Approximately 20% of patients with later manifestations of Lyme disease recall no history of tick bite or erythema chronicum migrans [11]. Similarly, patients with Lyme disease arthritis may have no history of earlier symptoms [6]. Although the skin rash and flu-like symptoms of early Lyme disease (stage 1) can be adequately treated in most instances with oral antibiotics [12, 13], later disease symptoms clearly require more aggressive antibiotic therapy [2, 4, 10, 14, 15].

Because the etiologic agent, Borrelia burgdorferi, is difficult to culture, the diagnosis of Lyme disease is based on the occurrence of [1] one or more of the classical features and [2] serum antibodies to the etiologic spirochete after the first 4 to 6 weeks of illness. This has resulted in increased testing for antibodies to B. burgdorferi, the causative agent of Lyme disease, in many patients with nonspecific somatic complaints or fatigue. Increasing numbers of such patients, who completely lack any of the classic clinical features for Lyme disease, are receiving empirical, parenteral antibiotic treatment for suspected late-stage, chronic Lyme disease [16-21]. This treatment consists of 2 to 3 weeks of daily intravenous cephalosporin (ceftriaxone, Rocephin; Roche Laboratories, Nutley, New Jersey) therapy costing between $3000 and $5000.

Although little doubt exists that a 2- to 4-week course of intravenous antibiotic therapy can often cure unequivocal, objective symptoms of disseminated (second-stage) and chronic (third-stage) Lyme disease, we found no reports in the peer-reviewed literature suggesting effectiveness of empirical, intravenous antibiotic treatment in patients with nonspecific complaints and a positive antibody titer for Lyme disease. Several reports exist indicating a lack of efficacy of this therapy [16-21]. Nonetheless, many patients with only nonspecific symptoms are given therapy empirically for months. A most striking example is the report of Genese and colleagues [20] of patients given as many as seven courses of intravenous antibiotic therapy. Twenty-five patients who received this therapy for as long as 170 days developed biliary disease, and 14 of these patients needed cholecystectomy. Twenty-two patients in the same cohort developed 29 bloodstream infections associated with intravenous catheter use after a median of 152.5 days of catheterization (range, 16 to 764 days) [20].

Our report examines the risks and benefits of two alternative therapeutic interventions: 1) empirical intravenous antibiotic treatment of the nonspecifically symptomatic patient who has a positive Lyme antibody titer or 2) no antibiotic treatment of such a patient. The analysis indicates that even in a best-case scenario, the risks and dollar costs of empirical therapy are substantially greater than the benefits.

Methods

Top Methods Results Discussion Conclusion Author & Article Info References

To evaluate the relative benefits and costs of the above two alternatives, the following procedures were developed. Data on the epidemiology of Lyme disease, the performance of serologic tests for the disease and the response to parenteral therapy were obtained by computerized search of peer-reviewed literature.

Serologic Tests for Lyme Disease

No standardized method exists for titering serum antibodies to B. burgdorferi. The two most commonly used methods include enzyme-linked immunosorbent assay (ELISA) and indirect immunofluorescent assay using crude homogenates of the etiologic organism [19-26]. Because the ELISA method was developed using standards from the indirect immunofluorescence assay, it behaves similarly to the latter, and it is the method typically used now, we have not considered the latter assay in this analysis.

Most investigators have considered a titer by ELISA to be positive if it exceeds by two standard deviations the mean for a group of normal sera [22-26]. Because the serologic test essentially defines the patients affected by this analysis, it is appropriate to examine both the sensitivity and the specificity of this test in different groups of patients. If one assumes a relatively normal distribution of titers in the normal population, then 2.5% of that normal population would be expected to exceed the mean by two standard deviations and thus would have false-positive results. If properly done, the test will detect cross-reactive antibody in patients with other spirochetal diseases. False-positive results have also been reported in varicella infections, infectious mononucleosis, rheumatoid arthritis, and systemic lupus erythematosus. However, after eliminating positive results due to cross-reactivity, published reports indicate a specificity in the ELISA assay of between 97% and 100%. In our baseline analysis, we assumed 100% sensitivity (because it is a positive titer that identifies patients for inclusion in this analysis in the first place) and a specificity of 98% (that is, 2% false positivity). Although investigators anticipate that using purified spirochetal epitopes as antigen in the ELISA in the future will greatly increase specificity, these epitopes are not available.

