The RDTs varied according to detection mechanism, time to results, storage temperature, and shelf life. Of the 6 RDTs, 3 were designed to detect influenza A and B viruses, 2 influenza A virus only, and 1 specifically H7 virus (test names and manufacturer information provided in Table 1). We followed manufacturers’ instructions and visually read the results. At the time of the study, 5 of the 6 tests had been approved for detection of seasonal influenza viruses in China, and approval was still pending for the Wondfo H7 test for A(H7N9). Since then, the Wondfo H7 test has been approved by the China Food and Drug Administration.

To evaluate detection limits of the RDTs, we propagated vaccine candidate A(H7N9) virus strain A/Anhui/1/2013 in MDCK cells and determined the mean 50% tissue culture infectious dose (TCID₅₀) per milliliter on the basis of at least 3 independent assays. Viruses were standardized to 1 × 10⁷ TCID₅₀/mL and serially diluted 10-fold in phosphate-buffered saline. The detection limit for 3 RDTs was 10³ TCID₅₀/mL and for 2 RDTs was 10⁴ TCID₅₀/mL; 1 RDT could not detect A(H7N9) virus. The following 3 RDTs with the highest sensitivity were chosen for further evaluation of A(H7N9) in clinical specimens: Wantai FluA, Wondfo FluA, and Wondfo H7 (Table 1). The specimens tested were throat swab or sputum (including tracheal aspirates) collected from patients with suspected A(H7N9) virus infection since late March 2013, confirmed by rRT-PCR with primers and probes described previously (1,3), and stored at −80°C.

To compare the efficiency of RDTs for detecting A(H7N9) virus and seasonal influenza A viruses, we also used RDTs and rRT-PCR to test seasonal influenza A(H3N2)–positive and A(H1N1)pdm09-positive throat swab samples collected during January–April 2012. rRT-PCR testing for seasonal influenza virus was conducted according to the World Health Organization protocol (5).

In total, 110 throat swab or sputum specimens from 53 A(H7N9)-infected patients and 115 A(H3N2) and 97 A(H1N1)pdm09 throat swab specimens were tested by using the 3 selected RDTs and rRT-PCR; each specimen was prepared and tested by all 4 assays at the same time. As cycle threshold (C_t) values increased, indicating lower levels of influenza virus in the clinical samples, the sensitivity of RDTs decreased significantly (Table 2). Viral load in throat swab specimens from A(H7N9)-infected patients was significantly lower than that from A(H1N1)pdm09- and A(H3N2)-infected patients (Figure 1).

We then further compared the sensitivity of RDTs for detecting virus in A(H7N9) specimens and seasonal influenza virus specimens with the same influenza A matrix gene C_t intervals. We found that for specimens with C_t <25, RDT sensitivity for A(H7N9) specimens and seasonal influenza virus specimens was similar. However, for specimens with C_t >25, RDT sensitivity was significantly lower when A(H7N9) specimens were compared with seasonal influenza virus specimens with the same C_t interval. Overall, RDT sensitivity for detecting A(H7N9) virus was significantly lower than that for detecting A(H1N1)pdm09 or A(H3N2) viruses (p<0.01). Wantai Flu A and Wondfo Flu A detection of A(H1N1)pdm09 and A(H3N2) viruses did not differ significantly (p>0.05). According to the Wondfo H7 subtype colloidal gold kit, 56 (51%) of the 110 A(H7N9) samples were positive and all 212 A(H1N1)pdm09 (n = 97) and A(H3N2) (n = 115) samples were negative (Table 2), demonstrating that this RDT can distinguish between clinical specimens positive for A(H7N9) and seasonal influenza viruses and that its rate of positivity for detecting A(H7N9) viruses is higher than that of the other 2 RDTs tested (Table 2). Ten throat swab samples that were influenza virus negative by rRT-PCR were also negative by the 3 RDTs.

Considering that most A(H7N9) virus–infected patients had pneumonia and that the virus replicates more efficiently in the lower respiratory tract than in the upper respiratory tract (6,7), A(H7N9) viral loads are probably higher in specimens from the lower respiratory tract. Viral loads were significantly higher in sputum/tracheal aspirates than in throat swab samples collected at the same time (Figure 2).

Although most RDTs examined in this study detected not only seasonal influenza virus but also A(H7N9) virus, the sensitivity of RDTs was lower for A(H7N9) virus than for seasonal influenza virus. Even for specimens with the same C_t intervals, RDT sensitivity to A(H7N9) virus was significantly lower than that for either A(H1N1)pdm09 or A(H3N2) virus. The most likely explanation is that cross-reactivity with the nucleocapsid protein–specific antibodies used in RDTs to detect seasonal influenza A virus was significantly lower for A(H7N9) virus. A previous study also indicated that detection sensitivity for swine-origin A(H1N1) viruses varies widely among seasonal influenza A virus RDTs; some tests are unsuitable for detecting several subtypes of avian influenza viruses because of low sensitivity (8).

The Wondfo H7 RDT evaluated in this study was based on a pair of anti-H7 monoclonal antibodies. We found that for each of the 3 C_t intervals, the sensitivity for detecting A(H7N9) was relatively higher for the subtype H7 RDT than for the other RDTs.

Our study indicates that throat swab samples, which have been widely used for influenza diagnosis in China, are not suitable for RDT detection of A(H7N9) virus because of the low levels of virus they contain (Figure 1). Viral loads are significantly higher in sputum samples/tracheal aspirates from the lower respiratory tract than from throat swab samples (Figure 2). If any previously designed influenza A virus–specific RDTs are to be used for detection of A(H7N9) viruses, the kits should be modified for use with sputum and tracheal aspirates by improving extraction. In summary, usefulness of currently available seasonal influenza RDTs for diagnosing A(H7N9) virus infections is limited because of their low sensitivity for detecting virus in upper respiratory tract specimens.