Journal of Hospital Infection (1990) 15 (Supplement
A), 13-22
Challenges in the design of trials to evaluate antibacterial agents in serious infections S. Ragnar
Norrby
Department of Infectious Diseases, University of Lund, Sweden
Introduction Decisions on the introduction of a new antibiotic for treatment of serious infections should be based on its (i) comparative in-vitro activity; (ii) pharmacokinetic properties; (iii) clinical and bacteriological efficacy in comparative clinical trials; and (iv) safety profile. For most new antibiotics abundant documentation is normally available on their antibacterial activities, including, when relevant, those against nosocomial pathogens such as Pseudomonas aeruginosa, Pseudomonas spp., Enterobacter spp., Serratia spp., Acinetobacter spp. and Citrobacter spp. The documentation of the pharmacokinetic properties of a new antibiotic is also quite extensive in most cases. However, it is not unusual that information is insufficient about the kinetics of an antibiotic in special patient groups, for example elderly patients, neonates and infants. Such data should be required before an antibiotic is accepted for routine therapy in hospitalized patients (Working Party of the British Society of Antimicrobial Chemotherapy, 1989). This article will deal with the optimal conduct of comparative efficacy and safety trials of antibacterial agents in seriously ill patients. As an example, an ongoing prospective, comparative, randomized, open, evaluator-blinded multi-clinic trial of ceftazidime vs. imipenem/cilastatin in patients with serious hospital acquired pneumonia, urinary tract infections or septicaemia, will be discussed. Study design The trial discussed here was designed as an open comparative trial in patients with serious hospital-acquired bacterial infections. Although a double-blind design would have been preferable to minimize investigators’ bias, it was decided to make the study open. The main reason for this was that, in Europe, it is very difficult to obtain an around-the-clock service Correspondence to: Professor S. R. Norrby, Hospital, S-22185 Lund, Sweden.
019&6701/90/0A0013+
Department
IO SO3.00/0
of Infectious
Diseases,
Lund
0 1990 The Hospital
13
University
Infection
Society
14
S. R. Norrby
from hospital pharmacy units, a prerequisite for an effective double-blind design of a trial comparing injectable antibiotics which do not have identical appearances and solubility properties. In other trials, for example of two aminoglycosides with similar dose regimens, blinding is possible. To obtain an unbiased evaluation of the results of this study, a co-ordinating committee was formed; the chairman of this committee has the responsibility of assessing all case report forms using a brief evaluation form (Table I). Case report forms, about which there is disagreement between the evaluations of the assessor and the investigator, are reviewed by the entire committee together with the responsible trial monitor from the sponsor. In the planning of the trial it was self-evident that it must be a comparative study. The choice of trial drugs was based on the fact that ceftazidime and imipenem/cilastatin have similar antibacterial activity against those pathogens which most commonly cause serious nosocomial infections (Neu & Labtharikul, 1982; Kropp et al., 1985). Randomization of the patients to the two trial arms is performed by sealed envelopes which are opened when the decision to enter a patient has been
Table I. Randomization No: Protocol violations: No Patient’s evaluability:
Form used for blinded evaluation of case reports Yes
Minor Major
Clinical efficacy only Clinical + bacteriological Non-evaluable
Specify Specify efficacy
Reason for exclusion from evaluation of bacteriological e&acy: Reason for exclusion from evaluation of clinical ejicacy: Clinical outcome: Not evaluable Cure Improvement
Relapse Failure
Deviation from investigator’s assessment: No Yes Bacteriological
outcome: Cleared Partial clearance Failure Relapse
Clinical adverse events: Investigator’s Laboratory
Colonization Superinfection New site Unassessable
assessment accepted:
adverse events: Investigator’s
assessment accepted
No Yes Suggested changes: No Yes Suggested changes:
Antibiotic
trials
in serious
infections
15
taken. This procedure is not optimal; preferably the breaking of the code should be made centrally, for example by a telephone call to an outside institution. However, practical difficulties make such procedures difficult. Prospective stratification (except by centre) is not used in this trial. The reasons for this are that the study includes a large number of patients, and that the number of prognostic factors affecting the outcome of hospital-acquired infections is so large that prospective stratification would result in too many treatment arms. Prospective stratification also invites mistakes when the randomization procedures are not centralized. In this stratification for important prognostic factors will be made study, retrospectively, and efficacy and safety will be assessed by a multi-variate analysis technique (see below). Calculation
of patient
sample
size
The sample size calculation is of vital importance for planning a clinical trial. In this study, clinical efficacy is used as the main end-point. Assuming that 85% of the patients randomized to the imipenem/cilastatin group will be assessed as clinically cured or improved, one would need 160 patients in each trial group to be able to accept the null-hypothesis that there is not a > 10% difference between the clinical response rates in the two treatment groups with a statistical power of 80% (type II error 0.2) and a level of significance (type I error) of 0.05. These calculations were routine ones. What turned out to be difficult to predict was the frequency of exclusions from evaluation of efficacy after randomization; that is, the frequency of protocol violations, premature discontinuation of treatment due to adverse effect or patients’ death or insufficient data during or after treatment. Before this study began it was calculated that 30% of randomized patients would have to be excluded from randomization due to such reasons. Thus, 460 patients would have to be entered into the study to achieve the goal of 160 evaluable patients in each treatment arm. When 158 case report forms had been assessed, 18% of the patients were found to be non-evaluable for efficacy, indicating that the investigators were performing better than expected. A consequence of the large number of patients required to test the null-hypothesis is that several centres had to be enrolled for the trial. Assuming that all centres entered 20 patients each, 23 centres were contacted and agreed to participate. A choice then had to be taken as to whether the study should be carried out as a fully co-ordinated multi-centre trial (MCT) or as a multiple independent trial (MIT) with less strict co-ordination. Since investigators in six countries are now involved, an MCT was considered to be too large an undertaking with the considerable effort and costs involved in arranging regular co-ordination meetings with all investigators. The choice therefore became an MIT study-but with one and the same trial protocol for each centre. This also had the advantage over
S. R. Norrby
16
a true MCT design in that centres could start the study at different and that new centres could be added during the course of the trial.
