Rapid antigen testing for the surveillance of influenza epidemics


Corresponding author and reprint requests: W. Wunderli, Central Laboratory of Virology, University Hospital of Geneva, 24, rue Micheli-du-Crest, 1211 Geneva 14, Switzerland
Tel: +41 22 372 40 86
Fax: +41 22 372 49 88
E-mail: werner.wunderli@hcuge.ch


Objective  To assess the use of a ‘near patient’ test for rapid antigen detection to obtain the more timely acquisition of data for the surveillance of influenza epidemics.

Methods  To the classical cell culture system used for the surveillance of influenza, a ‘near patient’ test was added. The cell culture method was applied for the detection of influenza virus in specimens sent to our laboratory. In contrast, the ‘near patient’ test was used directly by practitioners in their practices to screen patients for the presence of influenza virus antigen.

Results  The results for two seasons are presented. The ‘near patient’ test was able to detect a developing influenza epidemic with the same reliability as clinical consultation reports for influenza-like illness or the conventional culture method. However, the results obtained were available 9 days earlier on average, compared with cell culture. Because of this, results concerning the epidemics could be announced via the internet more rapidly. Although the ‘near patient’ test demonstrated a lower sensitivity than detection by conventional cell culture, the sensitivity was still sufficiently high to reveal the characteristics of the epidemics in the community.

Conclusions  Rapid influenza testing is a reliable tool for influenza surveillance and, compared with traditional methods (virus detection on cell culture and monitoring of influenza-like illness), provides faster results. Although the ‘near patient’ test has limited sensitivity compared with cell culture, results were consistent over two seasons, and suggest that rapid testing should be part of a surveillance program.


The detection of circulating influenza viruses is an important component of a global surveillance program organized by the World Health Organization (WHO). Surveillance can be conducted by monitoring epidemiologic markers such as consultation rates for influenza-like illness (ILI), hospitalization for pulmonary infections, consumption of certain drugs [1], and the detection and identification of the circulating virus strains [1]. Virologic surveillance can be carried out by different techniques, but few are standardized, and the choice of methods depends primarily on the laboratory in charge of the surveillance. Consequently, a direct comparison of results is often impossible.

The major goal of virologic surveillance is to obtain information on the types and variants of influenza virus circulating. These data are currently obtained through the collection and screening of a representative set of samples for the presence of influenza viruses. From a collection of isolates, the number of variants can be estimated by the application of appropriate techniques. Through the combination of results from the global network of influenza surveillance, the frequency and geographic distribution of variants provide important information for the adaptation of the composition of the vaccine. The rapid availability of information about an ongoing epidemic is of less importance for this purpose than precise and timely knowledge of the type and number of variants circulating over the season. However, for the initiation of preventative measures or for the use of antivirals, timely information on the kinetics of an ongoing influenza epidemic is of much greater importance. For this reason, the results produced by virologic surveillance should be available with as little delay as possible.

In Switzerland, monitoring of ILI is based on morbidity reporting by physicians participating in the Swiss Sentinel Surveillance Network (SSSN) and by the isolation of influenza viruses at the National Center of Influenza (NCI). A disadvantage of this surveillance network is that results are not rapidly available, for several reasons. To study the potential of a ‘near patient’ test as an epidemiologic tool, a national pilot study was conducted over 2 years. In addition to conventional virus detection and monitoring of ILI by the SSSN, influenza activity was assessed by an immunochromatographic assay for the detection of influenza A and B virus antigen in throat swabs. The aim of the present study was to assess whether the ‘near patient’ test was able to detect a potential influenza epidemic, and to determine whether a reduction in reporting time could be gained through this approach.

Materials and methods

Patients with influenza-like illness

Two hundred and twenty practitioners participate in general in the SSSN and report the number of influenza cases observed in their practice once weekly to the Federal Office of Public Health (medical contacts for ILI), according to criteria defined by the Network [2]. All data are managed by the Federal Office of Public Health, with updates provided once weekly.

