This is a critical abstract of an economic evaluation that meets the criteria for inclusion on NHS EED. Each abstract contains a brief summary of the methods, the results and conclusions followed by a detailed critical assessment on the reliability of the study and the conclusions drawn.

Oral ciprofloxacin was compared with intravenous ceftriaxone in the empiric domiciliary treatment of febrile episodes in low-risk neutropenic and nonneutropenic patients with cancer.

Treatment.

Cost-effectiveness analysis.

The study population comprised patients with cancer who were defined as being low-risk neutropenic or nonneutropenic. Study inclusion criteria were: fever (at least 38 degrees C) on two measurements, lasting no more than 24 hour and judged to be of infectious origin; and residency within 20 miles of the participating centres. Study exclusion criteria were: severe chemotherapy-induced neutropenia; severe comorbidity requiring hospitalisation; acute leukaemia and bone marrow transplant within the last two years; antibiotic treatment within the previous 5 days; use of haematopoietic growth factors; positive history of intolerance to quinolone derivatives or beta lactams; renal failure, under 18 years of age; and pregnant or nursing a baby.

The setting was primary and secondary care. The economic study was carried out in Italy.

The effectiveness and resource data were collected between June 1991 and May 1996. The price year was 1996.

The effectiveness data were derived from a single study.

Cost data were applied retrospectively to the same patient sample used to generate effectiveness evidence. The authors did not report whether all the cost data used were derived from the same patient sample used to generate the effectiveness evidence.

Power calculations were used to estimate the sample size required to test whether ciprofloxacin was at least as effective as ceftriaxone. The authors reported that the sample size was determined to test the hypothesis of non-equivalent efficacy. The authors defined the equivalence region as the lower limit of the 95% confidence interval. If the lower confidence limit was less than or equal to -15%, then the hypothesis of non-equivalence could be accepted. The power calculation was based on the assumption that the success rate of ceftriaxone treatment in febrile cancer patients is about 80% and to test the assumption that the success rate of oral ciprofloxacin therapy would not be more than 15% lower than that of ceftriaxone therapy, at the 5% significance level and 80% power. All consecutive adult outpatients with cancer referred to the participating centres were screened for the study. The study sample was appropriate for the clinical study question and comprised 173 febrile patients enrolled in the study from June 1991 to May 1993, accounting for 183 febrile episodes. Of these, 93 febrile episodes were randomised to receive ceftriaxone, and 90 febrile episodes were randomised to receive oral ciprofloxacin. Patients were not eligible to re-enter the study unless they were fully recovered from their last febrile episode.

The study was a multi-centre (number of centres not reported), prospective, randomised controlled trial. Patients who met the study inclusion criteria were randomised to oral ciprofloxacin or iv ceftriaxone using a computer-generated randomisation list of blocks permutated by participating centre. The duration of follow-up and number of patients lost to follow-up were not reported. The patients and health care providers were not masked to treatment allocation. The authors did not report whether the investigators who conducted the follow-up assessments or the data analysts were masked to method of treatment.

The analysis was performed on an intention to treat basis. The primary outcome used in the economic analysis was the clinical success rate. This was defined as when fever and the clinical signs of infection disappeared and eradication of the infecting micro-organism was obtained without changing the allocated antibacterial therapy, and the response lasted for at least 4 days after the discontinuation of therapy. The two patient groups were shown to be comparable at analysis in relation to age, gender, underlying disease and number of neutropenic patients. No adjustments for confounding factors were reported.

The oral ciprofloxacin group had a success rate of 82% (76 clinically successful outcomes in the 93 febrile episodes).

The iv ceftriaxone group had a success rate of 75% (68 clinically successful outcomes in the 90 febrile episodes).

The difference in success rate was 5% and the lower 95% confidence limit was -6%. This was within the equivalence range specified by the authors (-15%), indicating that the hypothesis of non-equivalence could be rejected.

Treatment failures occurred in 2 patients randomised to oral ciprofloxacin and in 4 patients treated with iv ceftriaxone.

Of the febrile episodes of the patients treated with oral ciprofloxacin 0 of 93 (0%) died from infection, compared to 2 of 90 (2%) receiving iv ceftriaxone.

The authors concluded that domiciliary treatment of febrile episodes in low-risk neutropenic and nonneutropenic cancer patients is feasible and effective and that there were no statistically significant differences between oral ciprofloxacin or iv ceftriaxone.

No summary measure of health benefit was defined for the economic analysis. The outcomes were reported in a disaggregated way. The authors indicated that there were no differences between the alternative therapies. However, the study did not recruit the sample size required (99 patients per treatment arm) to demonstrate that the 2 groups were statistically equivalent. This is a necessary condition for cost minimisation analysis, and therefore, in this instance, the study should be classified as a cost-consequences analysis.

Resource use and costs were not reported separately. Hospital related direct costs were reported. The authors reported that the direct costs included the acquisition and administration costs of drugs, including the cost of both materials used and the time spent by nursing staff in administering therapy. The unit price for drugs and materials was obtained from a published source. The authors reported that the mean cost per hour per worker during 1996 was L33,600 for nursing staff, and L60,000 per hour for the medical staff. The authors estimated the cost of failure for two alternative analyses. In the first, more costly analysis, it was assumed that all the patients who had a treatment failure needed hospitalisation and the costs were estimated from Italian Disease Related Groups data. In the second, less costly analysis, it was assumed that the costs of failure were the sum of the costs of patients actually admitted to hospital after treatment failure (9 for ciprofloxacin and 5 for ceftriaxone) plus the cost of antibiotic rescue treatment for those patients not hospitalised after treatment failure. The cost recorded for each failure without hospitalisation was that of 8 days treatment with iv imipenem (1g three times daily). The cost of side effects was estimated in a similar manner. The price year was 1996. Discounting was not undertaken because of the short time frame of the study.