times
Definitions
It is important that the terminology used in a trial protocol is interpreted in the same way by all parties involved in the study. Preferably, internationally accepted definitions should be used. However, such definitions are often lacking. Therefore, definitions were created for conformity in the trial and were to be agreed upon by the sponsor and the investigators. For this trial several terms were defined, as shown in Table II.
Table II.
Definitions of terms used in a clinical trial of cejtazidime vs. imipenem/cilastatin
Term Hospital-acquired infection Bacterial pneumonia
Clinical (culturenegative) pneumonia Bacteriologically documented septicaemia Clinical (culturenegative) septicaemia Acute bacterial pyelonephritis Complicated urinary tract infection Clinical cure Clinical improvement Clinical relapse Clinical failure
Definition An infection with onset 248 h after admission to hospital The isolation of an organism considered to be clinically relevant from purulent sputum, bronchial secretion, bronchial washing or lung tissue in the presence of new or progressive radiologically demonstrable pulmonary infiltrates without a known non-infectious cause New or progressive radiologically demonstrable pulmonary infiltrates without a known non-infectious cause The occurrence of one or more positive blood cultures with signs and symptoms of clinical septicaemia Clinical deterioration and rigors in association with unstable haemodynamic measurements and/or coagulopathy consistent with sepsis in the absence of positive blood cultures Fever associated with loin pain or tenderness, and accompanied by significant bacteriuria (>, 10’ cfu ml-’ of voided or catheter urine, or 2 lo3 cfu ml-’ in a sample obtained by bladder puncture), pyuria, and sometimes haematuria The presence of significant bacteriuria in a systemically unwell, febrile (> 38’C) patient in whom structural or functional abnormalities of the urinary tract, known to predispose to or permit persistence of infection, are present Clinical symptoms and signs of infection subsided with complete resolution of active infection during the treatment period Clinical signs and symptoms subsided but without complete resolution of active infection during the treatment period Recurrence of clinical signs and symptoms of the initial infection during the follow-up period, after initial cure or improvement Introduction of a new antibiotic agent less than 72 h after initiation of treatment with the study drug regimen because of a rapid deterioration in clinical condition, e.g. onset of unstable vital signs, development of shock etc., or no apparent response to at least 72 h of treatment with the study drug regimen resulting in the introduction of a new antibiotic agent or death due to infection
Antibiotic
Inclusion
trials
in serious
and exclusion
infections
17
criteria
The inclusion and exclusion criteria used in this study are given in Table III. It is of general importance that such criteria do not exclude too many patients from entry. Exclusion criteria which are too strict will markedly limit the external validity of the results obtained. Thus, exclusion criteria such as advanced age, organ failure or specific concomitant medication (other than with antibiotics) should not be used. It is also doubtful whether criteria such as ‘likelihood of death within 24 hours’ should be allowed; such patients are given antibiotics if they suffer from an infection and represent an important challenge for any new broad-spectrum parenteral antibiotic. An important exclusion criterion which should be used in all studies of this type is that patients who have already been included in the study once should not be allowed to enter it a second time. In studies of neutropenic patients with fever, multiple entry into clinical trials are common and may make the evaluation difficult. For example, in patients entered more than once it is not possible to know whether previous entries have resulted in selection of bacterial resistance in the normal flora, and it may be difficult to assess adverse reactions. In this context it is important that the patients included in the trial are representative of a larger group of patients with the same or similar conditions. To achieve that goal one can use a system where consecutive patients are entered; that is, all patients who fulfil the inclusion but not the exclusion criteria are included in the trial. However, normally that is not possible to achieve in a clinical setting where a large number of hospital staff members are involved in the care of the patients. Another alternative is then to use a reject log, meaning that all patients who fulfil the
Table
III.
