Abstract
Background
There is controversy about the preferred initial antibiotic therapy for peritoneal dialysis (PD)-related peritonitis. Quinolones have been used extensively in this setting, yet their long-term effectiveness is unknown.
Aim
To analyze the results of a protocol of treatment of PD-related peritonitis with ciprofloxacin, maintained over two decades.
Method
We analyzed the clinical outcome of 682 episodes of bacterial peritonitis treated with intraperitoneal ciprofloxacin monotherapy, and the time course of bacterial susceptibility to this antimicrobial, in a historical cohort of 641 PD patients (1988-2007). Main outcome variables included changes to initial therapy and rates of hospital admission, catheter removal, relapse, reinfection, PD dropout, and mortality. For comparisons we divided the study period into phases A (1988-1994), B (1995-2000), and C (2001-2007).
Results
The incidence of Staphylococcus aureus peritonitis decreased, while the incidences of polymicrobial and negative-culture peritonitis increased after phase A. In vitro susceptibility to ciprofloxacin decreased significantly only among coagulase-negative staphylococci (87.0% susceptible strains in phase A vs 70.0% in B and 70.1% in C, p = 0.006). Overall success rates (catheter not removed and ongoing PD after the episode) remained stable, at over 85%. However, the proportion of patients treated solely with ciprofloxacin declined from 75.7% (A) to 47.3% (B) to 32.4% (C) (p < 0.0005) and admission rates increased from 12.7% to 16.8% to 24.9% respectively (p = 0.001). These changes affected all the etiologic groups except culture-negative peritonitis. In vitro resistance to ciprofloxacin was a marker of multiresistance and correlated strongly with clinical outcome of peritonitis. Among isolates susceptible to ciprofloxacin, changing initial therapy for any reason also predicted a poor outcome.
Conclusions
Following satisfactory early results, the effectiveness of ciprofloxacin as monotherapy for PD-related peritonitis has declined markedly in the long term. This decline cannot be explained solely by a decrease of in vitro susceptibility to this antimicrobial, which was significant only among coagulase-negative staphylococci. Resistance to ciprofloxacin is a strong marker of in vitro multiresistance and poor clinical outcome of peritonitis.
Timely, appropriately designed, and properly implemented antibiotic therapy significantly increases the chances for an uncomplicated clinical course of PD-related bacterial peritonitis (4). It is obvious that the antibiotic susceptibility patterns of the etiologic agent(s) must guide the final therapy of any episode of peritonitis. On the other hand, the choice of the antibiotic schedule at diagnosis but before any microbiologic information is available is more controversial. There are some accepted general premises, including the convenience of empiric coverage of both gram-positive and gram-negative bacteria until an isolation is obtained (4) but, beyond these, there is a remarkable lack of unanimity concerning decisions such as the preferable antibiotic administration route, the use of continuous versus intermittent treatment schedules, and the duration of therapy. The spectrum of antibiotic regimes applied for the primary treatment of PD-related peritonitis is endless (5). Local differences in the epidemiology of these infections and a relative paucity of quality evidence to clarify the question (6) are two of the main factors favoring such heterogeneity of approaches.
Quinolones are a family of antibiotics that target bacterial topoisomerases II and IV and have been successful as antimicrobial agents during the past two decades (7). Their success is based on their favorable pharmacologic properties and wide spectrum of antibacterial activity. Older quinolones [including ciprofloxacin (Cpx)] have demonstrated an equilibrated activity against gram-positive and gram-negative bacteria (including Pseudomonas aeruginosa), while newer quinolones show improved activity against gram-positives, Chlamydia spp, Mycoplasma spp, mycobacteria, and selected anaerobes (8). Oral and intraperitoneal (IP) quinolones have been extensively used for the management of PD-related peritonitis, with overall satisfactory results (9-18). However, and similarly to other antibiotics, development of bacterial resistance may compromise the long-term results of this family of antimicrobials. In our center, we have used IP Cpx for initial treatment of PD-related peritonitis for the past two decades. This circumstance affords an excellent opportunity to assess the long-term effectiveness of this antimicrobial for the treatment of these infections, both on clinical grounds and in terms of in vitro susceptibility.
