Enhancing the predictive value of throat swabs in virological influenza surveillance

K. Leitmeyer, U. Buchholz, M. Kramer, B. Schweiger

Robert Koch-Institut, Berlin, Germany

According to a national survey in Germany, the influenza virological surveillance can be improved when sentinel practitioners take throat swabs specimens only from patients who consult early after the disease onset (ie, within 48 hours), and when they use the strict clinical case definition of influenza-like illness. PCR should be used for primary detection of influenza viruses.
Countries within the European region conduct surveillance for influenza because of its marked impact on morbidity, mortality, and the economy each winter season. The European Influenza Surveillance Scheme (EISS) facilitates the exchange of information on influenza activity collected by the national surveillance systems, and other countries are in the process of setting up influenza surveillance. Most countries’ surveillance for influenza consists of a clinical and a virological component. The latter may include the identification of viruses, their type, their subtype, and the characterisation of viruses by serological (haemagglutination inhibition) as well as molecular methods (sequencing). National reference laboratories usually identify influenza viruses either through virus isolation, enzyme linked immunosorbent assay (ELISA), immunofluorescence (IF) or reverse transcriptase polymerase chain reaction (RT-PCR). Internationally, virus isolation is still considered the gold standard.

The positive predictive value (PPV) for influenza of specimens taken from patients with respiratory tract infections is crucial for the ability to recognise the initial phase of an influenza epidemic. The PPV of a case definition plus a laboratory test depends on the sensitivity of the case definition, the specificity of the case definition, and the laboratory test method, as well as the prevalence of the disease in the population.

To develop recommendations for national influenza surveillance in Germany, we compared PCR and virus isolation regarding their sensitivity to identify influenza, and assessed other factors thought to be relevant to increase the yield of positive tests, such as the interval between disease onset and sample collection, as well as the time between sample collection and receipt of the sample by the laboratory.


We used data from the influenza season 2000–01. The National Influenza Reference Centre (NIC) in Berlin, Germany, which has been an active EISS member for several years, provided swab containers and pre-addressed and stamped envelopes to 105 sentinel physicians. The NIC advised them to take throat swabs from 3 to 5 patients per week if they presented with influenza-like illness (ILI), defined as acute onset of fever, dry cough, headache, myalgia or malaise. Sentinel physicians sent clinical samples to the NIC by mail. The NIC tested all samples by virus isolation and by PCR (1). All routine PCR tests included positive and negative control strains (quality controls).

We defined the time interval between disease onset and sample collection as “interval 1”, the time interval between sample collection and receipt of the sample by the laboratory as “interval 2”, and we performed the following calculations:

• Comparison of the overall proportion of influenza-positive samples by test method, i.e. virus isolation and PCR

• Relation between the proportion of influenza-positive samples and interval 1, comparing PCR with virus culture (information available for 1882 specimens)

• Relation between the proportion of specimens testing PCR-positive and interval 2, stratified by interval 1 (information available for 1879 specimens)


During the influenza season 2000–01, the NIC received 2592 processable specimens: 642 (25%) of the specimens tested positive by PCR, and 439 (17%) of them were culture-positive. No virus could be isolated in 204 (32%) of 642 specimens positive for PCR, but among 439 specimens from which an influenza virus could be cultured, PCR failed only once (0.02%) to detect influenza virus.

Of 1945 specimens where information on the date of onset and the date of specimen collection was available, 1430 (74%) were taken within 48 hours of disease onset. We observed a clear downward trend of the proportion of influenza-positive specimens both for PCR and virus isolation as interval 1 increased (chi2 for linear trend=42 and 39, both p-values<0.001). If the specimen was collected within 48 hours of disease onset, the proportion of specimens positive by PCR was between 23% and 40% (figure 1). Notably, for all time points of interval 1 between day 0 and day 6, use of PCR resulted in 8-16% more positive results than virus isolation.


A clear linear trend between the proportion of specimens positive by PCR and interval 2 could not be found for any of the strata of interval 1 (c2 for linear trend between 0 and 4, p-values>=0.04; Figure 2). Results from virus culture give a similar picture.



Our focus in this study was the role of modifiable factors that could influence the PPV of specimens taken from patients with ILI. First, the yield of positive samples can be enhanced when the swabs are taken early (between 24 and 48 hours) after disease onset, independent of the method used. This reflects the pattern of virus shedding. Influenza virus replication peaks at about 48 hours after infection and declines slowly thereafter, with little shedding after days 6 to 8 (2,3). The number of positive test results is significantly higher when PCR is used for detection of influenza viruses. This was demonstrated during the season 2000/2001 when 25% of the specimens were positive by PCR, compared with only 17% that could be cultured. Similar results were obtained during previous seasons (1,4). In addition, PCR results are usually available on the same day while virus isolation takes on average about a week (4).

