Plasma cytomegalovirus DNA, pp65 antigenaemia and a low CD4 cell count remain risk factors for cytomegalovirus disease in patients receiving highly active antiretroviral therapy

Objective: To study the natural history and the current risk factors for cytomegalovirus (CMV) disease in the context of highly active antiretroviral therapy (HAART).

Setting: Prospective multicentre cohort in 15 university hospitals in France.

Cited Here...: A group of 198 patients with CD4 cell count < 100 × 10⁶ cells/l (or < 200 × 10⁶ cells/l under HAART for at least 2 months), no previous CMV disease and CMV-positive serology were followed every 4 months clinically and for virological testing including HIV RNA and CMV blood markers (culture, pp65 antigenaemia, plasma CMV DNA and CMV late mRNA by the polymerase chain reaction).

Cited Here...: At inclusion, median CD4 was 77 × 10⁶ cells/l (0-308) and 85% of the patients received protease inhibitors. The percentage of patients receiving HAART reached 99% at 12 months. After a follow-up of 23.6 months, the incidence of CMV disease was 3.2/100 patient-years [95% confidence interval (CI) 1.3-5.0]. In univariate Cox models, all the CMV markers, a CD4 cell count remaining < 75 × 10⁶ cells/l and an HIV viral load > 100 000 copies/ml were predictive for CMV disease. The hazard ratios for CMV disease were 11 for blood culture; 14 and 70 for pp65 antigenaemia of ≥ 1 and ≥ 100 nuclei/200 000 cells, respectively; 35 for plasma CMV DNA; 6 for CMV mRNA; 29 for CD4 < 75 × 10⁶ cells/l; and 12 for HIV RNA > 100 000 copies/ml. In a stepwise multivariate analysis, only three covariates were independently associated with the occurrence of a disease: plasma CMV DNA, pp65 antigenaemia ≥ 100 nuclei/200 000 cells and a CD4 count < 75 × 10⁶ cells/l.

Conclusion: CMV blood markers and CD4 count < 75 × 10⁶ cells/l remain risk factors for CMV disease in patients receiving HAART. Analysis of plasma CMV DNA by the polymerase chain reaction is a reproducible and standardized tool that could be used as a decision marker for initiating CMV pre-emptive therapy.

Introduction

The impact of highly active antiretroviral therapy (HAART) on the incidence of opportunistic infections in patients with HIV infection has been extremely impressive. It is probably related to the increase in CD4 T cell count and to the restoration of immune responses that follow prolonged control of the plasma HIV load ^[1-3]. A dramatic reduction in the prevalence of cytomegalovirus (CMV) disease, of approximately 80%, has been observed in France between the second semesters of 1996 and 1997 ^[4,5]. HAART can restore a specific CD4 T cell reactivity against CMV antigens even in severely immunosuppressed patients ^[1].

However, the risk still persists since not all patients respond maximally to HAART; about 40% of the patients receiving HAART do not have a sustained reduction in HIV viral load. Moreover, a few CMV infections have been described to occur within the 5 to 7 weeks following initiation of HAART ^[6,7]. In the preHAART period, blood markers of CMV reactivation and a CD4 lymphocyte count < 50 × 10⁶ cells/l were shown to be independent predictors for the occurrence of CMV disease ^[8-10]. New tools for CMV detection have become available recently among which the use of the polymerase chain reaction (PCR) to detect CMV DNA from plasma has good prognostic value for the occurrence of CMV disease ^[11,12]. These new tools for CMV detection need to be evaluated in HIV-infected patients in the context of HAART and compared with the more conventional markers such as blood culture and pp65 antigenaemia.

The aim of this prospective study was to reassess the incidence and evaluate the risk factors of CMV disease in patients treated with HAART.

Methods

The Predivir cohort was a prospective, multicentre French cohort in which recruitment took place between June 1996 and April 1997. It aimed to follow virological markers and clinical CMV disease in patients with advanced HIV infection and to study and compare the prognostic value of these markers on the occurrence of CMV disease. Fifteen centres participated in France and informed consent was obtained from each patient. The protocol was approved by the ethical committee of Cochin Hospital in Paris.

