JAC Advance Access published online on March 29, 2007
Journal of Antimicrobial Chemotherapy, doi:10.1093/jac/dkm033
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Cost-effectiveness of empirical prescribing of antimicrobials in community-acquired pneumonia in three countries in the presence of resistance
1 i3 Innovus, Beaufort House, Cricket Field Road, Uxbridge UB8 1QG, UK 2 Summa Health System, 525 E. Market Street, Akron, OH, USA 3 Hospital Mutua de Terrassa, Pl. Dr. Robert 5, 08221 Terrassa, Barcelona, Spain 4 Bayer Healthcare, Friedrich-Ebert-Strasse 217, 42117 Wuppertal, Germany
* Corresponding author. Tel: + 44-1895-455380; Fax: +44-1895-520039; E-mail: mmartin{at}innovus.com
Received 10 July 2006; returned 15 August 2006; revised 17 January 2007; accepted 20 January 2007
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionTransparency declarationsReferences |
|---|
Objectives: To assess the cost-effectiveness of empirical outpatient treatment options for community-acquired pneumonia (CAP) in France, the USA and Germany, representing high, moderate and low antimicrobial resistance prevalence, respectively.
Methods: A decision analytic model was developed for mild-to-moderate CAP outpatient treatment. Treatment algorithms incorporated follow-up after treatment failure due to resistance or other reasons. First-line treatment included moxifloxacin, ß-lactams, macrolides or doxycycline; second-line treatment used a different antimicrobial class. Country-specific resistance and co-resistance prevalences to first- and second-line therapy for the major CAP pathogens were derived from surveillance studies. Clinical failure rates due to antimicrobial-susceptible and -resistant pathogens were obtained from the literature or estimated. Total costs were estimated using standard sources and a third-party payer perspective. Outcome measures included first-line clinical failures avoided, second-line treatments avoided and hospitalizations avoided. Incremental cost-effectiveness ratios (ICERs) were calculated.
Results: First-line moxifloxacin treatment followed by co-amoxiclav dominated all other treatments in France, the USA and in Germany for all outcome measures. Sensitivity analyses maintained moxifloxacin dominance in France and the USA but affected ICERs in some cases in Germany.
Conclusions: Antimicrobial resistance/spectrum have a significant impact on outcomes and costs in empirical outpatient CAP treatment. Despite low acquisition costs for generic antibiotics, first-line treatment effective against the major CAP pathogens, including strains resistant to other antimicrobials, resulted in better clinical outcomes in all countries and lower treatment costs for all.
Key Words: antimicrobial resistance , moxifloxacin , decision analytic model , cost-effectiveness analysis
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionTransparency declarationsReferences |
|---|
Community-acquired pneumonia (CAP) is a common and serious infection. Streptococcus pneumoniae is the most important bacterial pathogen in CAP, though a variety of other organisms such as Haemophilus influenzae and the atypical pathogens Mycoplasma pneumoniae and Legionella pneumophila are also implicated.1
The increasing prevalence of antimicrobial resistance in CAP pathogens, particularly S. pneumoniae, is of concern. There is evidence that CAP patients with penicillin-resistant pneumococci may be at greater risk of poor outcome.2,3 In addition, clinical failures of macrolide therapy due to macrolide-resistant S. pneumoniae have been well documented,4 and although the prevalence of fluoroquinolone-resistant S. pneumoniae is generally low, around 90 cases of clinical failure with levofloxacin have been observed.5,6 There are no reports of clinical failure due to failure to eradicate the infecting organism.
The majority of patients with mild-to-moderate CAP are treated in the community setting with empirical antimicrobial therapy. Patients with more serious disease or who are elderly or have co-morbidities may be hospitalized, though antimicrobial therapy is usually started empirically.7 Thus, it is important that the choice of antimicrobial therapy ensures appropriate coverage of potentially drug-resistant strains based on local antimicrobial resistance patterns.8
The prevalence of antimicrobial resistance varies substantially. France has the highest prevalence of penicillin-resistant or macrolide-non-susceptible S. pneumoniae in Europe, around 27% and 58%, respectively.9 In comparison, France's neighbour Germany is typical of more northern European countries with a relatively low prevalence of drug-resistant S. pneumoniae. The USA has a significant prevalence of drug-resistant S. pneumoniae, which appears to have stabilized at an intermediate level versus the range observed across Europe.10
CAP is a costly disease. In the USA, the annual cost of treating the 5 million cases of CAP in patients under 65 years of age is US$12.2 billion, and this cost increases as more elderly patients are included.11 Patients hospitalized with CAP present the major health and economic burden; hospitalization costs represent 
90% of total costs in the USA, approximately US$4 million/100 000 population.12 In France, of the 400 000 annual cases of CAP, around 80 000–100 000 result in hospitalization.13 In Germany, the annual cost of CAP is approximately 
500 million, with nearly 240 000 hospitalizations.14 In Europe, average costs per hospital stay for CAP tend to be lower than in the USA, one study estimated (from a main payer perspective) a cost of 3892 for Germany and 5928 for France versus 14347 for the USA (approximately US$17447).15 This study also found that appropriate antibiotic inpatient treatment contributed to lower costs.15 In addition, a study of antimicrobial treatment of lower respiratory tract infection conducted in a healthcare management organization found that when there was a match between pathogen antimicrobial susceptibility and initial antimicrobial therapy, the average cost of treatment per episode was $8821 versus $14754 (P = 0.02) when therapy and pathogen susceptibility were not matched.16
Despite concerns regarding the clinical impact of antimicrobial resistance on outcomes in CAP and the high costs of hospitalization, few data are available on the cost implications of antimicrobial resistance in this infection. From a third-party payer perspective, antimicrobial resistance has the potential to impact drug costs, personnel time, supplies, space and equipment; the major cost being due to patient hospitalization.17
The objective of this analysis was to develop a decision analytic model to evaluate the impact of antimicrobial resistance on the cost-effectiveness of first-line antimicrobial treatment strategies in CAP for patients treated empirically in the community. Three countries with different antimicrobial resistance prevalence profiles, France, USA and Germany, were selected for analysis using a third-party payer perspective.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionTransparency declarationsReferences |
|---|
Study population
The decision analytic model considered subjects with CAP classified as Fine risk categories I–III, i.e. with mild-to-moderate CAP. These patients would usually be treated empirically in the community. The Pneumonia Outcome Research Team (PORT) study reported the relative incidence of Fine risk criteria in an outpatient population as 62%, 26% and 8% for categories I, II and III, respectively.18 For the purpose of the model, these numbers were recalibrated to add up to 100%, resulting in a split between categories I, II and III of 65%, 27% and 8%, respectively, for the study population.