A restricted number of practitioners (55 on average) sent samples to our laboratory for virus detection by isolation on cell culture (throat and nasal swabs were combined in one tube with transport medium for each patient). Based on the same case definition for ILI, 198 practitioners selected patients for screening for virus secretion (virus antigen) by the ‘near patient’ test. For the 1999–2000 and the 2000–2001 seasons, 1104 samples were sent to our laboratory for virus isolation by conventional culture, and 4082 ‘near patient’ tests were performed by general practitioners.

Cell culture and antigen detection

Attempts to isolate influenza viruses by cell culture were made as previously described [3] on MDCK, LLC-MK2 and A549 cells at 37 °C with one major modification, since cultures were grown only in shell vials, instead of in tubes. MDCK cells were also used at 33 °C for virus isolation. Shell vials were inoculated with 400 µL of the sample and cultured for 7 days. Cells were removed mechanically from the shell vials, washed once in phosphate-buffered saline, applied to 12-well slides, air-dried, fixed in methanol, dried, and incubated with a pool of monoclonal antibodies against respiratory viruses (Respiratory Panel I Viral Screening and Identification Kit, Chemicon Products, Temecula, CA, USA) for 45 min. Slides were then washed again, and incubated with antimouse antibodies conjugated to FITC (included in the kit) for 30 min at 37 °C. Finally, slides were screened for the presence of typical granular fluorescent cells. If a positive result was obtained, the presence of influenza A or B was confirmed with monoclonal antibodies from the kit against these two viruses. Further characterization of variants of viruses obtained from positive cultures was carried out by the hemagglutination inhibition technique [3] with hyperimmune animal sera (ferret) obtained from the WHO reference center in London, UK.

For the detection of influenza virus antigen using only throat swabs, a test that detects a nucleoprotein from the virus by an immunochromatographic method was used (INFLUENZA A/B-RAPID TEST from Roche Diagnostics [4]). According to the manufacturer's information, the test showed a sensitivity of 85% and a specificity of 90% in comparison with the cell culture techniques [4].

Practitioners performed the ‘near patient’ test according to the manufacturer's instructions after specific training. Moreover, a ‘hotline’ for technical support was available during the winter season. In contrast to virus isolation on cell culture, practitioners took throat swabs for immediate testing from patients presenting with typical ILI as defined by the SSSN. The antigen detection test was validated for throat swabs only. Although selection criteria were identical for both methods, no patient underwent simultaneously the ‘near patient’ test and conventional culture for the detection of influenza virus. The results of the ‘near patient’ test were faxed to our laboratory by practitioners on a daily basis for registration.

For reporting purposes, the date of sampling was selected for both methods. Results were analyzed twice weekly and published on our website (http://www.influenza.ch).


Characteristics of the 1999–2000 and 2000–2001 influenza seasons

The results of the surveillance are presented in Figure 1. For the 1999–2000 season, the peak of the influenza activity was reached in the first week of 2000, whereas in the following year (2001), the maximum activity was observed only in the sixth week (maximum weekly consultation rates 7.6% and 3.3%, respectively). Detection of influenza virus by cell culture peaked in the first week of 2000, with 39 positive cases detected. During the following season, fewer positive samples were obtained, and the number reached a plateau in the third week, with almost 20 positive samples detected weekly, which remained the case until the sixth week.

Figure 1.

Epidemiologic curve of the influenza virus detected each week by cell culture and by the ‘near patient’ test during the 1999–2000 and the 2000–2001 seasons. Absolute number of influenza-positive samples detected by the ‘near patient’ test (open bars) or by the culture method (black bars). MC-ILI (%): percentage of medical contacts for ILI declared by the SSSN participants. Dotted line: threshold of 1.5% (MC-ILI) for an influenza epidemic. Values above 1.5% indicate an ongoing epidemic. Numbers on the horizontal axis indicate the number of the week of the corresponding year where the results were indicated.