Second, the specificity of ILI is substantially higher compared to, for example, acute respiratory infection (ARI), and is therefore used by most countries for obtaining specimens for influenza testing (5,6). The specificity of the primer/probe sets used in our TaqMan® PCR for amplification and detection of influenza virus genomes is very high (100%) as has been shown by investigating a number of influenza reference strains and clinical samples (1). Virus culture is also estimated to be well above 95%.

Third, PPVs are influenced by small increases in the prevalence or incidence of a disease from very low to low or moderate levels. This is the case at the beginning of an influenza epidemic, as observed during the influenza season 2000/01 (7). Therefore, the proportion of patients with ILI whose throat swabs are positive for influenza can be considered a sensitive monitor to assist in detecting the beginning of substantial influenza activities at the start of the season.

We were able to show that the proportion of PCR-positive specimens is affected very little by the time taken to transport specimens between sentinel physicians and the laboratory. This supports the practice of most surveillance systems, which use standard mailing of non-refrigerated specimens for transport.

Internationally, virus isolation, although time consuming, is still considered to be the gold standard for detection of influenza viruses, because more rapid tests, such as ELISA or IF, are of limited sensitivity and specificity. PCR is very specific, and the most sensitive technique which has been described for typing and subtyping of influenza viruses. This report confirms previous findings comparing virus isolation and PCR over several years (1,4,8), which are that PCR is substantially more sensitive than virus isolation. This difference may be explained by the fact that PCR requires viruses in only very low concentrations, and which do not need to be capable of replication. Its use in routine virological surveillance has now been demonstrated. Virus isolation will continue to be indispensable for the comprehensive characterisation of circulating influenza viruses.

In conclusion, to optimise early recognition of the beginning of an influenza epidemic through virological surveillance, we recommend that sentinel physicians be advised to take specimens only from patients who present early after disease onset (for example, within 48 hours of disease onset), that they use the strict clinical case definition of ILI, and that PCR be used for primary detection of influenza viruses.

References1. Schweiger B, Zadow I, Heckler R, Timm H, Pauli G. Application of a fluorogenic PCR assay for typing and subtyping of influenza viruses in respiratory samples. J Clin Microbiol 2000; 38: 1552-8. (http://jcm.asm.org/cgi/reprint/38/4/1552.pdf)

2. Murphy BR, Chalhub EG, Nusinoff SR, Kasel J, Channock RM. Temperature-sensitive mutants of influenza viruses. 3. Further characterization of the ts-1(E)] influenza A recombinant (H3N2) virus in man. J Infect Dis 1973; 128: 479-87.

3. Frank AL, Taber LH, Wells CR, Wells JM, Glezen WP, Paredes A. Patterns of shedding of myxoviruses and paramyxoviruses in children. J Infect Dis 1981; 144: 433-41.

4. Schweiger B, Timm H. Die TaqMan-PCR: Schnelle Typisierung und Subtypisierung von Influenzaviren direkt aus Patientenmaterial. [A TaqMan-PCR for rapid typing and subtyping influenza viruses in clinical specimens.] Bundesgesundheitsbl. –Gesundheitsforsch. Gesundheitsschutz 2000; 43: 788-95. [in German, English abstract available from http://yellow-fever.rki.de/FORSCH/PUBLIST/2000/00S17.HTM]

5. Webpage of the European Influenza Surveillance Scheme (EISS). Available at: http://eiss.org/html/introduction.html, accessed on November 14, 2001.

6. Aguilera JF, Paget WJ, van der Velden J, Watson JM. Heterogeneous case definitions used for the surveillance of influenza in Europe. Poster presentation, 6th EPIET Scientific Seminar; October 2001; Veyrier du Lac, France.

7. Uphoff H, Heckler R, Schweiger B. Ergebnisse der Influenzasurveillance im Winter 2000/01. [Results of the Influenza Surveillance Program for the 2000-2001 Influenza Season]. Bundesgesundheitsbl. – Gesundheitsforsch. – Gesundheitsschutz 2001, 44: 1169-73. [in German, English abstract available from http://www.rki.de/FORSCH/PUBLIST/2001/01U1.HTM]

8. Ellis JS, Fleming DM, Zambon MC. Multiplex reverse transcription-PCR for surveillance of influenza A and B viruses in England and Wales in 1995 and 1996. J Clin Microbiol 1997; 35: 2076-82.

Copyright(c) 2015 Eurosurv.All Rights Reserved