No statistical analysis of costs was conducted.

No indirect costs were included in the analysis.

The study reported costs in US dollars ($). The authors reported an exchange rate of 1$ = L1,600.

The authors reported a simple one-way sensitivity analysis based on low and high cost estimates for each intervention.

The reader is referred to the effectiveness results reported previously.

The authors reported that the total treatment cost of oral ciprofloxacin per febrile patient was $715 under the most costly hypothesis and $480 under the least costly hypothesis for outpatient treatment.

In comparison the total treatment cost of iv ceftriaxone was $1,062 under the most costly hypothesis, and $639 under the least costly hypothesis for outpatient treatment.

No synthesis of costs and benefits was carried out and no incremental analysis was reported.

The authors concluded that domiciliary (out-patient) treatment of febrile episodes in low-risk neutropenic and nonneutropenic cancer patients is feasible and effective and that either oral ciprofloxacin or ceftriaxone can be used. However, the authors stated that oral ciprofloxacin was less expensive and thus the more preferable treatment option.

The authors did not present a justification for the choice of comparator, although they noted that both the intervention and comparator could be used as outpatient treatment strategies for febrile episodes in low risk cancer patients. You, as a user of this database, should decide if these are widely used health technologies in your own setting.

The analysis was based on a randomised controlled trial design that was appropriate for the study question. The study sample was representative of the study population, although the authors cautioned that the study only included patients defined as low-risk and the results of this study are therefore relevant only to low-risk patients. The patient groups were shown to be comparable in the baseline characteristics specified by the authors. There were some differences in the distribution of gram negative and gram positive infection that may have influenced the success rates in each treatment group. However, the authors did not report whether these were statistically significant differences or whether they occurred by chance. The authors conducted, but did not clearly report, statistical analyses of the data, and used an intention to treat approach. The authors reported power calculations to estimate the study sample size. The authors appear to define non-equivalence for the power calculation as ciprofloxacin having a success rate (eradication of fever, clinical signs and infection) that was more than 15% lower than ceftriaxone. This implies that, even with a difference of up to 15% in the defined success rate, the 2 drugs would be accepted as equivalent in efficacy. The authors did not report whether this is a clinically acceptable range within which to define equivalence. If a clinically important difference between success rates is defined as being less than 15%, then the power calculations may have underestimated the sample size required to test the hypothesis. If it is more than 15% then the power calculations may have overestimated the sample size needed. The authors reported that 99 patients were required in each treatment arm (198 overall) to test the hypothesis of non-equivalence. However, only 173 patients with a total of 183 febrile episodes were recruited. This means that the absence of statistically significant differences reported by the authors may be due to a lack of power rather than equivalence in outcomes. The patients and health care providers were not masked to treatment allocation. The authors did not report whether data analysts or the investigators conducting follow-up assessments were masked to treatment allocation. The outcomes and/or subsequent therapy may have been influenced by knowledge of treatment allocation, which could have biased the results. From the data presented it is not possible to assess whether the analysis adequately controlled for confounding factors and biases.

The authors did not derive a summary measure of health benefit nor did they prove statistical equivalence in clinical outcome. The analysis was therefore categorised as a cost-consequences study.

All categories of cost relevant to the perspective adopted appear to have been included in the analysis. Although some costs, for example the cost of concurrent medications, were omitted from the analysis, their omission is unlikely to have affected the authors' conclusions. Costs and quantities were not reported separately. The authors did not report clearly how total costs had been generated from resource use and unit cost data. A limited sensitivity analysis was conducted based on changing the assumption regarding how treatment failures were managed. The authors did not conduct a sensitivity analysis of unit costs or resource use. This may limit the interpretation and generalisability of the study findings. The price year was reported and appropriate currency conversions were undertaken. Discounting was not undertaken, which was appropriate given the short time frame of the study.

The authors made appropriate comparisons of their findings with those from other studies but, due to the omission of an appropriate sensitivity analysis, did not fully address the issue of generalisability to other settings. The authors do not appear to have presented their results selectively. The authors acknowledged that the study did not enrol patients who had marked prolonged granulocytopenia and selected a low-risk population for therapy, including only nonneutropenic and moderately neutropenic patients, and this was reflected in the authors' conclusions. The authors reported a number of further limitations to the study namely the success rate of ciprofloxacin might be dependent on the causative bacteria (gram negative or gram positive) underlying the febrile episode. The authors went on to suggest that it may be necessary to use an agent with better coverage against gram positive bacteraemia at least in patients not responding to the initial empirical treatment.

The authors reported that domiciliary treatments cost, at the most, a third of the cost of inpatient treatments. Furthermore the authors stated that both domiciliary treatment regimes are clinically acceptable from aspects of efficacy and tolerability, providing that patients undergo initial evaluations and are followed-up carefully. However, the authors do state that, because of a lower cost and easier administration, that oral ciprofloxacin represents the best and preferred treatment option.

Supported by a grant from Bayer Italia.