Eligibility criteria were age of over 18 years, CD4 cell count < 100 × 10⁶ cells/l (increased to < 200 × 10⁶ cells/l under HAART begun at least 2 months previously), positive CMV serology, no previous CMV disease or CMV prophylaxis and written consent for participation.

Follow-up

Every 4 months, or 2 months in those with positive blood culture or pp65 antigenaemia, blood samples were taken for virological tests to detect markers of CMV reactivation and measure HIV load and the patients were assessed clinically.

Antigen assay

The assay for pp65 antigen was performed using CINAkit (Argène Biosoft, France) according to the manufacturer's instructions. The number of antigen-positive cells was counted and expressed as the number of positive cells per 200 000 leukocytes.

Cytomegalovirus blood culture

Buffy coat cells were isolated from 7 ml heparinized blood and resuspended in culture medium; 5 × 10⁶ leukocytes were inoculated into fibroblast monolayers seeded in 25 ml flasks. Cell cultures were incubated at 37°C, subpassaged after 2 weeks and maintained for another 2 weeks. They were observed twice weekly for the CMV characteristic cytopathic effect.

Plasma cytomegalovirus DNA

Plasma was obtained by centrifugating blood treated with ethylenediamine tetraacetic acid (EDTA). When plasma was not immediately processed, it was stored at -80°C. PCR for CMV DNA was performed using a commercially available kit (Amplicor PCR, Roche Diagnostic Systems, France) according to the manufacturer's instructions.

Cytomegalovirus late mRNA from blood cells

A reverse transcriptase PCR (RT-PCR) was used to detect late mRNA encoding for major capsid protein, as previously described ^[13].

Plasma HIV viral load

Plasma samples were analysed for HIV-1 RNA using a commercially available RT-PCR assay with a detection limit of 200 copies/ml (Amplicor HIV-1 Monitor, Roche Diagnostic Systems) or a branched DNA assay (Quantiplex bDNA, Chiron Corp., France) with a detection limit of 500 copies/ml. Both assays were performed according to the manufacturer's instruction. Thresholds of 1000 copies/ml and 10 000 copies/ml were used for the analysis of risk factors for the occurrence of CMV organ disease to overcome the problem caused by the different sensitivities of the HIV RNA measurement methods.

Clinical assessment

At each check up, patients were examined for CMV disease and a fundoscopy was performed. The diagnosis of CMV disease was based on the following criteria ^[10]. CMV retinitis was defined by its characteristic fundoscopic appearance and confirmed by photographs and angiography in doubtful cases. CMV disease at other sites was defined by the combination of suggestive clinical symptoms, macroscopic lesions, histological evidence of CMV intranuclear inclusions, and/or positive culture for CMV. All cases of CMV organ disease were subsequently validated by a validation committee.

Statistical methods

Only measurements performed every 4 months were taken into account in the prognostic analyses to avoid any bias linked to a more regular follow-up of patients with a positive CMV culture or antigenaemia.

The chi square test was used to compare qualitative variables and the Mann-Whitney test to compare quantitative variables. The incidence rate of CMV disease was calculated based on person-years. Cox proportional hazard models were used to study which markers coded as time-dependent covariates were prognostic of a CMV disease.

Results

Baseline characteristics of the population

The study initially included 205 patients between the period June 1996 to April 1997. Seven patients were not included in the analysis. Four did not meet the inclusion criteria (three were seronegative for CMV and one had CMV disease diagnosed at inclusion) and three patients attended only for the inclusion visit and were immediately lost to follow-up. Minor protocol violation occurred in 20 patients: five patients not receiving HAART had a CD4 cell count > 100 × 10⁶ cells/l, eight receiving HAART had counts > 200 × 10⁶ cells/l, two patients did not undergo fundoscopy at baseline and five had been taking HAART for less than 2 months. These patients were nonetheless followed and analysed and their results included in the data presented here. Of these, 84% were men. The median age was 36 years (range 25-69), and 61% had an AIDS-defining illness prior to inclusion. The baseline median CD4 cell count was 77 × 10⁶ cells/l (0-308) The median plasma HIV-1 RNA concentration was 5600 copies/ml (range < 200 to 5 200 000); this was measured using the Amplicor assay for 176 patients and the Quantiplex bDNA assay for 21. The viral load was below the detection limit for 22% of the patients and > 100 000 copies/ml in 20% of the patients. At entry, 85% of the patients had been treated with an antiretroviral regimen containing a protease inhibitor for a median duration of 3.2 months (range 0.0-11.4).