Inclusion
and exclusion
criteria
in a study of ceftazidime
Inclusion criteria 1 Hospitalized patients 3 18 years of age 2 Male or female sex 3 Suspected nosocomial bacterial pneumonia, septicaemia infections requiring parenteral antibiotic treatment 4 Onset of infection more than 48 h after randomization 5 Informed consent
vs. imipenem/cilastatin
or severe urinary
tract
Exclusion criteria 1 Hypersensitivity to cephalosporins or carbapenems 2 Immediate type hypersensitivity reactions to penicillins 3 Neutrophil count of less than 1000 mm-j 4 Likelihood of death within 24 h 5 Isolation before therapy of pathogens known or suspected to be intrinsically resistant to either of the study drug regimens 6 Primary diagnosis of a fungal, viral or mycobacterial infection 7 Therapy with ceftazidime or imipenem within the 4 weeks prior to study entry 8 Antibiotic therapy during the 3 days prior to entry into the protocol unless the patient remains symptomatic 9 Previous entry in this trial
18
S. R. Norrby
criteria for entering the study, but who do not do so, are listed with basic clinical information and reasons for not being included in the trial. At the completion of the study the data in the reject log are then compared with the corresponding data for patients entered, and no major differences in terms of prognostic factors should exist. In the ceftazidime vs. imipenem/cilastatin trial, such procedures were unfortunately not included in the protocol. We can therefore expect some difficulties in evaluating the representativeness of the study sample for a larger population of patients with hospital-acquired infections. Diagnostic
procedures
In antibiotic studies, diagnostic procedures aiming at verifying the aetiology of the infections treated are of vital importance. Such procedures are easily defined for patients with urinary tract infections or septicaemia where urine or blood cultures yield the results. However, in lower respiratory tract infections, the bacteriological diagnosis is very difficult, both in terms of requirements put on the investigators and in terms of interpretation of the findings. In this trial, the investigators are required to obtain sputum, transtracheal aspirate, bronchoscopic aspirate and/or lung biopsies. During the course of the study it has become apparent that the procedures for sputum bacteriology vary considerably from centre to centre. In most centres it is, for example, not routine to investigate sputum samples by microscopy. This results in the loss of two vital pieces of information: whether the samples contain squamous epithelial cells (indicating that they represent the pharyngeal and not the lower respiratory tract flora), and whether the organisms later isolated in culture were the predominant ones. Moreover, the preparatory procedures vary between laboratories. Some laboratories wash the sputum samples before further processing, while others have no such routines. Difficulties have also been encountered in the interpretation of results obtained with other samples from the lower respiratory tract. The term ‘transtracheal aspirate’ was found to be misinterpreted as equivalent to any sample taken through a tracheal tube. As a result of the above difficulties, a very large number of isolates have been reported from the lower respiratory tract samples taken from the patients assessed. Without guarantees that the samples do not represent the oropharyngeal flora, it has become difficult to assess the clinical relevance of the isolates. A lesson learned from this is that, before the trial, the laboratories involved should have been contacted to provide detailed information of their procedures for handling sputum and bronchial secretion samples. In some instances it would probably also have been advisable to try to change current routines for the purpose of the study in order to achieve conformity between the centres. Another field where considerable difficulties have been encountered is the interpretation of the radiological findings in patients with lower respiratory
Antibiotic
trials
in serious
infections
19
tract infections. A large proportion of the patients encountered have been treated in intensive care units and have received assisted ventilation. In such patients cardiac decompensation with lung oedema and/or pleural effusion are not uncommon and may make the diagnosis of inflammatory infiltrates of the lungs difficult. The optimal way to solve this problem would be to copy all X-ray films and have them assessed by one radiologist with special experience within the field of pulmonary radiology. Treatment
Clinical trials of parenteral antibiotics have a common weakness in that, to be evaluable for efficacy, the patient must have been treated with the trial antibiotic only, and that no subsequent oral treatment is allowed. The normal clinical procedure is to start treatment with a parenteral antibiotic and then to change to oral treatment when the patient has improved. According to tradition, and also to currently used guidelines for clinical trials of antibiotics, such procedures will render the patients non-evaluable for efficacy. In this trial the traditional approach is followed. However, the situation is less artificial since, of the patients so far enrolled, 73% have been in intensive care units and 48% have had rapidly fatal or ultimately fatal underlying diseases. Thus, most of the patients have been unable to take oral medications during the entire treatment course. The protocol states that patients should receive treatment with either ceftazidime-2 g every 12 h or imipenem/cilastatin-SOO/SOO mg every 8 h. In patients with septicaemia originating from the abdomen, metronidazole could be added when randomized to ceftazidime. Instructions are given for dose reductions in patients with impaired renal function and differ considerably between the two regimens. These differences in treatment between the two groups further emphasize the difficulties involved in making a study like this double-blind. In fact, in seriously ill patients with frequent impairment of renal function, a double-blind comparison of ceftazidime and imipenem/cilastatin seems impossible. Analysis
of efficacy
Serious infections treated with parenteral antibiotics occur only rarely in patients who are otherwise healthy. In most cases such infections result from complications to underlying diseases or surgical procedures. As a consequence of these factors it is often difficult to assess the role of the antibiotic itself in the outcome of the infection; therapeutic measures directed against the underlying condition may very well affect the outcome. In the analysis of efficacy in a trial of a parenteral antibiotic used for the treatment of serious infections, these factors should be taken into account. The results are clearly not suitable for a straightforward significance analysis in which cure rates of different treatment groups are compared.
