Viral load as predictor of Crimean-Congo hemorrhagic fever outcome.

We used quantitative real-time reverse transcription–PCR to measure viral load in serum from 24 patients in Kosovo who had acute Crimean-Congo hemorrhagic fever. Viral load correlated with clinical disease and antibodies and could be used as a predictor of disease outcome.


The Study
Serum samples were obtained from CCHF patients from Kosovo in 2001 (www.who.int/csr/don/2001_06_ 29e/en), 2003, and 2005. Clinical and biochemical data were provided by the Clinic of Infectious Diseases, Pristina, Kosovo. Presence of an acute febrile syndrome characterized by malaise, nausea, fever, and bleeding from various sites was reported as well as possible modes of infection. Leukocyte and platelet counts, aspartate and alanine aminotransferase levels, activated partial thromboplastin times, and creatinine values were available for most patients. Patients were categorized into 3 groups according to disease severity: fatal, severe, or moderate cases. On the basis of classifi cation by Swanepoel (4), severe cases were defi ned by the presence of hemorrhagic manifestations (epistaxis, hematemesis, and melena), lowered blood pressure (<100/60 mm Hg), and raised serum creatinine and transaminase levels.
Serologic testing for anti-CCHF virus immunoglobulin (Ig) M and IgG was done by ELISA. Molecular data were obtained by real-time reverse transcription-PCR (RT-PCR) with a limit of detection of 240 copies/mL of sample as recently described (10). Some assay modifi cations were necessary for accurate quantitation of viral load. Synthetic RNA was generated as a quantitative calibrator, and a competitive internal control was constructed as previously described to detect possible infl uences of PCR inhibitors (11). The original real-time RT-PCR protocol was complemented by the addition of 200 pmol/μL of internal control probe YFP2 (5′-ROX-ATCGTTCGTTGAGCGATTAG-CAG-BBQ-3′). This probe recognizes an alternative binding site introduced in the target gene by overlap-extension PCR (11). The standard curve for CCHF virus quantitation was based on synthetic calibrator RNA with concentrations from 24 × 10 5 to 24 × 10 1 copies/mL. Statistical analysis was performed with statistical software R version 2.2.1 (www.r-project.org) and the Statgraphics 5 package (Manugistics, Dresden, Germany).
A total of 24 patients had clinical, serologic, and molecular confi rmation of CCHF (Table). All 24 patients had an acute febrile syndrome, 8 reported tick bites, and 2 had been exposed in a hospital. For the 9 patients who died, only 1 serum sample was available from each. For the 9 patients with severe disease and the 6 with moderate disease, 2-3 consecutive samples were available.
From the 24 patients, 43 serum samples were tested by real-time RT-PCR and ELISA (Table). Viral loads ranged from 10 2 to 10 10 copies per milliliter of serum, depending on the day of illness, the severity of disease, and the results of serologic analyses (Table). Whether early laboratory fi ndings could serve as prognostic markers for outcome was explored. Because prognostic information is most relevant in the fi rst week of disease, only samples taken up to *Institute of Microbiology and Immunology, Ljubljana, Slovenia; †Clinic of Infectious Diseases, Pristina, Kosovo; ‡National Institute of Public Health, Pristina, Kosovo; and §Bernhard Nocht Institute for Tropical Medicine, Hamburg, Germany day 7 of symptoms from any patient were included in the analysis. Average sampling days did not differ between patients who died and those who survived (5.0 and 3.9 days, respectively, p = 0.28, analysis of variance [ANOVA]). No patient from either group had detectable IgG in the fi rst week. Although we suspected that early development of IgM might correlate with good outcome, such correlation was not found. Of those patients in whom IgM was detected up to day 7, 5 (62%) died. Of those without IgM in the fi rst week, 3 (42%) died (insignifi cant difference in 1way ANOVA, p = 0.6). Furthermore, the presence of IgM in the fi rst week did not correlate with viral load, which suggests that virus levels in the fi rst week can be regarded as an independent prognostic parameter. Namely, viral load seemed to be strongly related to the clinical classifi cation (p<0.001), with the average log value 9.25 (1.78 × 10 9 ) in the group of patients who died and 6.91 (8.06 × 10 6 ) in the group who survived (Figure).
To determine whether IgG could infl uence viremia, viral loads were correlated with log-transformed reciprocal antibody titers. Quantitative IgG levels showed a highly signifi cant inverse correlation with viral loads (p<0.0001) (Figure). It was thus reasoned that IgG levels could infl uence the later course of disease and, in particular, could re-  fl ect a discrimination between severe and moderate cases. In samples from both categories, no relationship between IgG and clinical classifi cation (p = 0.65) was determined after day 7. Also, no signifi cant relationship was found between clinical classifi cation and viral load (p = 0.74) in severe versus moderate cases. On average, viral load log value was 2.38 in severe cases and 2.69 in moderate cases (Figure).

Conclusions
This study describes the differential infl uences of CCHF viral load, IgM, IgG, and clinical outcome. CCHF viral load, but not IgM, could be used as a predictor of CCHF outcome. It was unexpected that IgM correlated with neither outcome nor viral load. On the contrary, quantitative IgG levels inversely correlated with viral loads, which suggests that IgG might neutralize virus in vivo. The fact that virus titers decreased in survivors independent of antibodies during the fi rst week implies involvement of innate or cellular immune mechanisms in the elimination of CCHF virus. CCHF viral load ranged from 10 2 to 10 10 copies/mL in the serum samples. It was shown that viral load of >10 8 copies/mL is a strong factor (p<0.001) for differentiating CCHF patients who died from those who survived. However, viral load does not help differentiate between severe and moderate cases according to common case defi nitions (4). The same was true for IgM levels. Viral load is also useful for estimating need for infection control measures. Viral loads measured in our patients were high, >10 9 copies/mL, higher than viral loads in other arboviral diseases that are not easily transmitted in the hospital, e.g., dengue (12). This fi nding could help explain why CCHF virus causes nosocomial infections on a regular basis. Another use for this fi nding is systematic monitoring of patients receiving ribavirin therapy. In the absence of suffi ciently large numbers of treated patients, however, we could not investigate this application. Figure. Correlation between clinical outcome, serologic data, and Crimean-Congo hemorrhagic fever (CCHF) viral load measurements. A) Viral load versus immunoglobulin (Ig) M result taken during the fi rst week of illness. B) Viral load versus outcome. Average viral loads were 1.6 × 10 9 copies/mL in persons who died and 5 × 10 6 copies/mL in persons who survived (difference highly signifi cant, p<0.0001). The dot is a datum point that has been identifi ed as an outlier. C) Statistically signifi cant difference (p<0.001) in CCHF viral load and day of illness between group who died and group who survived. D) No correlation in viral load and day of illness between severe and moderate CCHF cases. E) Inverse correlation of quantitative IgG levels with viral loads (p<0.0001) in samples taken after fi rst week of illness. Black dot, >1 sample; *, fi rst week samples.