CAP antimicrobial treatment algorithms
Antimicrobial treatment algorithms were developed for France, the USA and Germany on the basis of country-specific CAP treatment guidelines, expert opinion and the most frequently prescribed therapies for CAP based on market sales data for 2003 (Table 1).7,19,20 Treatment algorithms for all three countries included a fluoroquinolone (moxifloxacin), a macrolide or a ß-lactam as first-line options, and in the USA, doxycycline first-line was also considered. Second-line treatment had to be with a different antimicrobial class. Combination therapies were not included, as monotherapy is usually prescribed in the community setting; in addition, robust efficacy data for combination treatments, except for co-amoxiclav, are not routinely available. Treatment duration was 10 days for both first- and second-line treatments.
|
Decision analytic model and outcomes analysis
Figure 1 describes the decision analytic model, developed in Excel, used for the treatment of CAP in the community. Only the most common bacterial pathogens implicated in CAP (S. pneumoniae, H. influenzae and the atypical pathogens) were considered in the model at their approximate relative isolation frequencies. These were estimated on the basis of observed prevalences across German CAP patients treated initially as outpatients,21 reporting 40% for S. pneumoniae, 8% for H. influenzae and 12% for atypicals, confirming existing prevalence estimates from the published literature1 and expert opinion. For the purpose of the model, these prevalences were normalized to 100%, giving a frequency of 67% for S. pneumoniae, 10% for H. influenzae and 23% for the atypical pathogens. Viruses were not included in the base case, as ideally antimicrobials should not be prescribed for CAP infections of viral origin.
|
Antimicrobial treatment for these pathogens was based on the algorithms outlined above (Table 1). The model was designed to estimate the following outcomes: clinical failures avoided, second-line treatments avoided and hospitalizations avoided. In brief, first-line antimicrobial therapy can result in either clinical success or failure. Clinical failure is followed by either second-line antimicrobial therapy in the community or hospitalization. The rate of hospitalization after first-line therapy was estimated on the basis of the Fine criteria mix of the study population,18 with 6.9% of failures being hospitalized in the USA and 15.4%22 in France and Germany (with 9.2%23 in Germany tested in the sensitivity analysis), the remainder progressing to second-line therapy. The mortality rate after first-line therapy failure was considered to be zero. It was assumed that failure of second-line therapy in the community always resulted in hospitalization, with a mortality rate of 0.9%, based on the Fine criteria mix of the population.18 ICU admittance was not included in the base case analysis.
The model time scale for analysis was equivalent to the treatment duration of a CAP episode, including a possible second treatment and further hospitalization. As the total time, even considering the most extreme scenario, is shorter than 1 year, no discounting of costs or outcomes was required.
Estimate of clinical failure rate
Central to the model is the rationale that clinical failure can occur due to two main reasons: lack of response to treatment in patients with susceptible pathogens and failure caused by the presence of antimicrobial-resistant pathogens, though non-adherence, intolerance, empyema or wrong diagnosis can also be reasons for clinical failure. A series of equations describing the probability of first-line treatment failure for these two causes were derived from the model and are summarized in Table 2.
|
The failure rate due to lack of response against susceptible pathogens was estimated on the basis of antimicrobial efficacy data from clinical trials in CAP from the published literature. Studies were matched as far as possible on the basis of patient populations, causative pathogens, time horizon, dosage and outcome measures (Table 3).24–35
|
For each pathogen, expectations of clinical failure due to resistance were estimated. For S. pneumoniae, a 50% clinical failure rate was estimated in the base case analysis for the treatment of strains with MICs above the CLSI resistant breakpoint with fluoroquinolones and non-susceptible breakpoint for ß-lactams and doxycycline.
For macrolides, there are two main resistance mechanisms in S. pneumoniae; the mef(A) genotype was believed to confer mainly low-level resistance (MICs 1–32 mg/L), whereas the erm(B) genotype was expected to result in mainly high-level resistance (MICs 
8 mg/L).36 Until recently, it was believed that macrolides would be expected to retain efficacy against a proportion of macrolide-resistant strains with mef(A) and low MICs (1–4 mg/L), but be ineffective against erm(B) strains.37 However, recent evidence has indicated that both mechanisms can result in a high level of resistance, and therefore, an estimated clinical failure rate of 50% was used for both mef(A) resistance and strains with erm(B) resistance [±mef(A)]. The variance of mef(A) versus erm(B) prevalence in macrolide-resistant S. pneumoniae between the three study countries, 2.8% and 97.2% in France,38 63.8% and 35.3% in the USA38 and 51.0% and 40.8% in Germany,38 was included in the model, though in the base case this currently does not result in any differences because of the same rates being applied to mef(A) and erm(B).
Clinical failure rates for antimicrobial-resistant strains of H. influenzae were estimated as 50% for all antimicrobial classes. For the atypical pathogens, treatment with ß-lactams was assumed to be ineffective.7 However, a relatively high spontaneous cure rate (80–85%) has been observed for M. pneumoniae and Chlamydia pneumoniae in clinical practice isolated in mild-to-moderate community-acquired infection (assumed prevalence of 67% and 8%, respectively21), and 20% of patients with Legionella spp. and mild-to-moderate CAP will recover without therapy. Thus, based on the relative incidence of isolation of atypical pathogens, a spontaneous cure rate of 65% was assumed, giving an estimated clinical failure rate of 35% for ß-lactams against these organisms in this patient population.
Antimicrobial resistance prevalence
Antimicrobial resistance prevalence for the three study countries was based on published surveillance data (Table 4).9,10,39–42 For S. pneumoniae, CLSI resistance breakpoints were used for fluoroquinolones and non-susceptible breakpoints for ß-lactams and macrolides, except azithromycin for which the resistance breakpoint was used on the basis of the most appropriate data available. Doxycycline resistance was assumed to be the same as tetracycline resistance, based on the CLSI resistance breakpoint for S. pneumoniae. For roxithromycin, data for clarithromycin were used, based on the CLSI non-susceptible breakpoint for S. pneumoniae. Note that where ‘resistance’ is referred to throughout the article, it may relate to non-susceptible or resistance prevalences as detailed in Table 4.
|
For H. influenzae, breakpoints derived from pharmacokinetic/pharmacodynamic (PK/PD) criteria were used, except for moxifloxacin, for which the CLSI non-susceptible breakpoint was used.10 For roxithromycin, data for clarithromycin based on the PK/PD breakpoint were used. The differences between PK/PD- and CLSI-resistant breakpoints for H. influenzae are usually minor for fluoroquinolones, are more important for ß-lactams and have a significant impact on macrolide resistance prevalence. However, breakpoints based on PK/PD criteria are thought to better reflect the potential for clinical success/failure in H. influenzae and have been validated in animal models and clinical studies in otitis media and sinusitis and so were used in the base case analysis.10,43,44 For atypical pathogens, there are no reports of acquired resistance.
