TO THE EDITOR:

We read with interest the article by Sarda et al. (1). However, we would like to point out some major study design and data analysis flaws that undermine the quality of the work and the authors’ conclusion about the value of ^99mTc-ciprofloxacin in infection imaging.

The first flaw is the small sample size. With only 13 rabbits, 6 in the infected group (G1) and 7 in the uninfected group (G2), the study is heavily underpowered to demonstrate with a reasonable degree of confidence a statistically significant difference in imaging results between the 2 groups. Thus, although Figures 3 and 5 (these should have included error bars to indicate SDs) clearly demonstrate higher uptake of ^99mTc-ciprofloxacin in infected knees than in uninfected knees, the difference was not statistically significant. Sarda et al. (1) acknowledge the small size in their study, but instead of accepting that this is a serious limitation and hence that the results should be interpreted with caution, they justify themselves by giving feeble excuses. Although we sympathize with their wish to use as few animals as possible for ethical reasons, they should have realized that such an approach may well compromise the scientific quality and validity of their study. Another justification they give is the good reproducibility of their animal model. However, as the title of their article states, the model was developed to study prosthetic joint infection due to methicillin-resistant Staphylococcus aureus. The model has not been validated for infection imaging studies with radionuclides.

The second flaw is a failure to compare like with like. The 2 arms of the study should have included equal numbers of rabbits matched for age, sex, and weight. The rabbits varied in age (and weight) between 74 and 120 d, and one was 6 mo old. This is important because ciprofloxacin is taken up by growing cartilage in young mammals. Hence, only mature animals should have been studied. A similar volume of inoculum should have been injected. Instead, 0.5 mL of 10⁷ colony-forming units of methicillin-sensitive S. aureus was injected into G1 rabbits, and 1 mL of saline was injected into G2 rabbits. In addition, the suspension medium used to make the bacterial inoculum should have been used in the control arm; even better, a suspension of the dead bacteria could have been used as a control. Microbiologic evaluation should have been the same for both groups of rabbits. Quantitative bacterial counts of tissue samples were performed on 2 G1 rabbits but not on G2 rabbits. In G2 animals, only the knee exudates were cultured and found to be sterile. It is well known, and mentioned by the authors in their article, that culture of exudates lacks sensitivity for the diagnosis of prosthetic joint infection. Tissue samples generally yield higher culture positivity rates. Infection is an important and well-recognized complication of prosthetic joint surgery, the most common mechanism being contamination of the joint at the time of the operation, with rates of infection varying according to the type of joint (e.g., knee higher than hip). Infected and uninfected joints and rabbits should have been followed up identically with regard to the biodistribution, autoradiographic, and imaging studies.

The third flaw is use of ^99mTc-ciprofloxacin for animal studies. ^99mTc-ciprofloxacin has been standardized for human use only. For small-animal experiments, the preparation may have to be adjusted to give optimal results—for example, reducing the dose of ciprofloxacin from 2 mg to that appropriate for an adult rabbit weighing 2.9–3.5 kg.

Animal studies, when well designed and executed, may provide useful information, but care must be exercised when extrapolating results from these experiments directly to humans. The reasons given above are more credible explanations than those discussed by Sarda et al. (1) to account for the difference in results for ^99mTc-ciprofloxacin imaging between their studies on animals and others’ clinical studies on patients. In contrast with the study of Sarda et al. involving only 11 rabbits, we have studied the efficacy of ^99mTc-ciprofloxacin in a much larger sample size, 879 patients (2). The study, under the auspices of the International Atomic Energy Agency (IAEA), was well designed and incorporated internationally recognized criteria for defining infection and agreed criteria for the interpretation of ^99mTc-ciprofloxacin images. The overall sensitivity and specificity for detecting infection were 85.5% and 81.6%, respectively, but they varied according to the type of infection imaged. For example, the sensitivity and specificity in prosthetic joint infections were 96% and 91%, respectively (n = 194 patients). The IAEA study included 237 uninfected (control) patients, and hence the comment by Sarda et al. that we did not have control groups in our study is erroneous and should be retracted.