Detection of circulating influenza virus using the ‘near patient’ test showed a slight difference in the kinetics compared with ILI and virus isolation by conventional cell culture (Figure 1). The highest number of positive samples was observed in the second week of 2000; in the following season, no peak could be detected, because the number of positive samples remained stable between the fourth and the seventh weeks (2001). Nevertheless, both detection systems were able to reveal the beginning and the end of the epidemic with a comparable kinetic.

Time scale of availability of results

Compared with conventional cell culture, major differences in the availability of results and data on the type of virus circulating were observed using the ‘near patient’ test. The number of cases of ILI observed by SSSN was reported by participants at the end of each week to the Federal Office of Public Health. Analysis and calculations were done in a block at the beginning of the following week, with information about the medical consultation rate from the previous week being available on Tuesday evening. The time delay for the availability of the consultation rates from the previous week was therefore only 4–5 days.

The standard time for the availability of final results from cell culture was 11 days (± 3 days), which included 7 days for the initial cell culture, and an additional 4 days for immunofluorescence in positive cases. In total, 1104 samples were available for comparison from the two seasons.

The ‘near patient’ test data were transferred to our laboratory 1–3 days after the performance of the test by the practitioner. The time to transfer of results by fax was not registered, but did not usually exceed 2 days (95% of the results were available at that time). With 2-weekly updates, therefore, the average gain of time in comparison to the conventional system was about 9 days. Given the rapidly changing face of influenza epidemics, this represented a significant improvement for reporting purposes.

Positive rate for the detection of influenza viruses

As ‘near patient’ tests and those using the culture method were not carried out with samples from the same patient, the sensitivity and specificity of the corresponding tests cannot be calculated. Therefore, only the rates of positive samples observed by the two systems for virus detection during the same season can be compared. From an overview of results of the two techniques over the 2-year study period (Table 1), it can be concluded that the cell culture method was more sensitive than the ‘near patient’ test. In the 1999–2000 season, 33% of the samples were positive for influenza by cell culture, whereas in the 2000–2001 season, 23% of the samples were positive. In contrast, the ‘near patient’ test showed positive rates of 26% and 12%, respectively.

Table 1.  Positive rates observed with the different surveillance systems
SeasonTechnique appliedPositive samples/total% Positive samples
1999–2000Culture method167/49933
‘Near patient’ test663/242326
2000–2001Culture method141/60523
‘Near patient’ test197/65912
Both years combinedCulture method308/110428
‘Near patient’ test860/408221

These observations were analyzed in more detail (Table 2). The positive rate of the culture technique during the epidemic (beginning, peak, and end) was compared with that from the ‘near patient’ test. The rate of the culture system varied during the epidemic. It was highest when the peak of the medical consultation was reached, and was 57% in the first season and 38% in the second. This variability of the positive rate is well known, and has been generally observed over the 14 years of influenza surveillance in Switzerland [3]. The number and the percentage of positive cases increased and decreased in parallel with the progression and disappearance of the epidemic.

Table 2.  Positive rate of the two surveillance schemes during the epidemics
SeasonDuration of the
influenza epidemic
(% MC)
Positive rate of the
‘near patient’ test
Positive rate of the
culture method
  1. For the classification of the epidemic, a cut-off level of 1.5% was taken. MC, maximum consultation rate.

1999–2000Beginning (50th): 1.42336
Peak (1st): 7.62757
End (7th): 1.21645
2000–2001Beginning (3rd): 1.41134
Peak (6th): 3.21738
End (9th): 1.31720

The proportion of positive ‘near patient’ tests also changed, but was less dependent on the progression of the epidemic. In the first season, it increased from 23% to 27%, and in the second from 11% to 17%. The increase was less pronounced than for the culture system. This means that a number of positive samples were not detected by the rapid test, although the patients were selected according to the same criteria.


Data collected during the pilot study showed that the ‘near patient’ test used by the practitioners is a useful tool to obtain additional information for the surveillance of the activity of influenza viruses. The test was able to reveal differences in the intensity of epidemics, and results were consistent with those of conventional cell culture. Moreover, results were obtained on average 9 days earlier than with the latter method, with the added advantage that shipping of samples was avoided, which also contributed to the saving of time.