Although Western blot testing is available for a nominal sum, it has not been standardized, and its specificity is only 95%. The specificity of antibody reactivity with any of the several possible epitopes has not been established, and considerable disagreement exists about whether the number of lines, the position of lines, or the intensity of lines in Western blots distinguish true- from false-positive results [27, 28]. In Western blots, many of the antibodies in sera from true-positive patients recognize antigenic epitopes shared by other bacteria. Thus, persons with neither classical Lyme disease nor positive ELISA results have sera that can react with many lines on Western blots, so the specificity of this method and its role in diagnosis is undefined [27, 28]. Therefore, it is the current ELISA technology that is most relevant to our policy statement.

Regional Incidences of Lyme Disease and the Fatigue and Myalgia Syndrome

A major consideration in attempting to calculate the likelihood of risks and benefits of the parenteral antibiotic intervention is the prior likelihood or incidence of Lyme disease in a given area. Town-specific incidences in the state in which the disease was first defined were 815 per 100 000 (0.8%) in the most endemic area, with a county-specific incidence of 114 per 100 000 (0.1%) and a statewide incidence of 23 per 100 000 (0.02%) [29]. A national surveillance study by Tsai and colleagues [30] from 1987 to 1988 indicated the incidences shown in Table 1.

It should be noted that the incidence, even within states known to be endemic, can vary greatly. Epidemiologic studies have suggested that the rate of subclinical infection in endemic areas may vary from none to about the same rate as clinically apparent infections [11, 31, 32]. To bias this analysis in favor of antibiotic therapy, we have assumed the latter, that is, for every person contracting Lyme disease in Table 1, one infected asymptomatic seroconverter was at risk for having fatigue and myalgia as the sole symptom of Lyme disease. No data exist on the prevalence of Lyme disease in these areas; however, a period prevalence can be calculated by multiplying the incidence by the duration of the disease. We have assumed that a patient with the fatigue and myalgia syndrome will seek medical attention well within the first 2 years of symptom onset. Thus, in the mid-Atlantic region (where our assumption assumes an incidence rate of 6 nontypical/asymptomatic seroconverters per 100 000 population per year), the prevalence of seroconverters at the end of 2 years will be 12/100 000, which we have used as the baseline assumption for the prevalence of true-positive seroconverters in this analysis. In one report from a small island community in New York State, 9.7% of 176 persons had positive Lyme serologic results. We have examined the implications of this prevalence in our sensitivity analysis.

As mentioned earlier, the prevalence of the fatigue and myalgia syndrome, customarily called fibromyalgia, is between 2% and 5% [33-35]. A recent population-based random survey estimated a prevalence of 2% [36]; nothing suggests regional differences in prevalence. Although fibromyalgia is thought to be more common in women, data on the sex prevalence of positive Lyme disease serologic results is not known, and we, therefore, have not attempted to break our analysis down by sex. Calculations on the ratio of false-positive to true-positive patients with fatigue and myalgia are shown in Table 2.

If one assumes a best-case scenario (that is, in favor of antibiotics), that all true seroconverters who lack classical Lyme disease symptoms do have fatigue and myalgia, then in a population of 100 000 in the mid-Atlantic states, 2000 patients have fibromyalgia, 40 of whom (2%) will have false-positive Lyme disease titers. In the same population, 12 patients will have fatigue and myalgia who test as true positive. In the mountain states, the prevalence of fatigue and myalgia is identical to that in the mid-Atlantic states, whereas the prevalence of atypical Lyme disease seroconverters is 1/300th, that is, 1000 symptomatic false positives for every true positive.

Likelihood and Cost of Side Effects of Antibiotic Therapy

The 1991 edition of the "American Hospital Formulary Service Drug Information Report" lists the side effects of ceftriaxone therapy (Table 3). In addition to the toxicities in the table, ceftriaxone can infrequently cause cholecystitis requiring cholecystectomy. Although other side effects (eosinophilia, leukopenia, liver function abnormalities) can occur more frequently, these are not considered in our analysis because these phenomena would not cause substantial symptoms and would probably go unrecognized in an unmonitored patient. The above side effects are not necessarily additive in the treated population. Of the common side effects (>1%), diarrhea, upper gastrointestinal complaints, and rash occur in 1% to 4% of patients and can generally be considered mild. We have assigned an overall likelihood of 3% for the occurrence of one of these mild side effects. This gives somewhat greater weight to the problem of diarrhea because pseudomembranous colitis, although often mild, can occasionally be chronic and can cause substantial problems requiring therapeutic or diagnostic procedures. We have assigned a likely cost of managing these milder side effects at an average of $250 per occurrence, which includes the costs of topical or antipruritic therapy for rash; for antidiarrheal agents; for antinausea agents; for physician fees for the subset with side effects who require such; and for the occasional somewhat more extensive intervention that must be used in patients with severe pseudomembranous colitis.