Population and Method
Overall Design and Aim of the Study
Since September 1988, IP Cpx has been systematically used as initial monotherapy of PD-related peritonitis in our center. The purpose of the present study is to review retrospectively the results of this sustained approach, with an emphasis on the time courses of the clinical results and bacterial susceptibility to this antibiotic during this period. Follow-up was closed at March 2007.
Inclusion and Exclusion Criteria
We considered for preliminary analysis all episodes of PD-related peritonitis recorded during the study period and then applied the following exclusion criteria:
Noninfectious peritonitis (eosinophilic, chemical, icodextrin);
Patients treated with antibiotics other than Cpx (alone or in combination with Cpx itself) at the start of the episode;
Infections caused by agent(s) not susceptible to any common empiric antibacterial regime used for PD-related peritonitis (fungi, mycobacteria, anaerobic bacteria); and
Inadequate clinical records.
Specifically, seemingly infectious but negative-culture peritonitis were considered for the clinical analysis. For the cases of mixed infections and for bacterial infections treated with antibiotics other than Cpx, we reviewed the in vitro susceptibility of the isolated bacteria to Cpx but did not consider these episodes for clinical analysis.
Study Population
During the study period, 641 patients were treated with PD in our center for at least 2 months, including 359 males (56.0%) and 282 females (44.0%). Mean age was 58 ± 16 (range 7 – 87) years and mean follow-up on PD was 25 ± 22 (range 2 – 122) months. The etiologic spectrum of renal diseases was standard; 217 patients (33.9%) were diabetic. The baseline Charlson comorbidity score was 5.1 ± 1.8 (range 2 – 11). Continuous ambulatory peritoneal dialysis (CAPD) was the main mode of PD therapy in 402 patients (62.8%), while 238 patients (38.2%) underwent automated forms of PD. All CAPD patients used Y-type systems (Baxter Healthcare, Deerfield, IL, USA; and Fresenius Medical Care, Heidelberg, Germany) and volumetric cyclers (HomeChoice, Baxter) predominated, except for a few patients using gravity cyclers (PacXtra, Baxter) during phase A.
Definitions
All the terms used for definitions in this study, including infectious and noninfectious peritonitis, exit-site infection, refractory peritonitis, relapse, and reinfection, are in agreement with current recommendations (4).
Treatment of Peritonitis and Exit-Site Infections
The protocol for empiric therapy of peritonitis included IP Cpx with a loading dose of 100 mg/L dialysate, followed by 50 mg/L as maintenance dose. During the first 2 years of follow-up, oral Cpx (500 mg three times per day) was substituted for IP Cpx after the fifth day of therapy (provided full remission was obtained), but we used the IP route exclusively thereafter due to poor tolerance to oral Cpx in a significant proportion of patients and distrust of the consistency of this route to generate adequate IP levels of the drug (13). Therapy was maintained for a minimum of 2 weeks and extended to 3 or 4 weeks, as indicated, according to standard criteria for management of these infections (slowly responsive, clinically aggressive, or relapsing infections).
The criteria for catheter removal and for changes to the initial antibiotic therapy were individualized, also following usual recommendations in the field. Refractoriness to therapy and in vitro resistance to Cpx were the main criteria for substitution of Cpx, while an expected aggressive or relapsing behavior (e.g., P. aeruginosa) or, again, partial refractoriness to Cpx alone were common reasons for adding a second antibiotic to Cpx.
Treatment of exit-site infections varied widely during the study period. In general, oral antibiotics (according to bacterial susceptibility) or intravenous vancomycin for gram-positive infections were extensively used until the mid-1990s. During the past 10 years, we have increasingly used topical mupirocin, Cpx, and gentamicin for local treatment of superficial infections, reserving oral or parenteral therapy for refractory or aggressive infections.
Systematic screening and treatment (mupirocin) of Staphylococcus aureus nasal carriage started in 1989; screening and treatment of catheter exit-site colonization started in 1996. Asymptomatic colonization of the exit site by gram-negative bacteria is not routinely treated with antibiotics.