Estimating failure rate for second-line therapy: effect of multidrug resistance
Patients who failed first-line therapy were assigned to second-line therapy on the basis of the country-specific treatment algorithm (Table 1).
The effect of co-resistance between first- and second-line therapies was included in the model. An estimate was made of the probability that an S. pneumoniae isolate that was resistant to first-line therapy could also be resistant to second-line therapy, based on data for the prevalence of multidrug-resistant strains. This consideration was only incorporated for S. pneumoniae. As data on co-resistance between different antimicrobials are limited, 2003 data for the USA were used as a basis to calculate resistance prevalences to second-line therapy adjusted for the presence of co-resistance.41,45,46 For example, for a macrolide followed by co-amoxiclav, the prevalence of azithromycin-resistance in the US dataset used was 29.0%, with 97.7% of azithromycin-resistant strains also being co-amoxiclav-resistant. A multiplier of co-resistance could therefore be calculated as 97.7/29.0 = 3.4. This multiplier was combined with data on the prevalence of multidrug resistance (defined as resistance to at least three antimicrobial classes excluding penicillin) from the reported surveillance data for the different countries:10 35.6% for France, 16.2% for the USA and 4.7% for Germany into the following equation:
|
|
The calculated resistance prevalences for second-line therapy adjusted for the presence of multidrug resistance used in the model are shown in Table 5.
|
Resource use and cost estimates
Resource use was estimated from the literature and costs were obtained from established national cost sources using a third-party payer perspective for the year 2006 (Table 6).18,45,47–62 Where necessary, costs were inflated to 2006 levels.
|
Drug therapy costs were for 10 days of treatment, regardless of treatment outcome (Table 7); 10 days were chosen, as this is still considered the standard in clinical practice, though there is a trend towards shorter treatment. For France, a mean cost based on the cost of branded and generic treatments was assumed, as patients will often request branded drugs in this country.47 In Germany and in the USA, except in the case of moxifloxacin, drug treatment costs were for generic treatments, where available.48,49 Antimicrobial costs excluded patient co-payments (35% of cost in France and 10% in Germany with a minimum of
5.00 and maximum of 10.00). In the USA, average wholesale prices were used for consistency, without any adjustment for manufacturer's discounts, though patient co-payments were deducted where required, reflecting the third-party payer perspective.
|
Patients incurred medical visits at day 1 and day 3 of their first CAP episode and could incur further visits depending on their treatment pathway. Bacterial culture and susceptibility testing is not generally performed in the community setting and this cost was not, therefore, included. On the basis of the literature, the cost of two chest X-rays were included for 39% of patients in France,58 60% in the USA45 and 59% in Germany.60 A white blood cell count was assumed for all treatment failures. Hospitalization and mortality rates were derived from the decision model (Figure 1) as described above.
To estimate the total cost of a general practitioner (GP) visit in each country, home visits were included for France and Germany. On the basis of the literature, the frequency of home visits was estimated as 28% for France50 and 35% for Germany.56 Emergency room visits were also included for France and Germany at a frequency of 5%, and this was deducted from the home visits, giving a home-visit frequency of 23% for France and 30% for Germany to be used in the model. In the USA, home visits are not made and patients may visit an emergency room or outpatient clinic instead, with an estimated frequency of 28% and 5%, respectively.53
Hospitalization costs considered drug-related costs and medical costs (Table 7).52,54,61 On the basis of a publication by Colice et al.,11 which presents evidence on costs differences between patients surviving and dying in hospital, a multiplier of 1.45 was applied to the cost of patients dying in hospital because of CAP.
Incremental cost-effectiveness ratios (ICERs) were calculated for outcomes and costs using the following equation: ICER = (C1 – C0)/ (E1 – E0), where C1 is the cost of the intervention under study and C0 is the cost of the alternative with which it is compared, and E1 and E0 are their respective health outcomes.63 A strategy is said to be dominant when it is both less costly and more effective. Moxifloxacin/co-amoxiclav was used as the basis for comparison (C1, E1), as this was common to the treatment algorithms of all three countries studied. A variety of deterministic and probablistic sensitivity analyses were performed to test the model's main parameters.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionTransparency declarationsReferences |
|---|
Cost-effectiveness analyses—base case analysis
The results of the base case analysis are shown in Table 8. Moxifloxacin/co-amoxiclav dominated the other treatment options in all study countries (i.e. was less costly and more effective than all other comparators).
|
First-line treatment with a macrolide was less costly than co-amoxiclav in France and the USA and than cefuroxime axetil in Germany. However, first-line macrolide therapy resulted in worse outcomes when compared with co-amoxiclav in France and the USA and also required the higher rates of second-line therapy (Table 8).
Looking in more detail at the need for hospitalization, first-line therapy with moxifloxacin resulted in a reduction in the need for hospitalization of 68–72% for France, 48–60% for Germany and 46–78% for the USA compared with other treatment options (Table 8).
Sensitivity analysis: impact of antimicrobial resistance
The calculated rate of first-line treatment failure due to either susceptible or antimicrobial-resistant pathogens in the base case analysis is shown in Figure 2. For both France and the USA, macrolides as first-line therapy performed the poorest, whereas cefuroxime axetil had the highest potential for clinical failure in Germany. The impact of the higher resistance prevalence for France versus the USA versus Germany can be seen for the macrolides and ß-lactams. For macrolides, failures due to resistance represented 83.3% of total failures in France, 76.5% in the USA and 71.4% in Germany. Failure due to susceptible pathogens was found to be lower for macrolides than for moxifloxacin in all countries, since higher resistance rates result in the dominance of non-susceptible pathogen treatment pathways for macrolides. For ß-lactams, the proportion of failures due to resistance was 57.6% for France and 56.5% for the USA with co-amoxiclav and 52.8% for amoxicillin and 55.9% for cefuroxime axetil for Germany. For the USA, 52.2% of all failures with doxycycline were due to resistance. There was little effect on first-line failures due to resistance pathogens with moxifloxacin between the countries, with failures due to resistance contributing 3.3%, 1.6% and 2.0% of total failures for France, the USA and Germany, respectively.
|
The effect of resistance on hospitalizations was also examined (Figure 3). For the moxifloxacin first-line strategy, resistance contributed 2.0%, 0.9% and 1.4% to hospitalizations for France, the USA and Germany, respectively. For a macrolide first-line strategy for France, 65.5% of hospitalizations were due to resistance, compared with 47.6% for the USA and 57.1% for Germany. For a ß-lactam first-line strategy, resistance contributed to 36.4% of hospitalizations for France and 26.5% for the USA. In Germany, resistance was responsible for 36.1% of hospitalizations for amoxicillin first-line and 42.4% for cefuroxime axetil first-line. First-line use of doxycycline in the USA led to 61.5% of hospitalizations being attributable to drug resistance.
|
To determine the effect of resistance on cost-effectiveness, an analysis was performed, which assumed that all antimicrobials were 100% effective against susceptible pathogens. In this analysis, moxifloxacin/co-amoxiclav remained dominant over all other therapy options in all three study countries.