Evolution of HIV infection and survival

At the time of evaluation in May 1999, the median follow-up was 23.6 months (range 0.5-28.0); this equated to 348 patient-years. The percentage of patients receiving HAART (defined as a regimen containing a protease inhibitor) increased during the study to reach 99% at 12 months (Table 1). An improvement of immunological and HIV virological markers was observed within the whole cohort: the percentage with a CD4 cell count < 100 × 10⁶ cells/l increased from 35 to 78%, and the percentage of patients with undetectable plasma HIV RNA increased from 22 to 33% (Table 1). The cumulative probability of survival was 92% at 24 months [95% confidence interval (CI) 88-96]. There were 19 deaths from various causes: lymphoma (6), progressive multifocal encephalopathy (2), pancreatitis (1), adenocarcinoma (1), suicide (1), septic shock (1), infectious acute respiratory failure (2), cardiac and renal failure (1), myocarditis (1), extensive Kaposi sarcoma (1) and terminal AIDS (2).

Table 1

Incidence of cytomegalovirus markers and disease

At inclusion, at least one CMV blood marker was present in 22 patients (11%): 13 with blood culture (7%), 14 with pp65 antigenaemia, (7%), five with plasma CMV DNA (3%) and six (three undetermined) with CMV mRNA (3%). In parallel to the improvement of the whole cohort, the blood markers progressively disappeared from the blood (Fig. 1). This was true for 84% of the patients; in these patients the CD4 cell count increased to > 75 × 10⁶ cells/l. In the 16% of patients for whom the CD4 cell count remained at or decreased to < 75 × 10⁶ cells/l, the prevalence of CMV blood markers remained high although decreasing slightly during the study period (0-19% depending on the marker) (Fig. 1). Pp65 antigenaemia, leukocytic mRNA, and blood culture were found positive at least once in 13-14% of the patients, and plasma DNA in only 8%.

Fig. 1

After a follow-up of 23.6 months, the incidence rate of CMV disease was 3.2/100 patient-years (95% CI 1.3-5.0), and the cumulative incidence at 24 months was 6% (95% CI 2-9), (Fig. 2). CMV disease was diagnosed in 11 patients located initially as retinitis (7), oesophagitis (1), encephalitis (2) and disseminated (1). Of these 11 patients, nine had been receiving HAART for a median duration of 8.9 months. Their respective CD4 cell count, HIV viral load at the time of CMV disease and the time lapse between initiation of HAART and disease are described in Table 2. It is noticeable that three of the 11 with CMV infection had > 50 × 10⁶ cells/l (52, 54, 160). Before the diagnosis of CMV disease, at least one CMV blood marker was present in six of the 11 patients: viraemia (5), pp65 antigenaemia (6), plasma CMV DNA (6) and CMV mRNA (3). At the diagnosis of CMV disease, plasma HIV RNA was at a median of 1600 copies/ml (< 200 to 400 000 copies/ml) and at least one CMV blood marker was present in seven of nine patients.

Fig. 2

Table 2

Prognostic factors of CMV disease

The proportion of patients developing a CMV disease was higher in patients with a marker detected during the follow-up than in those without (Table 3). In particular, 63% of the patients with pp65 antigenaemia > 100 nuclei/200 000 cells developed a CMV disease compared with 3% without antigenaemia. Of the patients with plasma CMV DNA, 38% developed a CMV disease whereas only 3% developed disease without showing plasma CMV DNA. In those patients with such a marker who developed the disease, pp65 antigenaemia > 100 nuclei/200 000 cells and plasma CMV DNA preceded development of disease by 57 and 61 days, respectively.

Table 3

In the 16 patients with plasma CMV DNA, the cumulative probability of CMV disease was 33% at 3 months (95% CI 9-57), 33% at 4 months (95% CI 9-57) and 40% at 1 year (95% CI 15-66) (Fig. 3). In these 16 patients with plasma detection of CMV DNA at least once, nine had a CD4 cell count < 75 × 10⁶ cells/l, of whom six developed the disease. No CMV disease occurred in the seven patients who had a CD4 cell count > 75 × 10⁶ cells/l and were tested once positive for plasma CMV.

