Background
Infectious intestinal disease (IID) is a syndrome of mixed aetiology; many different pathogens can infect the human gastrointestinal tract and produce diarrhoea, vomiting and other characteristic symptoms. Mixed gastrointestinal infections are frequently detected, especially in infants and young children and when polymerase chain reaction (PCR) assays are used for diagnosis [
1,
2]. It is important to determine which pathogen is the cause of illness, in order to direct clinical management for individual patients and to advance epidemiological understanding of IID.
Reverse transcription- PCR (RT-PCR) is now the method of choice for detecting norovirus in clinical specimens. RT-PCR detects norovirus at lower concentrations and is less affected by specimen quality and preparation than electron microscopy [
3‐
5]; large numbers of specimens can be tested simultaneously, compared to the single throughput for electron microscopy. RT-PCR also detects a much wider range of norovirus genetic variants than enzyme-linked immunosorbent assays (ELISA) and may be more easily adaptable for detection of new strains [
6].
However, many healthy individuals, with no recent history of IID, are RT-PCR positive [
7‐
9], meaning that virus detection by RT-PCR is not well correlated with disease in norovirus infection. If RT-PCR positivity does not necessarily equate to norovirus-associated IID, it cannot be used alone to attribute illness to norovirus in IID cases; it is possible that the norovirus infection is 'asymptomatic' in the IID case, with another pathogen, detected or undetected, actually causing the symptoms. The poor diagnostic specificity of PCR and the associated difficulties for clinical interpretation of test results have been highlighted for other viral pathogens [
10,
11].
Previous studies have demonstrated differences in faecal norovirus load between symptomatically and asymptomatically infected individuals [
7,
12]. Histopathological investigations of experimentally inoculated volunteers and naturally infected individuals also indicate that the mechanism of pathogenesis in norovirus infection may rely on damage to the intestinal epithelium, caused by viral replication [
13‐
15], so that symptoms may be a result of high viral loads. The aim of this study was to use faecal viral load measurements to determine when illness is attributable to norovirus in IID cases.
Discussion
In this study we have demonstrated a difference in viral load between symptomatic and asymptomatic norovirus infection. A substantial proportion of IID cases who were positive only by RT-PCR had viral loads equivalent to those in healthy controls. This supports the hypothesis that norovirus is not always the cause of illness where it is detected by RT-PCR. We have shown that it is possible to use the viral load in clinical specimens to indicate where norovirus is the most likely cause of illness, by selecting a cut-off for the norovirus real time RT-PCR assay. We have also shown that the method of cut-off selection can be adapted for use with specimens that are routinely received and tested in clinical laboratories, to help other laboratories develop in-house cut-offs for their assays. This is essential because there is substantial variability between UK virology reference laboratories in the Ct values produced from standard reference specimens [
31]; the same cut-off may not be appropriate for all laboratories because of these differences in assay performance.
A major strength of this study is the availability of specimens from healthy controls. There are few community studies of IID with large control groups available, but they are essential for interpreting the RT-PCR data in IID cases. Importantly, it has been possible to validate the use of bacterial culture positive IID cases as a reference negative group, by comparison to the ROC analysis using healthy controls; this removes the need to collect further control specimens in future studies. We have also shown that RT-PCR positive IID cases, who are negative for other common bacterial, protozoal and viral pathogens, are a suitable reference positive group, so that the method can be used by laboratories without EM testing facilities. These reference groups can now be used by other laboratories for development of cut-offs for their assays.
The area under the ROC curve for the alternative reference groups is very low, possibly because the viral loads in many of the IID cases in the reference positive group were not representative of symptomatic norovirus infection; this is reflected in the low sensitivity for the cut-off at Ct value 31 when using these groups in the ROC analysis. However, selection of an appropriate cut-off is the main aim of this method and we have shown that this is possible with these reference groups. It is also important that the diagnostic performance (sensitivity, specificity, predictive values) of the cut-off is determined using an independent dataset, which was not possible in this study; the sensitivity and specificity calculated in the ROC analysis may misrepresent the performance of the cut-off in clinical application, because the cut-off is selected by optimising the diagnostic accuracy compared to the gold standard [
32,
33].