The less frequent side effects, which are reported to occur in 0.1% to 1.0%, we have arbitrarily assigned a frequency of 0.1%, in order not to bias this analysis against the use of antibiotic therapy. Because pancytopenia, renal impairment, fever and chills, and the rare anaphylactic reaction can lead to considerably greater diagnostic evaluation and therapeutic intervention, we have assumed that these more serious side effects will cost an average of $2000 per occurrence (including physician visits, medications, and diagnostic testing).

In one placebo-controlled, blinded trial of empirical ceftriaxone for arthritis in 60 patients [37], 29 had substantial diarrhea; in 2 patients it was severe enough to require discontinuation of therapy. Nine patients had acute allergic reactions, and 6 of 28 developed pseudogallstones. Thus, our basic assumption of infrequent side effects may underestimate the costs of empiric antibiotic therapy. We think that both of the above dollar valuations are conservative, and they would, therefore, be more likely to bias this analysis in favor of empirical antibiotic therapy.

Costs of Administering Antibiotic Therapy

We have ascertained the cost of 3 weeks of intravenous antibiotic treatment in one nonendemic area (Kentucky) and in two endemic areas (central Wisconsin and New Jersey). Depending on whether antibiotic is given once or twice a day and on the professional fee component of charges, the costs range from $2300 to $5000. We have, therefore, assumed the average cost of antibiotic therapy to be $3500.

We have assumed that an average patient, eager to divest him or herself of possible Lyme disease who is receiving intravenous antibiotics, would be willing to pay $100 to avoid the pain and time involved in receiving 3 weeks of intravenous therapy, to avoid travel to the treatment site, and to avoid other inconvenient procedures, although we have not directly heard this from our patients or their surrogates. We think this is a very conservative estimate.

We have not solicited a valuation from patients or surrogates concerning the psychosocial costs of not treating a positive Lyme titer in a nonspecifically symptomatic patient (with fatigue and myalgia). Differences in the value placed by patients for this anxiety (chagrin factor) will likely vary considerably depending on residence within or outside an endemic area, because in most endemic areas the public is considerably more aware and understandably more concerned about the risks for Lyme disease. To not bias this analysis against the use of antibiotic therapy and in the absence of survey data from patient groups, we assumed that a patient with fatigue and myalgia who is aware of a positive Lyme disease serologic result would be willing to pay $500 to be treated. The effect of assigning a different dollar value to this component was then examined in the sensitivity analysis. Physicians may also be anxious about not treating a patient with a positive serologic test that may augur later development of classical, late Lyme disease. Because the same physicians quite often benefit financially from the administration of therapy, we have avoided factoring into this analysis the degree to which the concern and the remuneration may be offsetting variables for the physicians involved.

Risk for Subsequent Development of Lyme Disease in Untreated Seropositive Patients

Patients with positive serologic results who do not receive antibiotic therapy may be at some risk for subsequent development of the symptoms of late-stage Lyme disease. No long-term follow-up studies exist of patients found to have positive serologic results with no or atypical symptoms, so the precise risk for late Lyme disease is not known. Nor is it known whether outcomes in patients developing symptoms later are worse if treatment has been delayed. Because the fatigue and myalgia syndrome is a chronic state, and most patients will have substantial evaluation and follow-up before the possibility of occult Lyme disease is entertained and antibiotic therapy is considered, it is unlikely that such patients will ever acquire either the rash of Lyme disease or cardiac involvement because these occur very early. Because arthritis and Lyme neuroborreliosis can be late symptoms of Lyme disease, they are the most likely symptoms of Lyme disease for which seropositive patients are at risk. We have, therefore, assumed that the seropositive patients who actually have persistent subclinical Lyme disease have a 5% chance for developing neurologic symptoms and a 10% chance for developing arthritis from the Lyme spirochete. It must be understood that these estimates are subjective, because no evidence exists to assist in the prediction of later symptomatic Lyme disease in seropositive patients. The effect of varying the above assumptions was examined in the sensitivity analysis discussed below.

