Limitations of Cytomegalovirus Testing

Douglas Jabs and coworkers performed a prospective study of 122 patients with cytomegalovirus (CMV) retinitis (5). They collected blood and urine CMV isolates and performed sensitivity testing of the isolates.

The interpretation of laboratory sensitivity testing of CMV strains is, however, problematic. CMV is known to be difficult to grow in cell culture, and indeed, only 60% of the cultures were positive in the study of Jabs and coworkers. A strong selection process is needed for growth in cell culture, and during this process features of the virus present in the blood or urine may be lost. Furthermore, the virus in blood or urine may be different from the virus causing the disease, in this case the virus in the retina. It is therefore questionable how representative a virus isolated in cell culture is for the disease causing virus in the retina.

The laboratory procedure for CMV sensitivity testing is less than optimal mainly because CMV shows large variation in growth in cell culture. The large variation between multiple sensitivity testing of the same isolate tested either in parallel or in sequence is rarely shown in scientific articles.

It is therefore not surprising that sensitivity testing results may vary between different laboratories where procedures may differ. The average 50% inhibitory concentration (IC₅₀) of CMV isolates obtained from untreated patients may vary between different laboratories by a factor of 2 (2).

The clinical relevance of CMV sensitivity testing is subject to much debate due to factors mentioned above and due to the limitation that sensitivity testing can be performed only for patients with positive blood or urine cultures. A positive and ganciclovir-sensitive blood or urine culture during adequate CMV treatment is itself associated with poor prognosis measured as subsequent contralateral eye disease (odds ratio, 3.82) (3). Jabs and coworkers have showed an association between resistance against ganciclovir and subsequent contralateral eye disease, while other studies have failed to detect an association with progression of CMV retinitis (for a review, see reference 2).

Of crucial importance for all laboratories is to carefully define their cutoff levels for CMV sensitivity testing. This is not a trivial problem, and the cutoff levels should be carefully scientifically motivated and the above-mentioned uncertainties should be accounted for. Ganciclovir has attracted most attention, and Drew and coworkers have proposed two cutoff levels (reviewed in reference 2). They considered IC₅₀s above 12 μM as indicating resistance and values below 6 μM as indicating sensitivity, while intermediate values were considered to indicate “intermediate susceptibility.” Jabs and coworkers used 6 μM as cutoff for resistance (3). The rationale for the selection of this level is, however, unclear. It is obvious, due to the uncertainties, that each laboratory needs to carefully assess CMV isolates obtained from untreated patients in order to ensure that the cutoff limits are set in an unbiased way. This is especially important for direct comparisons between different treatments.

Jabs and coworkers performed a statistical comparison between the relative risk for foscarnet resistance during foscarnet treatment and the relative risk for ganciclovir resistance during ganciclovir treatment (Table 3 in reference 5). This type of comparison is very much dependent on factors listed above but also and in particular on the patient population and the selected cutoff levels. The study performed by Jabs et al. was not randomized nor blinded. The treatment for each patient was assigned by their physician according to the clinical presentation of the disease. The influence of this design on the foscarnet-ganciclovir comparison is unknown, but the introduction of a bias probably cannot be excluded. The cutoff levels were unfortunately not justified, in contrast to previous work (1, 4).

Previous work by the same authors gives us the possibility to speculate (4). The authors had earlier reported IC₅₀s for CMV isolates obtained from untreated patients (4). It is important to note that the cutoff used in the recent study for ganciclovir resistance (6 μM) is 3.5 standard deviations (SD) higher than the average IC₅₀ (1.31 ± 1.34 μM) for blood isolates (2.6 SD higher than that [1.79 ± 1.65 μM] for urine isolates) while the cutoff used for foscarnet resistance (400 μM) is only 1.6 SD higher (IC₅₀, 209 ± 117 μM) (1.4 SD higher for urine isolates [IC₅₀, 214 ± 133 μM]) (Table 4 in reference 5). These data thus indicate that the cutoff used in the recent study for ganciclovir resistance may be further away from the average IC₅₀ for the virus isolates for untreated patients than the cutoff for foscarnet. If true, this would bias any comparison between the relative risks for ganciclovir and foscarnet resistance. It is therefore very important that Jabs and coworkers clearly show the scientific background for the selection of the cutoff levels and the justification for the comparison of the relative risks for ganciclovir and foscarnet resistance. Otherwise, their statement “These probabilities (relative risk for foscarnet resistance) appeared to be no less than that of developing a ganciclovir-resistant isolate, and our data suggested that they may be greater” may be misleading for clinicians treating AIDS patients. An unbiased comparison would probably include foscarnet cutoff levels in the range of 560 to 620 μM. The concentrations of foscarnet in plasma in patients during treatment is also in this range or above (6).

Given the uncertainties listed above, we strongly suggest that the word “resistance” be used with caution. We should reserve the word for clinical CMV isolates with an IC₅₀ indicating a clear relationship between laboratory resistance and clinical unresponsiveness. The cutoff level for this more stringent resistance criterion remains to be determined, but the level will be substantially higher than the levels used by Jabs et al. We should further refer to isolates for which the IC₅₀ is above the normal range (i.e., above a cutoff defined as the average for samples from untreated patients plus 2 to 3 SD) but below the resistance cutoff as “decreased sensitive” isolates. Otherwise, the laboratory conclusions from sensitivity testing of CMV isolates may be misleading for the clinicians treating AIDS patients.