Fig. 3

For the eight patients with pp65 antigenaemia > 100 nuclei/200 000 cells at least once, the cumulative probability of CMV disease was slightly higher: 50% at 3 months (95% CI 15-85), 63% at 4 months (95% CI 29-96) and 63% at 1 year (95% CI 29-96). Among these eight patients, five had a CD4 cell count < 75 × 10⁶ cells/l and developed CMV disease. The other three patients had a CD4 cell count > 75 × 10⁶ cells/l and did not develop disease.

In univariate analysis, it was shown that all the CMV markers, a CD4 cell count remaining > 75 × 10⁶ cells/l and an HIV viral load remaining at or increasing above 100 000 copies/ml were predictive for the occurrence of CMV disease (Table 4). In a stepwise multivariate analysis in which all the markers presented in Table 4 were taken into account, only three covariates were significantly associated with the occurrence of CMV disease: plasma CMV DNA, pp65 antigenaemia at or above 100 nuclei/200 000 cells and a CD4 cell count < 75 × 10⁶ cells/l. The respective risk ratios were 8.5 (95% CI 1.8-41.2), 17.2 (95% CI 3.3-88.4) and 10.7 (95% CI 1.2-91.3). None of the other factors, including plasma HIV RNA load, approached a significant level. When this analysis was limited to the patients who were receiving HAART since inclusion and who had a CD4 cell count of < 100 × 10⁶ cells/l at inclusion (105 patients; eight events of CMV infection), similar results were obtained for CMV markers. However, as all events occurred in those with CD4 cell counts below the threshold of 75 × 10⁶ cells/l, non-convergence occurred for CD4 count. The risk ratios were then 14.4 (95% CI 2.4-87.0) for plasma CMV DNA and 33.5 (95% CI 4.2-266.2) for pp65 antigenaemia at or above 100 nuclei/200 000 cells.

Table 4

Discussion

The Predivir cohort was implemented at the time when protease inhibitor therapy became available. This prospective study of 198 patients, a majority having a CD4 cell count < 100 × 10⁶ cells/l at the time of HAART initiation, focused on the predictive factors for and the characteristics of CMV disease in this new context. Only 11 CMV events occurred during the course of the study.

Before HAART, the prevalence and incidence of CMV end-organ disease in patients with a CD4 cell count of < 50 × 10⁶ cells/l, were 20-40%^[8,9] and 15-20/100 patient-years, respectively ^[14,15]. After the introduction of protease inhibitors in antiretroviral therapy, an impressive decrease of about fivefold in incidence of CMV disease was shown by several authors ^[14-16]. The incidence fell from 20.9 to 3.5/100 patient-years in a French cohort between January 1995 and June 1996 ^[15]. The low incidence of CMV disease, which was 3.2/100 patient-years (95% CI 1.3-5.0), in our cohort is consistent with the above findings. Moreover, most CMV-related events occurred within the first 10 months after initiation of therapy. The characteristics of the CMV infections still observed remained unchanged. The patients who developed a CMV disease were highly immunosuppressed, with a CD4 cell count < 50 × 10⁶ cells/l in 70%. The most common site of CMV disease remained the retina, although two patients developed encephalitis as the first end-organ site of disease.

At the end of the 24-month follow-up, 84% of the patients had a CD4 cell count of > 75 × 10⁶ cells/l. In parallel with the rise in CD4 cells to > 75 × 10⁶ cells/l, markers of CMV blood dissemination were more rarely detected. In patients who control their HIV viral load under HAART and whose CD4 cell count increased, specific anti-CMV lymphoproliferative response was shown to reappear after a few weeks. ^[2]. In absence of sustained HIV suppression, no restoration of CMV-specific lymphoproliferative response occurred. Moreover, anti-CMV lymphoproliferative response has been shown to reappear in patients during therapeutic success and to disappear after the treatment had eventually failed ^[17]. The restoration of lymphoproliferative response to CMV was associated with absence of retinitis relapses in patients given HAART after a previously diagnosed retinitis ^[17].