Effect of Treatment on Development of Lyme Disease

Considerable evidence exists that treatment with aggressive, intravenous antibiotic therapy in patients with more classical Lyme symptoms can be dramatic in curing or lessening those symptoms [1]. We, therefore, have assumed that a seropositive patient who receives intravenous antibiotic treatment with no overt objective symptoms of classical Lyme disease will successfully respond and be at no risk for subsequent development of late-stage Lyme disease. This is probably a very generous assumption and, again, it was made to bias our analysis in favor of antibiotic therapy.

Published series [16-19] indicate either no or transient responses of fatigue and myalgia to parenteral antibiotics. Although a transient effect might have a positive effect on any cost-effectiveness analysis, we have presumed that its transience would ultimately have a depressing and, therefore, negative effect as well. We have, therefore, treated such an effect as neutral here.

Results

Top Methods Results Discussion Conclusion Author & Article Info References

The baseline assumptions used in this analysis are shown in Table 4. The data derived from them are summarized in Table 5. They indicate that in a population of 100 000, 2% (2000 persons) will have the fatigue and myalgia syndrome due to causes other than Lyme disease. Of these, 2% (40 persons) will be falsely seropositive for Lyme disease. In the most endemic area, where the incidence of symptomatic Lyme disease is 6 per 100 000 persons, we have assumed an additional number are truly seropositive with persistent subclinical infection. In a 2-year span, their period prevalence would be 12/100 000. If 10% of these carriers are at risk for Lyme arthritis and 5% are at risk for Lyme neuroborreliosis, then 1.8 of the 12 carriers will develop these classical symptoms of late Lyme disease if left untreated.

If all seropositive patients (40 false-positive patients with fatigue and myalgia and 12 true-positive carriers) are treated with antibiotics for the possibility of Lyme disease, 3% (1.56 patients) will develop minor toxicity. Using a similar calculation for the 0.1% likelihood of major toxicity, 0.052 patients will develop a major drug reaction. If seropositive patients are treated only when those late joint or central nervous system symptoms occur, the number of patients per 100 000 developing minor and major toxicities will be 0.054 and 0.0018, respectively. Thus, in the most endemic region, the empirical strategy will cause 29 cases of drug toxicity for each one occurring using the alternative strategy.

The data on costs were derived by multiplying the dollar figures listed at the bottom of Table 4 by the numbers of people incurring those costs and were discussed previously in this report. Table 5 indicates that empiric treatment of all patients with the fatigue and myalgia syndrome and a positive Lyme serologic result would incur costs of $86 221 for each case of Lyme disease actually prevented or cured.

Sensitivity analyses have been done by varying the base assumptions, and the results are shown in Table 6. For example, in areas where the incidence of Lyme disease is 0.02 cases per 100 000, the use of the empiric strategy would result in a cost of more than $20 million per case of Lyme disease prevented or cured. If one presumes that the incidence of the idiopathic (that is, non-Lyme disease-related) fatigue and myalgia syndrome is only 1%, which would result in fewer false-positive patients being found, the cost per case prevented or cured decreases to $51 671.

The assumption that true-positive carriers of Lyme disease have a risk for late central nervous system or joint symptoms of 15% may be incorrect. When the risk for late central nervous system or joint disease is decreased to 5%, the cost per case prevented or cured increases to more than $265 000. Even if all the atypical true seroconverters ultimately get classical late Lyme disease, the cost of the empirical strategy is $9865 per patient.

When the prevalence of seropositivity was increased to 9.7% (as previously described), the cost of preventing a later case was $17 206. This occurred because only 15% of the treated patients were deemed at risk for late Lyme disease. Only when the risk for late Lyme disease was estimated to be 87% (at a 9.7% prevalence of seropositivity) did the empirical treatment strategy break even.

When the incidences of minor and major side effects were set at 25% and 10%, respectively, as suggested by Caperton and colleagues [37], the costs of preventing one case of late Lyme disease increased to only $93 276.

The dollar costs of toxicity may be different from those we assumed. As shown in Table 6, varying the costs of treating side effects over a large range had little effect on the costs of curing one patient. In general, differences in the incidence of side effects and costs of their treatment had little effect on the economics of the analysis.