The effect of multidrug resistance against second-line therapy was also examined. If multidrug resistance was assumed to be absent, hospitalizations were marginally reduced or unchanged for moxifloxacin/co-amoxiclav, macrolide/moxifloxacin and cefuroxime axetil/moxifloxacin treatment strategies. However, for a ß-lactam, followed by a macrolide, hospitalizations were reduced by 8.1% for France, 6.3% for the USA and 0.4% for Germany in the absence of multidrug resistance. These differences were reflected in cost reductions of 4.9%, 1.2% and 0.1% for France, the USA and Germany, respectively. In the USA, the doxycycline/azithromycin strategy resulted in a hospitalization reduction of 14.0%, leading to a 3.8% cost reduction. In terms of cost-effectiveness, the absence of multidrug resistance in the model had no effect on the dominance of moxifloxacin/co-amoxiclav over other treatment options for France, Germany and the USA.
Deterministic sensitivity analyses: other variables
As moxifloxacin/co-amoxiclav was dominant in the base case analysis, the effect of using moxifloxacin first-line or second-line was further examined. First, to equalize the effect of second-line therapy, an analysis was performed in which all second-line therapies were changed to co-amoxiclav. In this analysis, moxifloxacin/co-amoxiclav dominated all other treatment strategies for all three study countries. Second, to compare the effect of moxifloxacin used first- or second-line, moxifloxacin replaced all second-line therapies in the treatment algorithm; for this analysis, moxifloxacin/moxifloxacin dominated all other treatment options in all three countries studied.
The estimated clinical failure rates due to resistant pathogens assumed in the base case analysis were tested for their relative effect on model outcomes; a reduction of the base case clinical failure rate from 50% to 20% was examined. For France and the USA, reducing the clinical failure rate for individual antimicrobial classes separately had no effect on the dominance of the moxifloxacin/co-amoxiclav strategy. For Germany, reducing the macrolide clinical failure rate alone to 30%, 25% and 20% resulted in the loss of the dominance of the moxifloxacin/co-amoxiclav strategy, resulting in ICERs ranging between 38.02 and 15 123 per first-line clinical failure avoided, 44.94 and 178.76 per second-line treatment avoided and 289.06 and 1257.82 per hospitalization avoided.
In the base case analysis, a spontaneous cure rate for atypical infections treated with ß-lactams of 65% was used (i.e. a 35% failure rate). Changing the atypical spontaneous cure rate to 95% (i.e. a failure rate of 5%) had no impact on the dominance of the moxifloxacin/co-amoxiclav strategy in France and the USA versus all comparator treatments and in Germany versus roxithromycin/moxifloxacin and cefuroxime axetil/moxifloxacin. For the amoxicillin/roxithromycin strategy in Germany and a 95% spontaneous cure rate, the ICERs were 106.78 for first-line clinical failures, 126.19 for second-line therapy and 398.21 for hospitalizations avoided.
In the base case analysis, to estimate the resistance prevalence to H. influenzae, PK/PD breakpoints were used. These differ from CLSI breakpoints for macrolides, ß-lactams and doxycycline. A sensitivity analysis was, therefore, performed using H. influenzae CLSI breakpoints applied to the same antimicrobial resistance surveillance data set.10 For France and the USA, this analysis resulted in a reduction in first-line clinical failures and hospitalizations compared with the base case analysis, particularly for strategies using macrolides first-line. However, the overall dominance of the moxifloxacin/co-amoxiclav strategy was retained. In Germany, using CLSI breakpoints, moxifloxacin/co-amoxiclav dominance was retained against amoxicillin/roxithromycin and cefuroxime axetil/moxifloxacin. However, moxifloxacin/co-amoxiclav was no longer dominant against roxithromycin/moxifloxacin, with ICERs of 311.77 for clinical failure avoided and 368.53 per second-line treatment avoided and 2476.42 per hospitalization avoided.
The effect of changing hospitalization costs was also evaluated. For France or the USA, an increase or decrease in hospitalization costs of 50% had no effect on the dominance of moxifloxacin/co-amoxiclav. For Germany, increasing hospital costs by 50% also had no effect on the dominance of moxifloxacin/co-amoxiclav, whereas decreasing hospital costs by 50% resulted in ICERs for moxifloxacin/co-amoxiclav of 78.74, 93.07 and 537.71 versus roxithromycin/moxifloxacin for first-line clinical failures avoided, second-line treatment avoided and hospitalizations avoided, respectively. The effect of reducing the hospitalization rate for Germany (from 15.4% to 9.2%) produced similar results, with moxifloxacin/co-amoxiclav only losing dominance over the roxithromycin/moxifloxacin strategy with ICERS of 43.48, 47.88 and 518.33 for first-line clinical failures avoided, second-line treatment avoided and hospitalizations, respectively.
Viruses, when mistaken for bacteria in CAP, are often treated by antimicrobial drugs. Thus a change in the pathogen frequency split allowing the inclusion of viruses was also evaluated. Because viruses are self-limiting, it was assumed that viruses, treated by any of the antimicrobial drugs, are 100% resistant to treatment and yet experience a 100% clinical success rate, so patients are treated with a first-line treatment only. Applying a new frequency of 54% for S. pneumoniae, 8% for H. influenzae, 18% for atypical pathogens and 20% for viruses, the costs, clinical failure and hospitalization rates decreased in all countries since the microbial pathogens now represented fewer of the potential pathogens in the decision tree; however, the moxifloxacin/co-amoxiclav strategy remained dominant over all other treatment strategies in France, the USA and Germany.