20
S. R. Norrby
Instead, multi-variate analysis-in which prognostic factors are weighedshould be used. Such procedures have been described by Moore et al. (1984) in studies of aminoglycosides. In the ceftazidime vs. imipenem/cilastatin trial, the primary infection in patients so far entered is pneumonia in 65%, septicaemia in 22%, and severe urinary tract infection in 13%. This will lead to three subgroups which will have to be analysed separately and, especially in the pneumonia group, it is foreseen that there will be a multitude of important prognostic factors-for example, assisted ventilation, underlying cardiac or respiratory diseases, and causative bacterial agent. The following series of events are planned when the study is completed. Firstly, prognostic factors will be identified for a multi-variate analysis. Secondly, subgroups of infections will be decided upon. Thirdly, analyses will be performed, and only at that stage will the patients be grouped according to treatment groups. Thus, those performing the analyses will be blinded as to treatment allocation until all decisions about classifications of outcome and risk factors are decided upon. A similar type of analysis will be performed for safety. It is already clear that the frequencies of adverse effects will be low. It is our feeling that the main reason for this is that the investigators are less inclined to search for, or find, adverse events in severely ill patients than in those with mild or moderately severe infections. Lessons learned
from the ceftazidime
vs. imipenemlcilastatin
study
The most disturbing finding during the course of this study has been that patient recruitment has been considerably slower than expected (Figure 1). As a result, the time for recruiting a sufficient number of patients has had to be extended, and more centres than initially planned have been recruited. In retrospect, a basic mistake in the planning phase was to base the patient enrolment schedule on the estimates done by the investigators. We should instead have performed a pilot study which would have served several we would have achieved a better appreciation of the purposes. Firstly, number of patients available at each centre. Secondly, we could have tested the protocol and the case report form, and would then have been able to avoid most of the mistakes discussed above. Conclusions Multi-clinic or multi-centre trials in patients with hospital-acquired infections are difficult to perform with a high scientific quality. This is especially true in patients with hospital-acquired pneumonia where the main problems are that it is extremely difficult to achieve a correct microbiological diagnosis due to the difficulty of collecting samples representative of the lower respiratory tract, and also because of the multitude of bacterial species found, which makes it almost impossible to
Antibiotic
Figure 1. Actual and projected ceftazidime vs. imipenem/cilastatin
trials
in serious
rate of patient entry for the treatment
21
infections
at various of serious
stages of a clinical trial nosocomial infections.
of
separate commensals from pathogens. Optimally, a study of this kind should include a training period during which one ascertains that the procedures at the microbiological laboratories involved are optimal. During that period a pilot study should be performed in order to achieve adequate estimates of patient entry rate and to test the forms used in the study as well as the design of the protocol. Despite the above difficulties, it is quite clear already at this stage, that the material collected in a study of this kind will offer valuable information about the natural course of hospital-acquired infections, in addition to safety and efficacy data on the trial drugs. However, prerequisites for achieving adequate information are that the analysis be performed in a blinded fashion, although the design of the study is open, and that multi-variate analysis be used. References Kropp, H., Gerckens, L., Sundelof, J. G. & Kah an, F. M. (1985). Antibacterial activity of imipenem: the first thienamycin antibiotic. Reoiew of Znfectious Diseases 7 (Suppl. 3), S389410. Moore, R. D., Smith, C. R., Lipsky, J. J., Mellits, E. D. & Lietman, P. S. (1984). Risk factors for nephrotoxicity in patients treated with aminoglycosides. Ann& of Internal Medicine 100, 352-357. Neu, H. C. & Labthavikul, P. (1982). Antimicrobial activity and beta-lactamase stability of ceftazidime, an aminothiazolyl cephalosporin. Antimicrobial Agents and Chemotherapy 4, 11-18. Working Party of the British Society of Antimicrobial Chemotherapy (1989). The clinical ;;kation of antibacterial drugs. Journal of Antimicrobial Chemotherapy 23 (Suppl. B),
A), 13-22
Challenges in the design of trials to evaluate antibacterial agents in serious infections S. Ragnar
Norrby
Department of Infectious Diseases, University of Lund, Sweden
Introduction Decisions on the introduction of a new antibiotic for treatment of serious infections should be based on its (i) comparative in-vitro activity; (ii) pharmacokinetic properties; (iii) clinical and bacteriological efficacy in comparative clinical trials; and (iv) safety profile. For most new antibiotics abundant documentation is normally available on their antibacterial activities, including, when relevant, those against nosocomial pathogens such as Pseudomonas aeruginosa, Pseudomonas spp., Enterobacter spp., Serratia spp., Acinetobacter spp. and Citrobacter spp. The documentation of the pharmacokinetic properties of a new antibiotic is also quite extensive in most cases. However, it is not unusual that information is insufficient about the kinetics of an antibiotic in special patient groups, for example elderly patients, neonates and infants. Such data should be required before an antibiotic is accepted for routine therapy in hospitalized patients (Working Party of the British Society of Antimicrobial Chemotherapy, 1989). This article will deal with the optimal conduct of comparative efficacy and safety trials of antibacterial agents in seriously ill patients. As an example, an ongoing prospective, comparative, randomized, open, evaluator-blinded multi-clinic trial of ceftazidime vs. imipenem/cilastatin in patients with serious hospital acquired pneumonia, urinary tract infections or septicaemia, will be discussed. Study design The trial discussed here was designed as an open comparative trial in patients with serious hospital-acquired bacterial infections. Although a double-blind design would have been preferable to minimize investigators’ bias, it was decided to make the study open. The main reason for this was that, in Europe, it is very difficult to obtain an around-the-clock service Correspondence to: Professor S. R. Norrby, Hospital, S-22185 Lund, Sweden.