The specimens used in this study were originally collected during the mid 1990s and the viral RNA may have degraded during the prolonged storage and repeated freeze-thaw cycles for re-testing. Therefore the cut-off developed here should not be directly applicable to real-time RT-PCR results from fresh specimens without validation. Similarly, the cut-off should not be applied to assays with different protocols, because the Ct values may not equate to the same viral load per gram of faeces. It is unlikely, however, that there will have been differential degradation of RNA between specimens during storage, so it is still valid to compare the viral load between specimens in this collection, and to assume that the relative differences observed between IID cases and controls are a true reflection of symptomatic and asymptomatic infection. It is also important to note that any cut-off in viral load can only be applied to specimens collected from IID patients during acute symptoms, when the viral load is representative of disease aetiology. After symptoms resolve in norovirus-associated IID, the viral load quickly drops to levels seen in asymptomatic infection [
12] and the predictive value of the cut-off will be greatly reduced.
The cut-off developed here is not applicable to two of the rarer genotypes in genogroup II (GII-7 and GII-8), because the real time RT-PCR assay has poorer efficiency (a higher detection limit) for these genotypes (J. Gray, personal communication), so the Ct values do not represent the same faecal viral loads as for the other genotypes. At a population level, the degree of misclassification would be small because of the low prevalence of GII-7 and GII-8 [
34‐
36]. However, correct identification of illness caused by these genotypes may be important for clinical management, but would require development of genotype-specific cut-offs. Similarly, we have excluded genogroup I noroviruses from this analysis because the efficiency of the assay is highly variable for genotypes within in this genogroup. Development of a cut-off for GII-7 and GII-8 or genogroup I noroviruses would require collection of sufficient specimens for genotype-specific ROC analyses; clinical application would require genotyping to be part of routine diagnosis, which may not be economically or logistically feasible. Further work is also needed to characterise the kinetics of the real time RT-PCR assay, to determine whether a Ct value of 31 translates to the same faecal viral load for all genogroup II genotypes with the same assay efficiency. Selection of a single cut-off may also not be appropriate if the Youden index is similar for a range of Ct values between 28 and 33, as was the case in this analysis. With a larger sample size, in future studies, there may be better power to discriminate between potential cut-offs in this range. Nevertheless, the cut-off provides a major improvement in diagnostic specificity compared to the current qualitative use of RT-PCR in norovirus diagnosis.
The causal relationship between disease symptoms and viral load has not been established. However, if the relationship between the occurrence of disease and viral load is consistent, regardless of whether high viral loads are a cause or a consequence of disease, viral load will be a good marker of norovirus-associated IID and the approach developed here is valid. Viral load is routinely used to predict outcome and guide clinical management for a number of viruses that cause chronic infections, such as Epstein-Barr virus [
37] and cytomegalovirus [
38] in transplant patients, HIV [
39], hepatitis C [
40] and HTLV [
41]. However this is the first time, to our knowledge, that viral load has been used as a tool for diagnosing enteric viruses as the cause of acute IID.
Conclusion
As PCR diagnosis is applied to an increasing number of viral pathogens, the debate is growing about the clinical interpretation of positive results and the utility of PCR in diagnostic services [
10,
42,
43]. PCR has many advantages over traditional diagnostic methods, including higher throughput, shorter turnaround time, adaptability to new strains and production of data for molecular epidemiological surveillance. It is therefore important to ensure that clinically informative results are produced from PCR assays, to provide a high standard of patient care alongside these other benefits. The method developed here shows that the real-time RT-PCR output for norovirus can be used to attribute disease to norovirus in IID cases, where simple detection may not be sufficient to give a confident diagnosis of norovirus-associated IID. This semi-quantitative approach to diagnosis can improve both the accuracy of community-based estimates of norovirus associated IID incidence and the interpretability of diagnostic results provided to clinicians from clinical virology laboratories. However it is important that clinical and epidemiological information is considered in the diagnosis of disease aetiology for individual patients with Ct values close to the cut-off.
Independent validation of this method is required prior to application in other studies and laboratories; we have provided a method for validation without the need for collection of specimens from healthy controls or further use of EM. The method may also be useful for other viral pathogens, for which the same problems with the interpretability of PCR have been described. Future work will focus on applying this approach for estimation of norovirus associated IID incidence and describing the implications for diagnosis of norovirus outbreaks.
Acknowledgements
This study was funded by the Food Standards Agency. We would like to thank Corrine Amar, Lisa Griffiths, Fenella Halstead, Dalia Choudhury and Mihaela Cirdei who completed the laboratory testing.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
GP performed analyses and drafted the manuscript. CT and BL advised on analysis. JG and MIG led the laboratory work and advised on analysis. JG and DB led the study design. All authors contributed to the drafting and revisions of the manuscript.