Probabilistic sensitivity analysis
Probablistic sensitivity analyses (PSA) were performed using Monte Carlo simulations (10 000 runs) on distributions for those parameters that were found to have an impact on the deterministic sensitivity analyses. The spontaneous cure rate for atypical infections treated with ß-lactams and the clinical failure rates for each antimicrobial drug class were varied uniformly over the same ranges tested in the deterministic sensitivity analysis. The clinical success rates for each drug were varied in a triangular fashion using values found in the literature. Hospitalization costs for each country were tested by means of a lognormal distribution and resistance rates were varied using a beta distribution.
Since the means used for the base case of the sensitivity analysis differed slightly from the base case, due to the nature of the distributions chosen, the mean results of the PSA are not identical to the base case results. Nevertheless, the results proved to be robust to change. In each study country, the mean result after 10 000 runs maintained the dominance of the moxifloxacin/co-amoxiclav strategy.
Figures 4–6 show the scatter plots of the Monte Carlo simulations on the cost-effectiveness plan in the three countries for the outcome of first-line failure. In France and the USA we can see that no points cross the x-axis, meaning that the moxifloxacin/co-amoxiclav strategy is dominant in nearly all cases in France and the USA, respectively. In France, moxifloxacin first-line has a 100% and 99.8% probability, respectively, of being dominant against co-amoxiclav and clarithromycin first-line strategies. Similarly, in the USA, moxifloxacin is dominant 100%, 99.8% and 99.8% of the time against co-amoxiclav, azithromycin and doxycycline first-line treatments. In Germany, the cloud crosses the x-axis, indicating the loss of dominance in certain situations. The probability of dominance of moxifloxacin/co-amoxiclav is 67.1%, 54.4% and 100% against amoxicillin, roxithromycin and cefuroxime first-line strategies, respectively.
|
|
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionTransparency declarationsReferences |
|---|
The impact of bacterial resistance on the cost-effectiveness of antimicrobial therapy in CAP has not been systematically investigated. In the model described here, we aimed to explore the effect of differing bacterial resistance profiles for three countries on the cost-effectiveness of various treatment algorithms for the outpatient treatment of CAP.
The base case analysis showed that for France, Germany and the USA, first-line/second-line treatment with moxifloxacin/co-amoxiclav dominated all other evaluated strategies, including first-line therapy with a macrolide, ß-lactam or doxycycline. Moxifloxacin/co-amoxiclav was, therefore, both less costly and more effective than the comparator treatments. The countries included in this analysis, France, Germany and the USA, cover a representative range of resistance prevalence. France represents a country with a high prevalence of antimicrobial resistance, though higher prevalences are seen in the Far East.10 Germany is representative of countries with a low resistance prevalence, though there has been a trend for increased macrolide resistance in recent years.64 The results for these countries could, therefore, be seen as indicative for countries with similar resistance prevalence. This means that treatment strategies with moxifloxacin are expected to be more cost-effective in countries where resistance levels are high. Local treatment costs will have an effect, with generic treatments increasing the ICERs, but generic treatments were included in this model for all three countries (where available) and the higher treatment costs for moxifloxacin were offset in part or totally, depending on the sensitivity analyses, by the cost savings due to higher clinical efficacy.
The results of the model demonstrate that from a clinical point of view, it is more advantageous to use the antimicrobial that will be effective against the greatest proportion of CAP pathogens, including strains resistant to other antimicrobials, i.e. the moxifloxacin/co-amoxiclav strategy, and that, in most cases, this is also the least costly option. This was found even when the prevalence of resistance was relatively low, as in the case of Germany.
The effect of multidrug resistance in S. pneumoniae, specifically co-resistance between first-line and second-line therapies, was also tested in the model. A higher prevalence of multidrug resistance was associated with an increase in hospitalizations. The effect was most pronounced in France (8.0% increase in hospitalizations) where multidrug resistance is 36%, but was only slight in Germany (0.4% increase in hospitalizations) where 5% of pneumococcal strains are multidrug-resistant.
Deterministic sensitivity analyses showed that the model is relatively insensitive to change and the observed changes occurred in the expected direction of change. In the majority of the sensitivity analyses, moxifloxacin/co-amoxiclav remained the dominant treatment strategy. These findings indicate that reserving newer, efficacious therapies, for which bacterial resistance prevalence is relatively low for use as second-line treatment only, may be clinically and economically less favourable than using these agents as first-line therapy. In addition, the dominance of the moxifloxacin/co-amoxiclav strategy further illustrates this under the real-world assumptions that antimicrobials are prescribed for CAP infection of viral origin.
Although there is a lack of both clinical and economic studies in this area, there is some evidence supporting our findings. For example, a probability model that estimated the impact of antimicrobial resistance on clinical outcomes in CAP, though having key differences compared with our model and lacking a cost analysis, did have hospitalizations as an outcome and an analysis based on antimicrobial susceptibility data for France, thus allows limited comparison.65 In this study, for France, 12.7% of all patients diagnosed with CAP (Fine categories I–V) were hospitalized after amoxicillin/erythromycin and 13.7% after erythromycin/levofloxacin therapy and >40% of hospitalizations were due to antimicrobial resistance.65 Results for France in the current model are consistent with though perhaps more conservative than these findings, with 5.5% of patients hospitalized with the co-amoxiclav/clarithromycin treatment strategy and 4.9% with clarithromycin/moxifloxacin, although only Fine categories I–III patients were considered; 37% and 66%, respectively, of these hospitalizations were due to antimicrobial resistance. However, the previous study does support the overall conclusions reported in this article, finding that antimicrobial resistance may be a significant contributor to subsequent hospitalization in adults initially treated as outpatients for CAP.65
The split over Fine categories I–III was used from the original PORT study. Since that time, the USA has used this severity classification system in clinical practice. It is thought that this split has changed, as more Category III patients are now treated as outpatients. However, we have used the original split as this represents a conservative approach, as more cases are expected to be mild and the effect of treatment would be less pronounced.
It is important to recognize that this study has several limitations, most of them caused by a lack of suitable data. First, there is a lack of reported data on the relationship between resistance and clinical failure. In particular, the majority of clinical trials exclude patients who have pathogens that are non-susceptible to study medications. Thus, in the model, we were required to make a number of assumptions based mainly on expert opinion. There is evidence for a link between in vitro macrolide resistance in S. pneumoniae and clinical failure with macrolides from a well-designed case–control study.66 However, in this study, clinical failure is defined as breakthrough bacteraemia, rather than the need for further antimicrobial therapy or hospitalization; so comparison with the model objectives is not possible. A second limitation is that the model requires the use of population-specific efficacy data. Meta-analyses found in the literature were not suitable, as these included data from all settings (hospital and community) and all age groups (adults, children and the elderly). The lack of comparative trials in the specific population studied and differences in time horizons, outcome measures and treatment regimens in clinical trials posed problems, resulting in a restricted number of suitable publications for use in the model.