The rapid availability of information about the presence of influenza viruses can be particularly useful in certain situations. For example, when preventive measures have to be initiated rapidly, as in the case of outbreaks in nursing homes, hospitals, or on cruise ships, this information might be crucial. More details about strain variants can be obtained in a second approach.

As shown by data collated by the SSSN, influenza epidemics can vary considerably from one year to another [3]. Such changes were particularly striking over the two seasons included in the present study. In 1999–2000, a variant of influenza A/Moscow/10/99 (H3N2) was dominant, whereas in 2000–2001, the influenza A/New-Caledonia/20/99 (H1N1) was the most frequent virus strain detected in Switzerland. Apparently, between these two seasons, a switch occurred from the subtype H3N2 to H1N1. This change was reflected in much lower weekly medical consultation rates for ILI in the second season (maximum, 7.6% versus 3.2%). The epidemic lasted for 8 weeks in the 1999–2000 season, and for only 5 weeks in the following one. The new variant of influenza A (H1N1) was obviously less pathogenic, as reflected by a lower maximum of medical contacts for ILI. In the second season, the circulating virus caused milder disease, and therefore virus secretion decreased more rapidly. It is obvious that if an antigen detection test is used, the sensitivity of the test is crucial. The positive rate of such a test in a moderate season will be lower than in years with an intensive epidemic, as it fails to identify a certain number of patients secreting low amounts of virus. This phenomenon could be observed between the two seasons. However, for the surveillance of the epidemic, a maximal number of positive cases is of less importance than for a diagnostic application. If the test is able to detect an epidemic as early as possible, it can be used for the surveillance of influenza activity. The rapid test was able to identify and to characterize the two influenza epidemics.

The use of a rapid test for the surveillance of influenza epidemics has other advantages. First, it can be done without a sophisticated laboratory structure, e.g. in developing countries, where surveillance would not otherwise be possible. Second, its use can be integrated into an already existing surveillance scheme, delivering supplementary information. Finally, the use of the ‘near patient’ test makes information rapidly available about the actual epidemiologic situation. As shown in this study, the time gained in comparison with cell culture can be considerable, and was more than 9 days in our case. To our knowledge, this is the first time that such a test was used for surveillance purposes.

Another aim of the SSSN is to inform general practitioners about the ongoing situation of influenza epidemics. This information is important for the prescribing of antivirals, as several clinical studies have shown that treatment must be initiated as early as possible, ideally within 48 h after the onset of the disease. For the practitioner, the clinical diagnosis, together with knowledge of the actual situation, is of considerable help.

Several rapid tests for the detection of influenza A and B antigen have been commercialized [4–8], but have shown limited sensitivity (40–80%). Moreover, the predictive values of such tests were not well characterized. For these reasons, the role of rapid testing for treatment initiation needs to be validated in future studies. In addition, it is important to emphasize that the amount of virus found in different samples (throat swabs, nasal swabs, nasopharyngeal secretions, or nasal washes) varies [9], and the quantity of virus secreted also depends on the time the sample is taken after the onset of symptoms. Without strict control of this parameter, the sensitivity of the antigen detection test may not be sufficient for a diagnostic application, and therefore will not be useful in deciding whether a patient should be treated or not.

Rapid antigen detection tests have another tempting aspect for the general practitioner. Since the results are rapidly available and might give information on whether a particular patient is infected or not, practitioners may also use such tests for diagnostic purposes. As was discussed previously, these tests are not yet sufficiently characterized [4–8], and interpretation of a negative result should be regarded with caution, and further investigations should follow [5]. For a clear diagnosis, several factors have to be carefully controlled: the circumstances of sample-taking, the time after the onset of symptoms, and the specific training of the users of such tests.


This study was a collaboration between Roche Pharma AG, which made available the reagents free of charge, the Swiss Sentinel Surveillance Network (SSSN), and the Swiss National Influenza Center. The study was funded by grants from Roche Pharma AG and the SSSN.