The clinical diagnosis of influenza is often challenging, especially in young children. Thus, pediatricians often use in-office rapid influenza diagnostic tests (RIDTs) to confirm clinical suspicions of influenza. The positive and negative predictive values of these tests (ie, the accuracy of positive and negative tests)—and hence the tests’ usefulness—depend on the sensitivity and specificity of the test being used. However, these predictive values also depend on the frequency of influenza in the community; thus, the usefulness of RIDTs varies with the surges and peaks of disease activity over the course of a typical flu season.1,2 In a recent review, we suggested the usual circumstances under which the use of RIDTs is most likely to be helpful clinically.1
This year, however, is not at all “usual” for influenza. A novel strain of influenza A (H1N1) virus has been circulating worldwide, and disease activity has risen sharply in many parts of the United States well in advance of the normal flu season. At first glance, this would seem to be a year in which RIDTs might be very useful in monitoring influenza activity and making clinical decisions about antiviral treatment and prophylaxis. Unfortunately, however, RIDTs have not been as reliable in detecting the pandemic H1N1 influenza virus.3 In a recent CDC study, the sensitivity of one RIDT was only 47% for detection of the pandemic H1N1 influenza virus, and its specificity in this setting was only 86%.4 The sensitivity and specificity of this same RIDT for detection of seasonal influenza virus have been reported as 92.5% and 100%, respectively.5
Here, in a brief addendum to our earlier article,1 we examine how the usefulness of an RIDT that has the sensitivities and specificities mentioned above differs when it is used for the detection of pandemic H1N1 influenza as compared with its use for detection of seasonal influenza.
We do this by calculating the positive and negative predictive values of the RIDT—for both pandemic H1N1 flu and seasonal flu—at several points during a hypothetical influenza season.
CALCULATION OF PREDICTIVE VALUES OF RIDTs
Definitions. The positive predictive value is the proportion of patients with a positive rapid test who actually have influenza, and the negative predictive value is the proportion of patients with a negative rapid test who do not have influenza.
Simulation of a typical flu season. This year’s flu season may be highly atypical in many respects; nevertheless, we estimated disease activity at 5 different times over the course of a typical flu season, using data from a CDC review of national influenza activity during 6 recent winters.6
These 5 points were:
•12 weeks before peak.
•6 weeks before peak.
•6 weeks after peak.
•12 weeks after peak.
In addition, we calculated the positive and negative predictive values at a peak prevalence of 50% because there have been concerns that the spread of disease activity this year may be particularly brisk, with a large surge in cases.
Results. The positive and negative predictive values for an RIDT with a sensitivity of 92.5% and a specificity of 100% for detection of seasonal flu are displayed in Table 1. Because the specificity is 100%, the positive predictive value remains perfect at all levels of disease activity. The negative predictive value drops slightly during peak season but remains very high. When these calculations are repeated using an extremely high prevalence of 50%, the positive predictive value remains 100% while the negative predictive value decreases to 93.0% (Table 2).
Similar calculations were then performed for the same RIDT when used to detect pandemic H1N1 flu. We used the sensitivity of 47% and specificity of 86% that have been reported for this test for the detection of the pandemic H1N1 virus (Table 3). Under these conditions, the positive predictive values are much lower, especially early and late in the season, and this value only exceeds 50% at the peak of the season. The negative predictive values are considerably higher but drop to 82.2% during the peak season. When these calculations are repeated using an extremely high prevalence of 50%, the positive predictive value increases to 77.0% but the negative predictive value decreases to 61.9% (see Table 2).
Discussion. This analysis illuminates 2 important points. First, the predictive values of RIDTs—for both seasonal flu and pandemic H1N1 influenza— will vary over the course of the flu season depending on the level of disease activity in the community. The positive predictive value of an RIDT for either strain of flu can be very low at off-peak times, such as very early in the season, because the test is not perfectly specific and the disease frequency is low. Likewise, the negative predictive value can be low during a surge of clinical influenza cases because the test is not perfectly sensitive and the disease frequency is very high.
Pediatricians generally know when flu season has arrived because they are missing meals and the office telephone is constantly ringing. They can confirm these clinical suspicions, however, by communicating with their local health department or consulting the CDC’s influenza Web site (http://www.cdc.gov/flu).
The second important point this analysis serves to underscore is that if the predominant circulating influenza strain is the pandemic H1N1 flu, RIDTs will not be nearly as useful as they have been in recent flu seasons. The CDC has therefore urged clinicians to be especially cautious this year in their use of RIDTs.7 In those few circumstances in which laboratory confirmation of novel H1N1 influenza cases is desired, a real-time reverse transcriptase polymerase chain reaction (RT-PCR) test is recommended. Unfortunately, such testing is much more expensive and has a longer turnaround time than RIDTs.
WHAT THIS MEANS FOR YOUR PRACTICE
So far in the current US flu season, almost all of the influenza isolates have been novel H1N1 viruses. On the basis of what has been happening in Australia, it seems possible— and even likely—that this situation may persist through the entire flu season: in Australia, the pandemic strain has almost completely replaced the current seasonal H1N1 virus over the past couple of months.
The reported accuracy of RIDTs for the identification of the 2009 pandemic H1N1 virus is much lower than it has been for identification of seasonal influenza. Thus, if the 2009 pandemic H1N1 strain remains the predominant strain in the community (which seems likely), a positive test result early or late in the flu season is especially likely to be a false positive; similarly, a negative test during the peak of the season will often be a false negative. If the mix of causative viruses changes and a greater proportion of isolates are seasonal influenza viruses, then the predictive values of RIDTs may increase and these tests may become more clinically useful. For now, pediatricians are best advised to rely on clinical judgment in making the diagnosis of influenza.
If RIDTs that are more accurate in detecting novel H1N1 influenza become available, these can then be used to complement clinical judgment in evaluating children with influenzalike illness. In the meantime, if definitive laboratory confirmation of H1N1 virus infection is desired, real-time RT-PCR testing is recommended.