019&6701/90/0A0013+
Department
IO SO3.00/0
of Infectious
Diseases,
Lund
0 1990 The Hospital
13
University
Infection
Society
14
S. R. Norrby
from hospital pharmacy units, a prerequisite for an effective double-blind design of a trial comparing injectable antibiotics which do not have identical appearances and solubility properties. In other trials, for example of two aminoglycosides with similar dose regimens, blinding is possible. To obtain an unbiased evaluation of the results of this study, a co-ordinating committee was formed; the chairman of this committee has the responsibility of assessing all case report forms using a brief evaluation form (Table I). Case report forms, about which there is disagreement between the evaluations of the assessor and the investigator, are reviewed by the entire committee together with the responsible trial monitor from the sponsor. In the planning of the trial it was self-evident that it must be a comparative study. The choice of trial drugs was based on the fact that ceftazidime and imipenem/cilastatin have similar antibacterial activity against those pathogens which most commonly cause serious nosocomial infections (Neu & Labtharikul, 1982; Kropp et al., 1985). Randomization of the patients to the two trial arms is performed by sealed envelopes which are opened when the decision to enter a patient has been
Table I. Randomization No: Protocol violations: No Patient’s evaluability:
Form used for blinded evaluation of case reports Yes
Minor Major
Clinical efficacy only Clinical + bacteriological Non-evaluable
Specify Specify efficacy
Reason for exclusion from evaluation of bacteriological e&acy: Reason for exclusion from evaluation of clinical ejicacy: Clinical outcome: Not evaluable Cure Improvement
Relapse Failure
Deviation from investigator’s assessment: No Yes Bacteriological
outcome: Cleared Partial clearance Failure Relapse
Clinical adverse events: Investigator’s Laboratory
Colonization Superinfection New site Unassessable
assessment accepted:
adverse events: Investigator’s
assessment accepted
No Yes Suggested changes: No Yes Suggested changes:
Antibiotic
trials
in serious
infections
15
taken. This procedure is not optimal; preferably the breaking of the code should be made centrally, for example by a telephone call to an outside institution. However, practical difficulties make such procedures difficult. Prospective stratification (except by centre) is not used in this trial. The reasons for this are that the study includes a large number of patients, and that the number of prognostic factors affecting the outcome of hospital-acquired infections is so large that prospective stratification would result in too many treatment arms. Prospective stratification also invites mistakes when the randomization procedures are not centralized. In this stratification for important prognostic factors will be made study, retrospectively, and efficacy and safety will be assessed by a multi-variate analysis technique (see below). Calculation
of patient
sample
size
The sample size calculation is of vital importance for planning a clinical trial. In this study, clinical efficacy is used as the main end-point. Assuming that 85% of the patients randomized to the imipenem/cilastatin group will be assessed as clinically cured or improved, one would need 160 patients in each trial group to be able to accept the null-hypothesis that there is not a > 10% difference between the clinical response rates in the two treatment groups with a statistical power of 80% (type II error 0.2) and a level of significance (type I error) of 0.05. These calculations were routine ones. What turned out to be difficult to predict was the frequency of exclusions from evaluation of efficacy after randomization; that is, the frequency of protocol violations, premature discontinuation of treatment due to adverse effect or patients’ death or insufficient data during or after treatment. Before this study began it was calculated that 30% of randomized patients would have to be excluded from randomization due to such reasons. Thus, 460 patients would have to be entered into the study to achieve the goal of 160 evaluable patients in each treatment arm. When 158 case report forms had been assessed, 18% of the patients were found to be non-evaluable for efficacy, indicating that the investigators were performing better than expected. A consequence of the large number of patients required to test the null-hypothesis is that several centres had to be enrolled for the trial. Assuming that all centres entered 20 patients each, 23 centres were contacted and agreed to participate. A choice then had to be taken as to whether the study should be carried out as a fully co-ordinated multi-centre trial (MCT) or as a multiple independent trial (MIT) with less strict co-ordination. Since investigators in six countries are now involved, an MCT was considered to be too large an undertaking with the considerable effort and costs involved in arranging regular co-ordination meetings with all investigators. The choice therefore became an MIT study-but with one and the same trial protocol for each centre. This also had the advantage over
S. R. Norrby
16
a true MCT design in that centres could start the study at different and that new centres could be added during the course of the trial.