Susceptibility Testing
Susceptibilities to Cpx, amoxicillin–clavulanic acid, cefazolin, cefotaxime, ceftazidime (only gram negatives), imipenem/cilastatin, vancomycin (only gram positives), and gentamicin were tested using a standard minimal inhibitory concentration (MIC) method (19) in the majority of cases. In 91 strains isolated during the first 3 years of the study, a standard disc diffusion method was used. An MIC of >2 μg/mL, or any category beyond full susceptibility by disc diffusion, was used for categorization of bacterial resistance to Cpx. We applied the following cutoff points for susceptibility according to MIC for other antibiotics: ≤8 μg/mL for amoxicillin-clavulanic acid, cefazolin, cefotaxime, and ceftazidime, and ≤4 μg/mL for imipenem/cilastatin, vancomycin, and gentamicin.
Strategy of Analysis
As designed, the study had two well-defined subjects of analysis: First, for each episode of peritonitis included in the study, we analyzed the following outcome variables: Cpx as final treatment (versus change/addition to/of other antibiotics), main reason for change of antibiotic (when indicated), hospital admission, catheter removal, relapse, reinfection, immediate dropout to hemodialysis after the episode, and death (under any circumstances) during the episode or immediately thereafter. Catheter removal, dropout to hemodialysis, and death formed a composite variable of overall adverse outcome. Second, we analyzed susceptibility to Cpx for each individual species of bacteria isolated from dialysate during the study period. Thus, we included separately for this analysis the different bacteria isolated in each polymicrobial infection, and also (as indicated above) bacteria isolated in mixed infections and bacterial infections treated with antibiotics other than Cpx, which were not considered for clinical analysis.
To analyze the time course of clinical results and bacterial susceptibility to Cpx, we arbitrarily divided the study period into three phases, namely A (1988 – 1994), B (1995 – 2000), and C (2001 – 2007).
Statistics
Numeric variables are presented as mean ± standard deviation (SD) and range when appropriate. Comparison between categorical variables was performed according to chi-square distribution and Fisher's exact test. Comparison between numeric variables was carried out using Student's t-test and analysis of variance. The SPSS 15.0 software (SPSS Inc., Chicago, IL, USA) was used to produce these analyses.
Results
Overview
The total number of patients treated with PD was 194 during phase A, 291 during phase B, and 341 during phase C. Mean age at the start of PD increased from 57.2 years during phase A to 64.6 during phase B, and decreased slightly to 61.0 during phase C (p = 0.001 ANOVA). The prevalence of automated PD versus total PD increased progressively, from 7.5% of patients during phase A to 34.3% during phase B to 40.5% during phase C (p < 0.0005). On the contrary, comorbidity scores, gender distribution, and prevalence of diabetes did not change significantly between the study phases.
During the study period, we recorded 832 episodes of peritonitis. The incidence of peritonitis decreased from 0.68 episodes/patient/year during phase A to 0.47 during phase B to 0.38 during phase C (p < 0.0005). Seventy episodes were fully excluded from analysis due to their noninfectious origin or common bacteria-free etiology, and two others due to incomplete clinical information. Seventy-eight episodes were considered only for Cpx susceptibility analysis (initial therapy including antibiotics other than Cpx and/or mixed flora with anaerobic bacteria or yeasts), leaving 682 episodes that met the inclusion criteria for clinical analysis. The etiologic spectrum in this final group for each study phase is presented in Table 1. The incidence of S. aureus peritonitis fell, while incidences of polymicrobial and negative-culture peritonitis increased after phase A; phases B and C showed similar distributions of etiologic agents of infection. The overall incidences of infections by methicillin-resistant Staphylococcus aureus (MRSA) and non-fermenting gram-negatives (including P. aeruginosa) were low. We recorded a single case of peritonitis by an extended-spectrum beta-lactamase (ESBL)-producing Escherichia coli during the whole study period.