This model does not address any adverse effects related to treatment. This aspect was omitted for several reasons: first of all because most adverse events are not very serious; second because it is difficult to measure the economic effect of these adverse events; third, because this would have increased the complexity of the tree while removing the focus from the main issue, antimicrobial resistance; and finally because the nature of the serious adverse events is often not revealed in publications, making costing extremely difficult.
Published surveillance data were used to obtain resistance prevalence for the countries under consideration. These data are at best 1 year old and therefore may not reflect currently prevailing resistance patterns. Data from surveillance studies will depend on the centres included in the study and may not be representative of more local trends. In addition, the majority of samples considered in surveillance programmes are obtained from hospitalized patients, though there is some evidence that for community-acquired pathogens, hospital resistance prevalence data are consistent with findings from the community.67 Despite these limitations, surveillance data still represent the best available data source for the measurement of resistance in different countries in terms of quality control, consistency and availability.
This model used resistance rates for the general population, as we could not assess the relative risk for resistance for patients in the model, i.e. the model does not take into account any co-morbidities which could influence resistance, and we assumed for the purpose of the model that patients had not had any prior antimicrobial use in the preceding 3 months. This would have led to an underestimation of the effect of treatment in all countries concerned. However, the size of this effect was explored in a sensitivity analysis in which resistance rates were decreased by 50%, which did not change the results, i.e. the moxifloxacin/co-amoxiclav strategy remained less costly and more effective than other treatment strategies.
This model includes only direct medical costs. A large percentage of the CAP population included in this analysis, Fine risk categories I–III, will be employed. A reduction in hospitalizations and second-line treatments is therefore expected to result in a decrease of lost productivity, something which is generally assessed in an analysis using a societal perspective, i.e. including productivity costs. This type of analysis would be expected to result in a further decrease in the ICERs without affecting the current hierarchy of results.
Despite its limitations, this model is the first to provide both clinical and economic information on the effect of resistance on empirical prescribing and may assist local decision-makers in allocating scarce healthcare resources. As the relationship between resistance and clinical failure in the model was based primarily on assumptions and expert opinion, further research providing information on this relationship would benefit further modelling work in this area.
To our knowledge, this model is the first health economic model to assess the cost-effectiveness of the antimicrobial therapy of CAP in the presence of antimicrobial resistance and the first to consider the effect of multidrug resistance on the efficacy of second-line therapy. Our model indicates that resistance has considerable economic and clinical consequences, with reduced efficacy of macrolides and ß-lactams versus moxifloxacin, even at the low-resistance prevalence included in the model for Germany. In the base case and in most instances in the sensitivity analyses, moxifloxacin/co-amoxiclav maintained dominance over other therapy options. In addition, moxifloxacin used first-line was dominant over other treatment strategies using a macrolide, ß-lactam or doxycycline first-line and using moxifloxacin as second-line therapy. In conclusion, these results indicate that choosing first-line antimicrobial therapy in CAP, which will be effective against the majority of pathogens, including antimicrobial-resistant strains, is the most clinically effective and generally also the least costly treatment option.
|
Transparency declarations |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionTransparency declarationsReferences |
|---|
The study was carried out independently by i3 Innovus and the manuscript was approved by each author before submission. Bayer Healthcare had the opportunity to comment on a draft of the manuscript and the authors were free to decline publication. M. M. and S. Q. are full-time employees of i3 Innovus. M. K. and A. K. are full-time employees of Bayer Healthcare. T. F. recent research funding: Ortho-McNeil, Oscient, Pfizer, sanofi-Aventis; Consultant: Bayer, GlaxoSmithKline, Ortho-McNeil, Oscient, Pfizer, sanofi-Aventis, Schering-Plough; Speakers' Bureau: Abbott, GlaxoSmithKline, Merck, Ortho McNeil, Oscient, Pfizer, sanofi-Aventis, Schering Plough, Wyeth. J. G. reports no conflicts of interest.
|
Acknowledgements |
|---|
We thank Naomi Richardson of Magenta Communications Ltd for editorial assistance in the preparation of this manuscript. This study was carried out by i3 Innovus under a grant from Bayer Healthcare.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionTransparency declarationsReferences |
|---|
1 . Bartlett JG, Dowell SF, Mandell LA, et al. (2000) Practice guidelines for the management of community-acquired pneumonia in adults. Infectious Diseases Society of America. Clin Infect Dis 31:347–82.[CrossRef][Medline]
2 . File TM Jr. (2004) Streptococcus pneumoniae and community-acquired pneumonia: a cause for concern. Am J Med 117:Suppl 3A, 39S–50S.
3 . Tleyjeh IM, Tlaygeh HM, Hejal R, et al. (2006) The impact of penicillin resistance on short-term mortality in hospitalized adults with pneumococcal pneumonia: a systematic review and meta-analysis. Clin Infect Dis 42:788–97.[CrossRef][Web of Science][Medline]
4 . Garau J. (2003) Why do we need to eradicate pathogens in respiratory tract infections? Int J Infect Dis 7:Suppl 1, S5–12.
5 . Perez-Trallero E, Marimon JM, Iglesias L, et al. (2003) Fluoroquinolone and macrolide treatment failure in pneumococcal pneumonia and selection of multidrug-resistant isolates. Emerg Infect Dis 9:1159–62.[Web of Science][Medline]
6 . Scheld WM. (2003) Maintaining fluoroquinolone class efficacy: review of influencing factors. Emerg Infect Dis 9:1–9.[Web of Science][Medline]
7
.
Niederman MS, Mandell LA, Anzueto A, et al. (2001) Guidelines for the management of adults with community-acquired pneumonia. Diagnosis, assessment of severity, antimicrobial therapy, and prevention. Am J Respir Crit Care Med 163:1730–54.
8
.
Ball P, Baquero F, Cars O, et al. (2002) Antibiotic therapy of community respiratory tract infections: strategies for optimal outcomes and minimized resistance emergence. J Antimicrob Chemother 49:31–40.
9 . Jones ME, Blosser-Middleton RS, Critchley IA, et al. (2003) In vitro susceptibility of Streptococcus pneumoniae, Haemophilus influenzae and Moraxella catarrhalis: a European multicenter study during 2000–2001. Clin Microbiol Infect 9:590–9.[CrossRef][Web of Science][Medline]
10
.