Finally, the dollar value assigned to the anxiety experienced by a patient with the fatigue and myalgia syndrome and positive serologic results was varied. Although not a perfect technique, in utility theory this approach assumes that patients have sufficient anxiety and that they are willing to pay the cost of relieving it out of their own pockets. This sensitivity analysis shows that, using baseline assumptions, if patients were willing to pay $3485 for relief of their anxiety, or approximately what the course of therapy costs, the strategy described would break even.

Discussion

Top Methods Results Discussion Conclusion Author & Article Info References

Review and analysis of current data indicate that little rationale exists for the empirical administration of intravenous antibiotics to a patient with a positive Lyme antibody titer whose clinical symptoms are nonspecific (for example, fatigue, myalgia). Although such symptoms may be reported in early Lyme disease, recent publications [12, 13] indicate that oral antibiotic therapy is sufficient to eliminate the Lyme disease spirochete during the acute phase and that more aggressive (intravenous) antibiotic therapy will not eliminate later constitutional symptoms [16, 18]. Whereas evidence exists that a 2- to 4-week course of such therapy can improve unequivocal, objective manifestations of disseminated or chronic Lyme disease [2, 14, 15], no data exist to justify the empirical use of intravenous antibiotics as treatment for a patient with nonspecific complaints and a positive Lyme antibody titer. The cost-effectiveness analysis, described in our report, suggests that empirical treatment with intravenous antibiotics of patients from endemic areas, whose only symptoms are nonspecific fatigue and myalgia, on the basis of a positive Lyme serologic result alone, incurs costs and risks that far exceed any benefit that might be attained. When patients from nonendemic areas are considered, these risks are even greater. Our analysis estimates that it will cost $86 221 to prevent one case of late Lyme disease if all seropositive patients with fatigue and myalgia in the most endemic region are treated with intravenous antibiotics. The cost-effectiveness ratio would increase up to 300 times in a nonendemic area, such as the mountain states.

A major variable, substantially influencing the current analysis, was the value that patients (with nonspecific symptoms and a positive Lyme antibody test) assigned to their anxiety about not getting treatment and, therefore, possibly developing classical manifestations of late Lyme disease. In utility theory, one method of valuing this is to express the state of such patients in quality-adjusted life years. This method requires an estimation of the number of life years remaining to the patient. Because the age-specific incidence of Lyme disease infection is not known and may vary from region to region, we have elected to use an alternative method, that is, assigning a dollar amount that patients would hypothetically be willing to pay to eliminate that anxiety. In our cost-effectiveness analysis, when the value of anxiety increased as high as $3485, fortuitously the value assigned to the cost of antibiotic therapy in our baseline analysis, the empirical strategy, broke even.

In our analysis, we have attempted to bias our baseline assumptions to favor empirical antibiotic treatment. In many areas of the country, a course of intravenous antibiotic therapy for Lyme disease costs more than $5000. In addition, the incidence of side effects assumed in our analysis may be extremely conservative given the report of Caperton and his associates [37]. We have assumed a 15% likelihood of truly positive patients developing objective symptoms of late Lyme disease. Even if 100% of such patients later develop classical Lyme disease, the empirical strategy would cause side effects in four patients who are serologically false positive for every true-positive patient ultimately destined to develop late Lyme disease symptoms. Further, our analysis does not consider the cost-effectiveness aspects of repetitive cycles of intravenous antibiotics, which would obviously greatly increase costs and increase the risk for side effects without demonstrable effects on benefits.

Conclusion

Top Methods Results Discussion Conclusion Author & Article Info References

Our analysis indicates that no rationale exists for the administration of intravenous antibiotics to a patient with a positive Lyme antibody titer whose only symptoms are nonspecific fatigue or myalgia. Although such symptoms may be seen in early Lyme disease, current experience indicates that oral antibiotic therapy is sufficient to eliminate the Lyme disease spirochete at this stage and that more aggressive (intravenous) antibiotic therapy will not eliminate the constitutional symptoms of fatigue and myalgia. Our cost-effectiveness analysis suggests that empirical treatment with intravenous antibiotics of patients from endemic areas, who only have nonspecific constitutional symptoms (that is, fatigue and myalgia), incurs costs and risks that far exceed any benefits that might be obtained. When patients from nonendemic areas are considered, these risks are naturally even greater. Using the baseline assumptions in our analysis, only when the patient's anxiety (chagrin) about not treating a positive Lyme titer exceeds $3500 do the benefits of that strategy exceed the risks.