times
Definitions
It is important that the terminology used in a trial protocol is interpreted in the same way by all parties involved in the study. Preferably, internationally accepted definitions should be used. However, such definitions are often lacking. Therefore, definitions were created for conformity in the trial and were to be agreed upon by the sponsor and the investigators. For this trial several terms were defined, as shown in Table II.
Table II.
Definitions of terms used in a clinical trial of cejtazidime vs. imipenem/cilastatin
Term Hospital-acquired infection Bacterial pneumonia
Clinical (culturenegative) pneumonia Bacteriologically documented septicaemia Clinical (culturenegative) septicaemia Acute bacterial pyelonephritis Complicated urinary tract infection Clinical cure Clinical improvement Clinical relapse Clinical failure
Definition An infection with onset 248 h after admission to hospital The isolation of an organism considered to be clinically relevant from purulent sputum, bronchial secretion, bronchial washing or lung tissue in the presence of new or progressive radiologically demonstrable pulmonary infiltrates without a known non-infectious cause New or progressive radiologically demonstrable pulmonary infiltrates without a known non-infectious cause The occurrence of one or more positive blood cultures with signs and symptoms of clinical septicaemia Clinical deterioration and rigors in association with unstable haemodynamic measurements and/or coagulopathy consistent with sepsis in the absence of positive blood cultures Fever associated with loin pain or tenderness, and accompanied by significant bacteriuria (>, 10’ cfu ml-’ of voided or catheter urine, or 2 lo3 cfu ml-’ in a sample obtained by bladder puncture), pyuria, and sometimes haematuria The presence of significant bacteriuria in a systemically unwell, febrile (> 38’C) patient in whom structural or functional abnormalities of the urinary tract, known to predispose to or permit persistence of infection, are present Clinical symptoms and signs of infection subsided with complete resolution of active infection during the treatment period Clinical signs and symptoms subsided but without complete resolution of active infection during the treatment period Recurrence of clinical signs and symptoms of the initial infection during the follow-up period, after initial cure or improvement Introduction of a new antibiotic agent less than 72 h after initiation of treatment with the study drug regimen because of a rapid deterioration in clinical condition, e.g. onset of unstable vital signs, development of shock etc., or no apparent response to at least 72 h of treatment with the study drug regimen resulting in the introduction of a new antibiotic agent or death due to infection
Antibiotic
Inclusion
trials
in serious
and exclusion
infections
17
criteria
The inclusion and exclusion criteria used in this study are given in Table III. It is of general importance that such criteria do not exclude too many patients from entry. Exclusion criteria which are too strict will markedly limit the external validity of the results obtained. Thus, exclusion criteria such as advanced age, organ failure or specific concomitant medication (other than with antibiotics) should not be used. It is also doubtful whether criteria such as ‘likelihood of death within 24 hours’ should be allowed; such patients are given antibiotics if they suffer from an infection and represent an important challenge for any new broad-spectrum parenteral antibiotic. An important exclusion criterion which should be used in all studies of this type is that patients who have already been included in the study once should not be allowed to enter it a second time. In studies of neutropenic patients with fever, multiple entry into clinical trials are common and may make the evaluation difficult. For example, in patients entered more than once it is not possible to know whether previous entries have resulted in selection of bacterial resistance in the normal flora, and it may be difficult to assess adverse reactions. In this context it is important that the patients included in the trial are representative of a larger group of patients with the same or similar conditions. To achieve that goal one can use a system where consecutive patients are entered; that is, all patients who fulfil the inclusion but not the exclusion criteria are included in the trial. However, normally that is not possible to achieve in a clinical setting where a large number of hospital staff members are involved in the care of the patients. Another alternative is then to use a reject log, meaning that all patients who fulfil the
Table
III.