The statement of Sarda et al. (1) that an animal model of prosthetic joint infection closely mimics acute postoperative infection in humans sadly demonstrates a profound lack of knowledge and insight into the pathogenesis and pathobiology of this important disease in humans and hence an inability to evaluate critically the difference between the animal model and the human condition. The animal model of prosthetic joint infection is by its very nature artificial, designed to produce quick results and use the fewest animals possible (time and cost are often overriding factors, as well as or in addition to any ethical consideration)—for example, a study of the efficacy of different antibiotics as a prelude to clinical trials in humans. Hence, a large inoculum of organisms is used (10⁷ colony-forming units in 0.5 mL by Sarda et al., i.e., a visible suspension of bacteria) to ensure an almost 100% infection rate (compared with up to a 5% infection rate after prosthetic joint surgery in humans) and a rapid onset of infection (clinically manifesting within a day in the model described by Belmatoug et al. (3) and used by Sarda et al.), which overwhelms the animal’s defense system and sometimes results in death. By contrast, in human infections, the number of organisms contaminating the joint at the time of the operation (this period, not just the period after wound closure, is recognized as of greatest risk and was used in the animal model by Sarda et al. to inoculate the organism into the joint) is much smaller (hence, the operation is classified as a clean procedure). Thus, the time taken for the human infection to manifest clinically (i.e., the incubation period) is variable: usually a few weeks to a few months but sometimes longer than a year, depending on the type and virulence of the organisms and the host factors that determine susceptibility to and recovery from infection. The animal model is therefore atypical and does not represent well the spectrum and severity of infection occurring in real life and in real patients.

In addition, we would like to comment on the discussion of Sarda et al. (1). First, although quinolones might be taken up and concentrated inside neutrophils and macrophages, this effect must be slight with ^99mTc-ciprofloxacin in vivo, as evidenced by the lack of uptake in bone and bone marrow (sites rich in these cells) in both human and animal studies, including that of Sarda et al. In addition, quinolones are not retained within these and other cells as long as are, for example, the newer macrolide antibiotics such as azithromycin. As the blood concentration drops, they leach out of the cells and tissues into the tissue fluid and then into the blood, to be excreted mostly in the urine. In healthy volunteers, the serum half-life of intravenous ciprofloxacin is 3.5–4.8 h, little remains at 12 h, and there is no accumulation. By contrast, ^99mTc-ciprofloxacin remains bound to and is retained at sites of bacterial infection, giving a high target-to-background ratio, which is the basis of the bacterial specific imaging with this agent.

Second, the inhibitory effect of ciprofloxacin on growing cartilage has been well documented in beagle puppies given large doses of the antibiotic. This effect formed the basis of the contraindication for ciprofloxacin in children and pregnant women. However, ciprofloxacin has been and continues to be extensively and effectively used worldwide as an oral treatment for the troublesome and debilitating Pseudomonas aeruginosa infection in thousands of children with cystic fibrosis, for which no alternative oral treatment is available. The antibiotic has been well tolerated, and the adverse effect on cartilage growth seen in puppies has not been a problem in children. This is a classic example of the fact that for a variety of reasons, which include species difference, experimental data on animals may not always be directly applicable to humans. One must therefore be cautious about such extrapolation.

Our third comment regards inhibition of mammalian DNA gyrase by ciprofloxacin. This inhibition is unlikely to be significant with ^99mTc-ciprofloxacin, which contains only a tracer dose of ciprofloxacin (2 mg, which is 1/200th of a single therapeutic intravenous dose of ciprofloxacin). Moreover, compared with bacterial DNA gyrase, the affinity of ciprofloxacin is 100–1,000 times lower for mammalian topoisomerase II, and the ready reversibility of the binding is compatible with ciprofloxacin pharmacokinetics in humans. Also important, the concentration of ciprofloxacin used in the studies cited by Sarda et al. (1) and similar published studies has varied from around to greatly above the therapeutic range. For example, a ciprofloxacin concentration of 5–50 mg/L was used in the study by Bryant and Mazza (the reference for which has not been given accurately by Sarda et al. and is provided here (4)), and a trovafloxacin concentration of 0.5–25 mg/L was used in the study by Pascual et al. (5), compared with a peak serum concentration of 2–4 mg/L 1 h after intravenous administration of the standard dose of 400 mg of ciprofloxacin. The therapeutic range is many times higher than is obtained by the administration of ^99mTc-ciprofloxacin (assuming linear kinetics, the serum concentration of ciprofloxacin 1 h after intravenous administration of ^99mTc-ciprofloxacin, which contains 2 mg of ciprofloxacin, is expected to be about 0.01–0.02 mg/L). Hence, the relevance of these studies to ^99mTc-ciprofloxacin imaging is doubtful.

In conclusion, although interesting, the study of Sarda et al. (1) is fundamentally flawed by poor design. Regrettably, Sarda et al., instead of acknowledging the limitations of their study, try to justify their approach and support their results by uncritically citing the results of other investigators. This attempt is also reflected by their reference to the work of Welling et al., about which we have had many exchanges in letters to the editors of both The Journal of Nuclear Medicine and the European Journal of Nuclear Medicine. These authors have stated that their research has never been centered on ^99mTc-ciprofloxacin but, rather, that they have used it as a control agent for their experiments with ^99mTc-labeled cationic peptides for infection detection in animal experiments. There is, no doubt, much to be gleaned from animal experiments, but when poorly designed and evaluated they may yield misleading results and be misinterpreted, as well as causing a great deal of unnecessary suffering to and waste of animal life.