Etiologic Spectrum of Bacterial Peritonitis
p < 0.05.
p = 0.01; other differences between phases not significant.
Table includes only episodes of bacterial peritonitis available for full clinical and in vitro ciprofloxacin susceptibility analysis. Figures denote number (%).
In Vitro Bacterial Susceptibility to Cpx
The main results are presented in Table 2. Overall, bacterial susceptibility to Cpx was reasonably well preserved during the study period. Coagulase-negative staphylococcus was the only etiologic group displaying a significant decline in susceptibility to Cpx, with E. coli displaying a similar nonsignificant trend. Susceptibility of enterococci to Cpx was stable at nonoptimal levels during follow-up. The small number of non-fermenting gram-negative bacteria did not permit a valid conclusion on the behavior of this subgroup after long-term use of Cpx.
In Vitro Bacterial Susceptibility to Ciprofloxacin According to Study Period
ID = data insufficient to estimate MIC90; MIC = minimal inhibitory concentration; NS = not significant.
Analysis includes all peritonitis bacterial isolations obtained during the study period. Figures denote MIC90 (μg/mL) and, in brackets, percentage of bacteria presenting full susceptibility (disc diffusion) or MIC ≤2 μg/mL for ciprofloxacin. Comparisons by chi-square analysis.
In vitro susceptibility to Cpx was strongly associated with susceptibility to the other antibiotics tested. Strains resistant to Cpx were also more frequently resistant to amoxicillin–clavulanic acid (22.5%, compared with 68.5% if susceptible to Cpx, p < 0.0005), cefazolin (35.5% vs 78.2%, p < 0.0005), cefotaxime (20.7% vs 90.2%, p < 0.0005), ceftazidime (13.3% vs 27.8%, p = 0.09), imipenem/cilastatin (10.5% vs 64.0%, p < 0.0005), and gentamicin (13.0% vs 52.3%, p < 0.0005). The proportion of gram-positive bacteria resistant to cefazolin increased from 23.5% during phase A to 43.0% and 48.9% during phases B and C respectively (p < 0.001). We observed no single instance of vancomycin resistance among gram-positives during follow-up.
Clinical Results
The main results of therapy are presented in Table 3. As shown, there was a marked progressive decrease in the fraction of patients whose final therapy was Cpx alone. Incidence of hospital admission increased during phase C; other outcome variables did not change significantly. The reasons for change of initial therapy included clinical refractoriness to Cpx alone (61.2%), in vitro resistance to Cpx despite adequate clinical response (9.7%), and medical decision despite ongoing clinical response and in vitro susceptibility to Cpx (28.9%). The main reasons for the latter were presumed or confirmed aggressiveness of the infection (as estimated by the clinician in charge), demanding antibiotic associations (Enterococcus spp, Enterobacteriaceae, P. aeruginosa, polymicrobial), clinically benign but slowly resolving infections (as commonly observed in peritonitis by Streptococcus spp), and, less frequently, simultaneous extraperitoneal infection by other bacteria (e.g., in urinary tract). The relative distributions of the three main causes for switching initial Cpx therapy were similar for the three study phases.
Clinical Results
PD = peritoneal dialysis; NS = not significant.
Figures denote % of cases. Comparisons by chi-square analysis.
In the 44.9% of cases in which Cpx was not the final therapy, that therapy was based most frequently on vancomycin alone (13.3%), beta-lactams alone (9.8%), or antibiotic associations (20.7%). The latter most frequently included vancomycin (76.0%), aminoglycosides (49.6%), Cpx itself (27.0%), beta-lactams (26.4%), and rifampicin (16.1%).
The decline in the fraction of patients treated solely with Cpx was significant in all the etiologic groups except negative-culture peritonitis, but was particularly patent among gram-positive bacteria and polymicrobial infections (Table 4). Hospital admission rates tended to increase in all the etiologic groups, but the difference was significant only in the case of gram-positive (10.6% in phase A, 13.7% in phase B, 20.8% in phase C; p = 0.01) and polymicrobial infections (9.1%, 18.2%, and 32.1%, respectively; p = 0.04). No etiologic group showed a decline in the composite variable of overall adverse outcome of peritonitis (Table 5).
