkly Rep. 1994; 43:564-5.
65. Lyme disease-United States, 1994. MMWR Morb Mortal Wkly Rep. 1995; 44:459-62.
66. Jacobs P, Bachynsky J, Baladi JF. A comparative review of pharmacoeconomic guidelines. PharmacoEconomics. 1995; 8:182-9.
This article has been cited by other articles:
|
|
 |
|
  C. D. Tibbles and J. A. Edlow Does This Patient Have Erythema Migrans? JAMA, June 20, 2007; 297(23): 2617 - 2627. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. E. Aguero-Rosenfeld, G. Wang, I. Schwartz, and G. P. Wormser Diagnosis of Lyme Borreliosis Clin. Microbiol. Rev., July 1, 2005; 18(3): 484 - 509. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. P. Smith, R. T. Schoen, D. W. Rahn, V. K. Sikand, J. Nowakowski, D. L. Parenti, M. S. Holman, D. H. Persing, and A. C. Steere Clinical Characteristics and Treatment Outcome of Early Lyme Disease in Patients with Microbiologically Confirmed Erythema Migrans Ann Intern Med, March 19, 2002; 136(6): 421 - 428. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. C. Steere Lyme Disease N. Engl. J. Med., July 12, 2001; 345(2): 115 - 125. [Full Text] [PDF]
|
|
|
|
 |
|
 N. A. Shadick, M. H. Liang, C. B. Phillips, K. Fossel, and K. M. Kuntz The Cost-effectiveness of Vaccination Against Lyme Disease Arch Intern Med, February 26, 2001; 161(4): 554 - 561. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. A. M. Blaauw, A. M. van Loon, J. F. P. Schellekens, and J. W. J. Bijlsma Clinical evaluation of guidelines and two-test approach for Lyme disease Rheumatology, November 1, 1999; 38(11): 1121 - 1126. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. J. Masters, K. B. Kirkland, and D. T. Dennis Erythema Migrans in the South Arch Intern Med, October 26, 1998; 158(19): 2162 - 2165. [Full Text] [PDF]
|
|
|
|
|
|
R. C. Bransfield, D. Hassler, M. Maiwald, T. N. Petney, M. A. Piras, E. M. Porqueddu, P. Porcu, A. Aceti, R. E. Anderson, A. D. Fix, et al. Diagnosis, Treatment, and Prevention of Lyme Disease JAMA, September 23, 1998; 280(12): 1049 - 1051. [Full Text] [PDF]
|
|
|
|
 |
|
 P. T. Fawcett, C. D. Rosé, K. M. Gibney, and R. A. Doughty Comparison of Immunodot and Western Blot Assays for Diagnosing Lyme Borreliosis Clin. Vaccine Immunol., July 1, 1998; 5(4): 503 - 506. [Abstract] [Full Text]
|
|
|
|
 |
|
 A Lyme Decision Tree Journal Watch Infectious Diseases, March 1, 1998; 1998(301): 18 - 18. [Full Text]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Article
