With the exception of retinitis, end-organ manifestations of CMV disease in HIV-infected individuals have unspecific clinical characteristics, and histopathological confirmation is required for definite diagnosis. Because biopsy of affected organs may be difficult to obtain, there is a need for microbiological diagnostic methods that can be used on readily available specimens. A variety of tests to detect CMV nucleic acids have been evaluated in different patient populations. To our knowledge, no previous tests have been evaluated in relation to autopsy findings, an important shortcoming, as histological verification of CMV disease is often first obtained at autopsy. The main strength of this study is that it explores the utility of quantitative PCR in the diagnosis of CMV disease in HIV-infected individuals with available autopsy results. Also, it includes one of the highest reported number of cases of CMV disease in HIV-infected patients evaluated with CMV PCR.
Several qualitative and quantitative PCR methods for CMV in plasma in various patient populations have demonstrated sensitivity in the range 35–93% [13, 18–20, 23, 26, 27]. In our study, maximum sensitivity of 47% for the total number of cases, using a cut-off at the limit of detection, was in the lower end of this range. In a small previous study on HIV-infected individuals using COBAS AMPLICOR CMV Monitor, sensitivity was 92% . The relatively low sensitivity in our study could be due to samples in many cases having been taken as part of routine monitoring and therefore too long, a median of 69 days, before development of CMV end-organ disease. However, CMV viraemia is known to often precede development of CMV disease by several months [2, 13, 17]. Also, in contrast to previous studies, most of our cases of CMV disease were first diagnosed at autopsy, and it is possible that these post-mortem manifestations are associated with a shorter time with viraemia and/or lower level of viraemia. Interestingly, our results show that sensitivity was significantly higher for patients with CMV disease in more than one organ compared to patients with single organ disease.
Previous studies of CMV disease diagnosed before death have yielded specificities in the range 47–100% [13, 18–20, 23, 26, 27]. Our study shows specificity in the upper range (89%), with only 8 of 72 (11%) cases having detectable CMV viraemia but no diagnosis of CMV disease (false positive tests). Studies that do not include autopsy results may underestimate the specificity, as CMV end-organ disease may be a difficult clinical diagnose to make.
Our data were also analysed with regard to detection of CMV viraemia in one versus two plasma samples taken at different times. We found that cases with detectable CMV viraemia in both samples all developed CMV disease. Thus, by requiring confirmation in a second test if the first was positive, a specificity of 100% was attained, but at the cost of lower sensitivity.
From a clinician's point of view, the predictive value of a positive and negative test result has greater practical implications than do sensitivity and specificity. The predictive value of any test is generally dependent on the prevalence of the disease in question in the study population, NPV decreasing and PPV increasing with increasing prevalence. Previous studies have reported NPV in the range 80–95% [20, 23, 26, 27], which is somewhat higher than in our study. The prevalence of CMV disease in our study was comparatively high – 42% (53 of 125 patients) – thus contributing to a low NPV. For the same reason, the PPV demonstrated in our study was among the highest reported. In our patient population at high risk of CMV disease we attained a PPV of 100% with a cut off-value of 10 000 CMV DNA copies/mL or a requirement for viraemia in two consecutive tests. Studies that do not include autopsy results may underestimate the proportion of cases with CMV disease, resulting in an overestimation of the NPV and an underestimation of the PPV.
The study covers a period both before and after HAART became standard of care. Even in patients on HAART, antiretroviral treatment was suboptimal for many, probably due to HIV resistance having developed after previous ineffective therapy with single drugs, or combination of two antiretrovirals. This may affect the external validity of our results in the diagnosis of CMV disease in HIV-infected patients who have access to more effective HAART today and, as a result, have lower risk of CMV disease. However, results from our study are likely to be relevant for patients who interrupt treatment, who fail to respond to antiretroviral therapy, or who are diagnosed too late for effective HIV therapy to be initiated.
This was a retrospective study in which the time intervals varied considerably between the last plasma sample and CMV diagnosis (alive or at autopsy), or between the last plasma sample and death in cases without CMV disease, as samples were taken as part of routine monitoring of the patients rather than on clinical suspicion. It is likely that the sensitivity and PPV of CMV PCR will be higher in a setting where plasma is analysed at a time of clinical suspicion of CMV end-organ disease, rather than as part of routine monitoring of patients.
The availability of more than one plasma sample from individual patients at relevant time points was limited, but a high positive predictive value for CMV disease was found for patients with CMV viraemia in two consecutive tests with a median time interval of between two and three months. However, our data do not allow us to conclude when plasma CMV quantitative PCR should be performed, or how soon a patient should be retested after a positive test result.
In solid organ transplant patients and allogenic bone marrow transplantation, routine monitoring of virological markers for cytomegalovirus and administration of pre-emptive therapy has been shown to significantly reduce the risk of CMV disease [30, 31], and post-transplant monitoring of CMV viraemia by PCR or other methods is used in many centres. Among HIV-infected individuals the value of pre-emptive therapy is less well documented, but monitoring of high risk patients for this purpose is recommended in French guidelines .
Many studies have shown that the risk of CMV disease sharply increases when CD4 cell counts fall below 100/mm3 [2, 3, 32]. We suggest that HIV infected patients with low CD4 cell counts who have clinical features suggestive of CMV disease, should have plasma tested for CMV by quantitative PCR as part of the diagnostic procedure. In patients with repeated viraemia or viraemia above 10 000 CMV DNA copies/mL, pre-emptive therapy should be considered after a careful clinical examination for signs and symptoms of CMV end-organ disease. However, our findings need to be validated in prospective studies on patients living with HIV today.