Clinical Results According to Etiologic Agent. Ciprofloxacin as Final Therapy
NS = not significant.
Figures denote % of cases. Comparisons by chi-square analysis.
Clinical Results According to Etiologic Agent. Adverse Outcome
NS = not significant.
Figures denote % of cases resulting in catheter removal, dropout to hemodialysis, and/or death. Comparisons by chi-square analysis.
Clinical Results According to Bacterial Susceptibility to Cpx
Resistance to Cpx was a strong marker of poor outcome of peritonitis (Table 6). This circumstance stood to a similar degree when gram-positive and gram-negative bacteria were considered separately.
Clinical Outcomes According to In Vitro Susceptibility to Ciprofloxacin
PD = peritoneal dialysis.
Figures denote % of cases. Susceptibility defined by full susceptibility (disc diffusion) or minimal inhibitory concentration ≤2 μg/mL for ciprofloxacin. Polymicrobial peritonitis categorized as resistant if at least one isolate was resistant to ciprofloxacin. Comparisons by chi-square analysis.
In the particular case of coagulase-negative staphylococci, Cpx alone was the final therapy in 68.6% and 7.1% of susceptible and resistant strains respectively (p < 0.0005). Relapse rates were 12.7% and 28.6% (p = 0.02), hospital admission rates were 8.5% and 19.6% (p = 0.01), and catheter removal was necessary in 4.8% and 7.1% (NS), respectively (PD dropout and mortality rates were too low to permit statistical analysis).
Changes to initial Cpx therapy were also predictive of outcome of infections by bacteria susceptible to Cpx (Table 7). Clinical refractoriness confirmed its well-known prognostic significance, but medical decision also predicted a poor outcome, probably due to the type of infections leading to such a decision (see above).
Clinical Outcome of Infections Due to Bacteria Susceptible to Ciprofloxacin, According to Maintenance/Change of Initial Therapy
PD = peritoneal dialysis; NS = not significant.
Figures denote % of patients. Polymicrobial peritonitis included only if all causative micro-organisms susceptible to ciprofloxacin. Comparisons by chi-square analysis.
Discussion
There is now ample experience in the use of quinolones for the treatment of PD-related peritonitis. Ciprofloxacin has been reported on most frequently (9-14,18,20-25) but other related drugs, including ofloxacin, levofloxacin, and pefloxacin, have also been used in this setting (15-17). Quinolones have been used alone and in association with other antimicrobials, and both oral and IP routes have been employed successfully. Several pharmacokinetic studies have demonstrated adequate plasma and dialysate levels with both routes of administration, although with significant interindividual variability (9,11,12,22,26-29). The results of a recent meta-analysis suggest that the IP route may yield more consistent results (6).
The results of our study could, at first sight, endorse the efficiency of a protocol based on IP Cpx for primary therapy of PD-related peritonitis, with overall success rates consistently over 85% during a period approaching two decades (Table 5). However, a more meticulous analysis of the data downplays this impression. Hospital admission rates increased significantly during follow-up and the fraction of patients treated solely with Cpx decreased progressively during the study period (Table 3), indicating declining effectiveness of this antimicrobial in controlling the full course of these infections. Bacterial resistance might explain this decline only in the case of coagulase-negative staphylococcal species (Table 2). Quinolones are not drugs of choice for the treatment of infections by this family of bacteria, and their relative ineffectiveness in the setting of PD-related infections has been reported previously (9,14,30). Concerning other gram-positives, as also with gram-negative bacteria (with the possible exception of E. coli), in vitro susceptibility to Cpx remained remarkably stable across the study periods in our study. Resistance to Cpx and other quinolones is being increasingly reported in infections affecting the population at large, especially among aggressive bacteria such as MRSA, Enterococcus spp, and Acinetobacter spp, as also with ESBL-producing and non-fermenting gram-negative bacteria (31). Infections by these microbes demand individualized antimicrobial approaches and are not adequately covered by most regimes commonly used for initial treatment of PD-related peritonitis (4,5). Resistance to quinolones is known to occur by at least three mechanisms (31). The first involves single (as in S. aureus or P. aeruginosa) or multiple (as in E. coli) gene mutations for topoisomerases II (gram-positives) and IV (gram-negatives). The second implicates reduced drug permeation and altered regulation of active efflux mechanisms; this mechanism is frequently associated with multiresistance to different antimicrobials. Finally, low-level resistance can be mediated by plasmids producing the Qnr protein.