Inclusion
and exclusion
criteria
in a study of ceftazidime
Inclusion criteria 1 Hospitalized patients 3 18 years of age 2 Male or female sex 3 Suspected nosocomial bacterial pneumonia, septicaemia infections requiring parenteral antibiotic treatment 4 Onset of infection more than 48 h after randomization 5 Informed consent
vs. imipenem/cilastatin
or severe urinary
tract
Exclusion criteria 1 Hypersensitivity to cephalosporins or carbapenems 2 Immediate type hypersensitivity reactions to penicillins 3 Neutrophil count of less than 1000 mm-j 4 Likelihood of death within 24 h 5 Isolation before therapy of pathogens known or suspected to be intrinsically resistant to either of the study drug regimens 6 Primary diagnosis of a fungal, viral or mycobacterial infection 7 Therapy with ceftazidime or imipenem within the 4 weeks prior to study entry 8 Antibiotic therapy during the 3 days prior to entry into the protocol unless the patient remains symptomatic 9 Previous entry in this trial
18
S. R. Norrby
criteria for entering the study, but who do not do so, are listed with basic clinical information and reasons for not being included in the trial. At the completion of the study the data in the reject log are then compared with the corresponding data for patients entered, and no major differences in terms of prognostic factors should exist. In the ceftazidime vs. imipenem/cilastatin trial, such procedures were unfortunately not included in the protocol. We can therefore expect some difficulties in evaluating the representativeness of the study sample for a larger population of patients with hospital-acquired infections. Diagnostic
procedures
In antibiotic studies, diagnostic procedures aiming at verifying the aetiology of the infections treated are of vital importance. Such procedures are easily defined for patients with urinary tract infections or septicaemia where urine or blood cultures yield the results. However, in lower respiratory tract infections, the bacteriological diagnosis is very difficult, both in terms of requirements put on the investigators and in terms of interpretation of the findings. In this trial, the investigators are required to obtain sputum, transtracheal aspirate, bronchoscopic aspirate and/or lung biopsies. During the course of the study it has become apparent that the procedures for sputum bacteriology vary considerably from centre to centre. In most centres it is, for example, not routine to investigate sputum samples by microscopy. This results in the loss of two vital pieces of information: whether the samples contain squamous epithelial cells (indicating that they represent the pharyngeal and not the lower respiratory tract flora), and whether the organisms later isolated in culture were the predominant ones. Moreover, the preparatory procedures vary between laboratories. Some laboratories wash the sputum samples before further processing, while others have no such routines. Difficulties have also been encountered in the interpretation of results obtained with other samples from the lower respiratory tract. The term ‘transtracheal aspirate’ was found to be misinterpreted as equivalent to any sample taken through a tracheal tube. As a result of the above difficulties, a very large number of isolates have been reported from the lower respiratory tract samples taken from the patients assessed. Without guarantees that the samples do not represent the oropharyngeal flora, it has become difficult to assess the clinical relevance of the isolates. A lesson learned from this is that, before the trial, the laboratories involved should have been contacted to provide detailed information of their procedures for handling sputum and bronchial secretion samples. In some instances it would probably also have been advisable to try to change current routines for the purpose of the study in order to achieve conformity between the centres. Another field where considerable difficulties have been encountered is the interpretation of the radiological findings in patients with lower respiratory
Antibiotic
trials
in serious
infections
19
tract infections. A large proportion of the patients encountered have been treated in intensive care units and have received assisted ventilation. In such patients cardiac decompensation with lung oedema and/or pleural effusion are not uncommon and may make the diagnosis of inflammatory infiltrates of the lungs difficult. The optimal way to solve this problem would be to copy all X-ray films and have them assessed by one radiologist with special experience within the field of pulmonary radiology. Treatment
Clinical trials of parenteral antibiotics have a common weakness in that, to be evaluable for efficacy, the patient must have been treated with the trial antibiotic only, and that no subsequent oral treatment is allowed. The normal clinical procedure is to start treatment with a parenteral antibiotic and then to change to oral treatment when the patient has improved. According to tradition, and also to currently used guidelines for clinical trials of antibiotics, such procedures will render the patients non-evaluable for efficacy. In this trial the traditional approach is followed. However, the situation is less artificial since, of the patients so far enrolled, 73% have been in intensive care units and 48% have had rapidly fatal or ultimately fatal underlying diseases. Thus, most of the patients have been unable to take oral medications during the entire treatment course. The protocol states that patients should receive treatment with either ceftazidime-2 g every 12 h or imipenem/cilastatin-SOO/SOO mg every 8 h. In patients with septicaemia originating from the abdomen, metronidazole could be added when randomized to ceftazidime. Instructions are given for dose reductions in patients with impaired renal function and differ considerably between the two regimens. These differences in treatment between the two groups further emphasize the difficulties involved in making a study like this double-blind. In fact, in seriously ill patients with frequent impairment of renal function, a double-blind comparison of ceftazidime and imipenem/cilastatin seems impossible. Analysis
of efficacy
Serious infections treated with parenteral antibiotics occur only rarely in patients who are otherwise healthy. In most cases such infections result from complications to underlying diseases or surgical procedures. As a consequence of these factors it is often difficult to assess the role of the antibiotic itself in the outcome of the infection; therapeutic measures directed against the underlying condition may very well affect the outcome. In the analysis of efficacy in a trial of a parenteral antibiotic used for the treatment of serious infections, these factors should be taken into account. The results are clearly not suitable for a straightforward significance analysis in which cure rates of different treatment groups are compared.