Jacobs MR, Felmingham D, Appelbaum PC, et al. (2003) The Alexander Project 1998–2000: susceptibility of pathogens isolated from community-acquired respiratory tract infection to commonly used antimicrobial agents. J Antimicrob Chemother 52:229–46.
11 . Colice GL, Morley MA, Asche C, et al. (2004) Treatment costs of community-acquired pneumonia in an employed population. Chest 125:2140–5.
12 . De Graeve D and Beutels P. (2004) Economic aspects of pneumococcal pneumonia: a review of the literature. Pharmacoeconomics 22:719–40.[CrossRef][Web of Science][Medline]
13 . Leophonte P. (2003) Community acquired pneumonia: higher mortality than traffic accidents. Rev Prat 53:1413–6.[Medline]
14 . Welte T, Marre R, Suttorp N. (2004) Competence network "community acquired pneumonia" (CAPNETZ). A first interim report. Internist (Berl.) 45:393–401.
15 . Neiss A, Elkharrat D, Kohno S, et al. (2004) Hospitalisation costs for community-acquired pneumonia (CAP) in France (F), Germany (D), Japan (J) and USA. Eur Resp J 24:Suppl 48, 8s.
16 . Boles M, Harbarth S, Chinn C, et al. Assessing the cost implications of microbiological sensitivity to antibiotic treatment in lower respiratory tract infections (LRTI). Abstracts of the Fortieth Interscience Conference on Antimicrobial Agents and Chemotherapy, Toronto, Canada, 2000(American Society for Microbiology, Washington, DC, USA) pp. 509 Abstract 2129.
17 . McGowan JE Jr. (2001) Economic impact of antimicrobial resistance. Emerg Infect Dis 7:286–92.[Web of Science][Medline]
18
.
Fine MJ, Auble TE, Yealy DM, et al. (1997) A prediction rule to identify low-risk patients with community-acquired pneumonia. N Engl J Med 336:243–50.
19 . Agence Francaise de Sécurité Sanitaire des Produits de Santé. (2002) Antibiotherapie par voie generale en practique courante: infections respiratoires basses de l'adulte http://www.agmed.sante.gouv.fr (6 February 2006, date last accessed).
20 . ERS Task Force Report. (1998) Guidelines for management of adult community-acquired lower respiratory tract infections. European Respiratory Society. Eur Respir J 11:986–91.[CrossRef][Web of Science][Medline]
21 . Welte T, Marre R, Suttorp N. (2006) What is new in the treatment of community-acquired pneumonia? Med Klin (Munich) 101:313–20.
22 . Laurichesse H, Gerbaud L, Baud O, et al. (2001) Hospitalization decision for ambulatory patients with community-acquired pneumonia: a prospective study with general practitioners in France. Infection 29:320–5.[CrossRef][Web of Science][Medline]
23 . Sanguinetti CM, De Benedetto F, Donner CF. (2005) GP management of community-acquired pneumonia in Italy: the ISOCAP study. Monaldi Arch Chest Dis 63:23–9.[Medline]
24
.
Torres A, Muir JF, Corris P, et al. (2003) Effectiveness of oral moxifloxacin in standard first-line therapy in community-acquired pneumonia. Eur Respir J 21:135–43.
25 . Petitpretz P, Arvis P, Marel M, et al. (2001) Oral moxifloxacin vs high-dosage amoxicillin in the treatment of mild-to-moderate, community-acquired, suspected pneumococcal pneumonia in adults. Chest 119:185–95.
26 . Hoeffken G, Meyer HP, Winter J, et al. (2001) The efficacy and safety of two oral moxifloxacin regimens compared to oral clarithromycin in the treatment of community-acquired pneumonia. Respir Med 95:553–64.[CrossRef][Web of Science][Medline]
27
.
Faich GA, Morganroth J, Whitehouse AB, et al. (2004) Clinical experience with moxifloxacin in patients with respiratory tract infections. Ann Pharmacother 38:749–54.
28 . Fogarty C, Grossman C, Williams J, et al. (1999) Efficacy and safety of moxifloxacin vs clarithromycin for community-acquired pneumonia. Infect Med 16:748–63.
29 . O'Doherty B and Muller O. (1998) Randomized, multicentre study of the efficacy and tolerance of azithromycin versus clarithromycin in the treatment of adults with mild to moderate community-acquired pneumonia. Azithromycin Study Group. Eur J Clin Microbiol Infect Dis 17:828–33.[CrossRef][Web of Science][Medline]
30 . Mathers DL, Hassman J, Tellier G. (2004) Efficacy and tolerability of once-daily oral telithromycin compared with clarithromycin for the treatment of community-acquired pneumonia in adults. Clin Ther 26:48–62.[CrossRef][Web of Science][Medline]
31
.
Tellier G, Niederman MS, Nusrat R, et al. (2004) Clinical and bacteriological efficacy and safety of 5 and 7 day regimens of telithromycin once daily compared with a 10 day regimen of clarithromycin twice daily in patients with mild to moderate community-acquired pneumonia. J Antimicrob Chemother 54:515–23.
32 . Ortqvist A, Valtonen M, Cars O. (1996) Oral empiric treatment of community-acquired pneumonia. A multicenter, double-blind, randomized study comparing sparfloxacin with roxithromycin. The Scandinavian Sparfloxacin Study Group. Chest 110:1499–506.
33 . Leophonte P, File T, Feldman C. (2004) Gemifloxacin once daily for 7 days compared to amoxicillin/clavulanic acid thrice daily for 10 days for the treatment of community-acquired pneumonia of suspected pneumococcal origin. Respir Med 98:708–20.[CrossRef][Web of Science][Medline]
34
.
Higuera F, Hidalgo H, Feris J. (1996) Comparison of oral cefuroxime axetil and oral amoxycillin/clavulanate in the treatment of community-acquired pneumonia. J Antimicrob Chemother 37:555–64.
35 . Gupta SK, Shirbhate NC, Khanna NN, et al. (1982) Enterolithiasis. J Postgrad Med 28:225–8.[Medline]
36
.
Shortridge VD, Zhong P, Cao Z, et al. (2002) Comparison of in vitro activities of ABT-773 and telithromycin against macrolide-susceptible and -resistant streptococci and staphylococci. Antimicrob Agents Chemother 46:783–6.