Analysis of our data provides some clues to the reasons why admission rates increased and initial Cpx monotherapy was substituted in a majority of cases during phases B and C (Table 3). First, some etiologic groups with a higher risk of treatment failure, including polymicrobial infections (Table 1), were clearly more incident in these phases. Second, the prevalence of elderly patients increased during phases B and C, with higher expected morbidity and hospital admission rates in this subgroup. Third, even though in vitro susceptibility to Cpx decreased significantly only among coagulase-negative staphylococci, these bacteria caused 32.2% of the infections, bringing overall susceptibility rates below 80% during phases B and C (Table 2). Finally, our group has progressively adopted clinical recommendations emphasizing the convenience of obtaining rapid and consistent remission of peritonitis (4) to maximize the chances for patient and technique survival. According to this notion, ongoing refractoriness and even slow remission lead to more precocious revisions of therapy, while complicated etiologic subsets have become systematically treated with antibiotic associations. Agile and good-quality bacteriologic information is essential to implement these clinical decisions.
Our results show that in vitro resistance to Cpx is a strong prognostic marker of the outcome of PD-related peritonitis (Table 6). The explanation for this circumstance seemingly goes beyond simple delay to an effective antibiotic therapy. Naturally, aggressive bacteria were overrepresented among Cpx-resistant strains (Table 2), and Cpx resistance is known to be commonly associated with multiresistance patterns (7,30), as observed in our study. On the other hand, even among Cpx-susceptible bacteria, changes to initial Cpx therapy portended a poorer prognosis (Table 7), which is not difficult to explain, both in the case of clinically refractory infections and when changes responded to a medical decision, the latter frequently taken in the presence of infections by aggressive or multiple bacteria.
Limitations of our study include its retrospective nature and lack of a control group. The low incidence of infections by non-fermenting gram-negatives, MRSA, and ESBL-producing bacteria does not permit analyses of these specific etiologic subsets. On the other hand, the results of our study may not be applicable to protocols combining Cpx with other antibiotics for initial therapy for PD-related peritonitis (9,18,23-25). Also, newer quinolones that display higher activity against gram-positive bacteria (8) could represent a significant improvement in the management of these infections, although evidence for this is still needed. Finally, our study does not permit adequate analysis of the results of oral Cpx for PD-related peritonitis. We renounced this route of administration due to poor tolerance to the prescribed dose (500 mg three times per day) in a significant proportion of patients and to distrust of the consistency of this route to generate adequate IP levels of the drug, based on our own results (13). Other groups have reported better outcomes using lower doses of Cpx (250 – 500 mg two times per day) (9,11,14,21,23). In these studies, plasma and dialysate levels of the drug were usually adequate although, as previously mentioned, with significant interindividual variability.
In summary, after two decades, IP Cpx alone does not stand as a desirable option for initial therapy of PD-related peritonitis, as its efficiency for full treatment of these infections has declined markedly in the long-term. This decline cannot be explained solely by a decrease in in vitro susceptibility to this antimicrobial, which was significant only among coagulase-negative staphylococci. Bacterial resistance to Cpx is a strong marker of antibiotic multiresistance and of poor clinical outcome of peritonitis. The progressive complexity of the resistance patterns of bacteria to antimicrobials, the growing prevalence of high-risk patients in PD programs, and the convenience of obtaining rapid and consistent remission of peritonitis to improve patient and technique survival may favor the choice of expanded rather than restricted antibiotic regimes for the primary treatment of PD-related peritonitis.
Footnotes
M. Pérez Fontán is a regular consultant to Baxter Inc.