20
S. R. Norrby
Instead, multi-variate analysis-in which prognostic factors are weighedshould be used. Such procedures have been described by Moore et al. (1984) in studies of aminoglycosides. In the ceftazidime vs. imipenem/cilastatin trial, the primary infection in patients so far entered is pneumonia in 65%, septicaemia in 22%, and severe urinary tract infection in 13%. This will lead to three subgroups which will have to be analysed separately and, especially in the pneumonia group, it is foreseen that there will be a multitude of important prognostic factors-for example, assisted ventilation, underlying cardiac or respiratory diseases, and causative bacterial agent. The following series of events are planned when the study is completed. Firstly, prognostic factors will be identified for a multi-variate analysis. Secondly, subgroups of infections will be decided upon. Thirdly, analyses will be performed, and only at that stage will the patients be grouped according to treatment groups. Thus, those performing the analyses will be blinded as to treatment allocation until all decisions about classifications of outcome and risk factors are decided upon. A similar type of analysis will be performed for safety. It is already clear that the frequencies of adverse effects will be low. It is our feeling that the main reason for this is that the investigators are less inclined to search for, or find, adverse events in severely ill patients than in those with mild or moderately severe infections. Lessons learned
from the ceftazidime
vs. imipenemlcilastatin
study
The most disturbing finding during the course of this study has been that patient recruitment has been considerably slower than expected (Figure 1). As a result, the time for recruiting a sufficient number of patients has had to be extended, and more centres than initially planned have been recruited. In retrospect, a basic mistake in the planning phase was to base the patient enrolment schedule on the estimates done by the investigators. We should instead have performed a pilot study which would have served several we would have achieved a better appreciation of the purposes. Firstly, number of patients available at each centre. Secondly, we could have tested the protocol and the case report form, and would then have been able to avoid most of the mistakes discussed above. Conclusions Multi-clinic or multi-centre trials in patients with hospital-acquired infections are difficult to perform with a high scientific quality. This is especially true in patients with hospital-acquired pneumonia where the main problems are that it is extremely difficult to achieve a correct microbiological diagnosis due to the difficulty of collecting samples representative of the lower respiratory tract, and also because of the multitude of bacterial species found, which makes it almost impossible to
Antibiotic
Figure 1. Actual and projected ceftazidime vs. imipenem/cilastatin
trials
in serious
rate of patient entry for the treatment
21
infections
at various of serious
stages of a clinical trial nosocomial infections.
of
separate commensals from pathogens. Optimally, a study of this kind should include a training period during which one ascertains that the procedures at the microbiological laboratories involved are optimal. During that period a pilot study should be performed in order to achieve adequate estimates of patient entry rate and to test the forms used in the study as well as the design of the protocol. Despite the above difficulties, it is quite clear already at this stage, that the material collected in a study of this kind will offer valuable information about the natural course of hospital-acquired infections, in addition to safety and efficacy data on the trial drugs. However, prerequisites for achieving adequate information are that the analysis be performed in a blinded fashion, although the design of the study is open, and that multi-variate analysis be used. References Kropp, H., Gerckens, L., Sundelof, J. G. & Kah an, F. M. (1985). Antibacterial activity of imipenem: the first thienamycin antibiotic. Reoiew of Znfectious Diseases 7 (Suppl. 3), S389410. Moore, R. D., Smith, C. R., Lipsky, J. J., Mellits, E. D. & Lietman, P. S. (1984). Risk factors for nephrotoxicity in patients treated with aminoglycosides. Ann& of Internal Medicine 100, 352-357. Neu, H. C. & Labthavikul, P. (1982). Antimicrobial activity and beta-lactamase stability of ceftazidime, an aminothiazolyl cephalosporin. Antimicrobial Agents and Chemotherapy 4, 11-18. Working Party of the British Society of Antimicrobial Chemotherapy (1989). The clinical ;;kation of antibacterial drugs. Journal of Antimicrobial Chemotherapy 23 (Suppl. B),