37 . Anzueto A and Norris S. (2004) Clarithromycin in 2003: sustained efficacy and safety in an era of rising antibiotic resistance. Int J Antimicrob Agents 24:1–17.[Web of Science][Medline]
38 . Farrell DJ, Morrissey I, Bakker S, et al. (2002) Molecular characterization of macrolide resistance mechanisms among Streptococcus pneumoniae and Streptococcus pyogenes isolated from the PROTEKT 1999–2000 study. J Antimicrob Chemother 50:Suppl S1, 39–47.[Abstract]
39 . Morrisey I, Colcolough A, Viljoen L, et al. (2004) The comparative in vitro activity of moxifloxacin against respiratory tract pathogens isolated during 2003 from Libra-targeted surveillance. Clin Microbiol Infect 10:Suppl 3, 117–8 Abstract P527.
40 . Brown SD and Rybak MJ. (2004) Antimicrobial susceptibility of Streptococcus pneumoniae, Streptococcus pyogenes and Haemophilus influenzae collected from patients across the USA, in 2001-2002, as part of the PROTEKT US study. J Antimicrob Chemother 54:Suppl 1, 7–15.
41
.
Felmingham D, Reinert RR, Hirakata Y, et al. (2002) Increasing prevalence of antimicrobial resistance among isolates of Streptococcus pneumoniae from the PROTEKT surveillance study, and compatative in vitro activity of the ketolide, telithromycin. J Antimicrob Chemother 50:Suppl S1, 25–37.
42 . Doern GV and Brown SD. (2004) Antimicrobial susceptibility among community-acquired respiratory tract pathogens in the USA: data from PROTEKT US 2000-01. J Infect 48:56–65.[CrossRef][Web of Science][Medline]
43 . Jacobs MR. (2003) How can we predict bacterial eradication? Int J Infect Dis 7:Suppl 1, S13–20.
44 . Dagan R. (2003) Achieving bacterial eradication using pharmacokinetic/pharmacodynamic principles. Int J Infect Dis 7:Suppl 1, S21–6.
45 . Bayer. Data on file. 2005.
46
.
Rybak MJ. (2004) Increased bacterial resistance: PROTEKT US—an update. Ann Pharmacother 38:S8–13.
47 . Dictionnaire Vidal (2005) 81st edition (Editions du Vidal, Paris).
48 . Red Book. Pharmacy's Fundamental Reference (2005) (Thompson Healthcare, Stamford, CT).
49 . Rote Liste (2005) , Frankfurt a.M. Rote Liste Service GmbH.
50 . Fantin B, Aubert JP, Unger P, et al. (2001) Clinical evaluation of the management of community-acquired pneumonia by general practitioners in France. Chest 120:185–92.
51 . Union des Caisses Nationales de Securite Sociale (2004) Nomenclature Generale des Actes Professionels (NGAP).
52 . PMSI (Programme de Médication des Systèmes d'Information). Agence Technique de l'information sur l'Hospitalisation, Paris http://www.atih.sante.fr (4 February 2006, date last accessed).
53 . Niederman MS, McCombs JS, Unger AN, et al. (1998) The cost of treating community-acquired pneumonia. Clin Ther 20:820–37.[CrossRef][Web of Science][Medline]
54 . American Medical Association. (2004) CPT Current Procedural Terminology 2005(Ingenix Inc., Salt Lake City).
55 . Blue Cross Insurance. (2005) Emergency Department Cost USA http://www.bcbs.com/mcrg/index.html (2 November 2005, date last accessed).
56 . Boerma W and Groenewegen P. (2001) GP home visiting in 18 European countries. Eur J Gen Pract 7:132–7.
57 . Kassenärztliche Bundesvereinigung. Einheitlicher Bewertungsmaßstab für ärztliche Leistungen (EBM) http://www.kbv.de/ (2 February 2006, date last accessed).
58
.
Laurichesse H, Robin F, Gerbaud L, et al. (1998) Empirical therapy for nonhospitalized patients with community-acquired pneumonia. Study Group of General Practitioners. Eur Respir J 11:73–8.
59 . Oregon State University. Student Health Services: Most Common Fees http://studenthealth.oregonstate.edu/costs/index.php (6 February 2006, date last accessed).
60 . Harbarth S, Albrich W, Brun-Buisson C. (2002) Outpatient antibiotic use and prevalence of antibiotic-resistant pneumococci in France and Germany: a sociocultural perspective. Emerg Infect Dis 8:1460–7.[Web of Science][Medline]
61 . Deutsche Krankenhausgesellschaft, Spitzenverbände der Krankenkassen & Verband der privaten Krankenversicherung. Website der Selbstverwaltung für German refined — diagnosis Related Groups http://www.dkgev.de (11 December 2005, date last accessed).
62 . ENL Medical. Costs White Blood Cell Count, WBC http://www.enlmedical.com (6 December 2004, date last accessed).
63 . Garber A. (2000) Advances in cost-effectiveness analysis of health interventions. Handbook of Health Economics (Elsevier Science B, New York)182–219.
64 . Reinert RR, Lutticken R, Reinert S, et al. (2004) Antimicrobial resistance of Streptococcus pneumoniae isolates of outpatients in Germany, 1999–2000. Chemotherapy 50:184–9.[CrossRef][Web of Science][Medline]
65
.
Singer ME, Harding I, Jacobs MR, et al. (2003) Impact of antimicrobial resistance on health outcomes in the out-patient treatment of adult community-acquired pneumonia: a probability model. J Antimicrob Chemother 51:1269–82.
66 . Lonks JR, Garau J, Medeiros AA. (2002) Implications of antimicrobial resistance in the empirical treatment of community-acquired respiratory tract infections: the case of macrolides. J Antimicrob Chemother 50:Suppl S2, 87–92.[Abstract]
67 . Gordon KA, Biedenbach DJ, Jones RN. (2003) Comparison of Streptococcus pneumoniae and Haemophilus influenzae susceptibilities from community-acquired respiratory tract infections and hospitalized patients with pneumonia: five-year results for the SENTRY Antimicrobial Surveillance Program. Diagn Microbiol Infect Dis 46:285–9.[CrossRef][Web of Science][Medline]
68 . Doern GV, Richter SS, Miller A, et al. (2005) Antimicrobial resistance among Streptococcus pneumoniae in the United States: have we begun to turn the corner on resistance to certain antimicrobial classes? Clin Infect Dis. 41:139–48.
69 . Reinert RR, Rodloff AC, Halle E, et al. (2004) Antibacterial resistance of community-acquired respiratory tract pathogens recovered from patients in Germany and activity of the ketolide telithromycin: results from the PROTEKT surveillance study (1999-2000). Chemotherapy 50:143–51.[CrossRef][Web of Science][Medline]
70
.
Reinert RR, Reinert S, van der LM, et al. (2005) Antimicrobial susceptibility of Streptococcus pneumoniae in eight European countries from 2001 to 2003. Antimicrob Agents Chemother 49:2903–13.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||