The most frequently encountered markers for CMV infection were pp65 antigenaemia, CMV RNA and CMV blood culture; these occurred more frequently than plasma CMV DNA. Although in univariate analysis, all the CMV markers as well as a CD4 cell count < 75 × 10⁶ cells/l and a plasma HIV viral load > 100 000 copies/ml were predictive for the occurrence of a CMV disease, only three factors remained independently associated with the disease in a multivariate analysis: pp65 antigenaemia > 100 nuclei/200 000 cells was the most predictive, followed by plasma CMV DNA and a CD4 cell count remaining at or decreasing to < 75 × 10⁶ cells/l. In this analysis, no threshold of HIV viral load was predictive of a CMV disease. As only 21 patients were followed for HIV viral load using the Quantiplex b-DNA technique, the results for the prognostic value of viral load were derived mainly from the results of the Amplicor Monitor technique. When the analysis was limited to the data from the latter technique, the results were identical (data not shown). Because of the small number of subjects, no subgroup analysis limited to patients followed by the former technique was possible.

It should also be noted that, given the small numbers of CMV events, the results of the multivariate analysis have wide confidence intervals and should be interpreted with caution.

A bias possibly limiting the interpretation of the results could have been the differential evaluation of patients; those having a positive assay for CMV were examined more frequently for ethical reasons. To avoid this, only the biological results collected every 4 months were used in the statistical analyses. Moreover, CMV disease does not usually remain asymptomatic for a prolonged period and no retinal scar, suggestive of a spontaneously cured retinitis, was evidenced on the fundoscopy performed every 4 months in those attending only at 4-month intervals.

Hence, in the HAART period, the presence of markers of CMV blood dissemination remains a major risk factor for further development of CMV disease. The PCR assay for detecting CMV DNA from plasma is a convenient diagnostic procedure. It can be performed within 5 h using a commercially available test. Only small amounts of plasma, minimally processed are sufficient and specimens can be frozen without requirement for immediate processing. However, there was only a 2-month interval between detection of CMV DNA in plasma and the diagnosis of CMV disease for 40% of the patients with plasma CMV DNA who developed the disease; by comparison, there was an interval of 4 months between pp65 antigenaemia or CMV isolation from blood and onset of disease. Will plasma CMV DNA be a marker that can be detected early enough to allow initiation of pre-emptive treatment in HIV infected patients? The risk for the patient would be that treatment would only be initiated after a CMV-related event has occurred. However, 60% of the patients in whom CMV DNA was detected once in plasma did not develop the disease, probably because a subsequent immune restoration under HAART occurred.

Antigenaemia with pp65 is an earlier predictor of CMV disease. The assay can be completed within 5 h after sampling and the level is easily assessed by counting the marked cells in a population. Its high predictive value when interpreted quantitatively has been recently shown ^[18] The major limitation of this test is the need for rapid processing of the specimens to avoid autolysis of blood leukocytes, which leads to false-negative results. Moreover, this is a labour-intensive test and is not suitable for large series of specimens; in addition, the results depend on the anti-pp65 monoclonal antibodies used in the test. However, its high predictive value after measurement plus its early detection (before CMV DNA) in plasma makes it an interesting tool until a more sensitive technique using PCR is commercially developed. CMV mRNA detection by PCR and CMV culture appeared less interesting as prognostic tools.

The above data suggest that, in the HAART period, it is possible to define a group of patients at risk for CMV disease for which a strategy of pre-emptive therapy can be proposed. This includes those who have a CD4 cell count of < 75 × 10⁶ cells/l and a blood marker demonstrating CMV reactivation (plasma CMV DNA or pp65 antigenaemia > 100 nuclei/200 000 cells). Such a strategy has the advantage of avoiding treatment of all patients with a CD4 cell count < 75 × 10⁶ cells/l and CMV seropositivity, among whom only a few will develop CMV disease. It can be proposed more easily now that new oral drugs are available, such as proganciclovir, that can be taken once or twice daily. An important consideration, however, will be whether the predictive factors for CMV disease remain constant in patients for whom HAART therapy starts to fail.

Acknowledgements

We are grateful to the patients for their collaboration and to Sandrine Jacquet for the preparation of the manuscript.