AUTHORS' REPLY

We thank Navarro et al. for their interest in our work and regret their difficulties in being able to reproduce results similar to those of our group. In fact, their results with a small group of patients and controls are rather surprising, as are some of the comments they make to justify them. In theory at least, it is paradoxical that the PCR false negatives should occur in precisely those patients with a positive blood culture, and we do not think it reasonable to attribute this to a bacterial inoculum below the PCR detection threshold. Zimmerman et al. demonstrated the existence of a close relationship between the size of the inoculum and the positivity of the blood culture and growth rate of the organism (8). The results obtained by Navarro and colleagues would, in fact, point in just the opposite direction to these findings. In one of our works we reported that the detection level of the technique was around 10 fg of DNA, equivalent to the DNA from two cells (5). This amount of inoculum is so small that it is difficult to believe that it would not exist in 0.5 to 1 ml of blood from most patients with positive blood cultures. Moreover, since our first report we have studied a further 52 cases of brucellosis by means of PCR, 36 (69.2%) patients with positive blood cultures and 16 (30.8%) diagnosed according to conventional clinical and serological criteria. Overall, 48 of these 52 patients (92.3%) had a positive PCR, corresponding to 34 (94.4%) of those with a positive blood culture and 14 (87.5%) of those diagnosed clinically and serologically. Although we are unaware of the conditions of extraction and storage for the samples described by Navarro et al., degradation of the DNA sample seems an unlikely reason for the high rate of false PCR negatives since we obtained satisfactory results from samples maintained at 20°C for 6 months prior to processing.

The area where our center is situated and that where Navarro et al. work are both regions where brucellosis is endemic. Virtually 100% of reported cases of brucellosis in Spain are caused by B. melitensis, which is recognized as the most virulent biovar of the Brucella genus. Gotuzzo et al. reported that the rate of clinical infection with B. melitensis in an exposed population was higher than 50% (4). The existence of asymptomatic brucellosis is well-known, but this does not appear to be very common in the case of B. melitensis, and, at the present time, in order to speak strictly of an asymptomatic infection it is necessary to isolate the causative agent or demonstrate some type of specific serological response. Since we included in all our PCRs a sample from a healthy subject as a control of the process of DNA extraction, and to date we have had no false-positive results due to this, we do not think that the existence of an asymptomatic infection is the cause of false positives; nor, therefore, does it contribute significantly to a reduction in the specificity of the technique.

The close phylogenetic relationship between O. anthropi and Brucella spp. is acknowledged, as is the observation of similar products amplified by using the 31-kDa Brucella protein, the heat shock proteins (DnaK, DnaJ, HtrA and GroEL), and 16S rRNA primers (1). Nevertheless, we agree with Romero et al. that "it is unlikely that O. anthropi would cause a false-positive result in a test for Brucella spp. with the PCR assay ... since O. anthropi has rarely been found to be pathogenic" (6). Given the low virulence of this microorganism, it would be surprising if it infected healthy persons and produced asymptomatic bacteremia leading to a false-positive PCR result. With respect to this, it is interesting that only a very few cases of infection by O. anthropi have been reported to date and that almost all occurred in severely immunosuppressed patients or those with debilitating illnesses; most infections were nosocomial or in patients with catheters or other foreign bodies (2, 3). From a clinical point of view, this is a situation diametrically opposite to infection with Brucella spp., which is always a community infection affecting generally immunocompetent subjects.

Finally, we agree with Navarro et al. that in-house PCR results can sometimes be difficult to reproduce. PCR, although a theoretically simple concept, requires dedicated and experienced personnel. The adaptation and acceptance of this technology in the forum of clinical diagnosis has been slow, due mainly to a number of technical obstacles (7). We are sure that familiarization with the technique will eventually lead to these authors producing results similar to those communicated by our group.