Evaluation of the point-of-care Becton Dickinson Veritor™ Rapid influenza diagnostic test in Kenya, 2013–2014

In many tropical countries, the capacity for influenza virus diagnostic testing in clinical settings is limited. Although many point-of-care rapid influenza diagnostic tests (RIDTs) have been evaluated in temperate settings [1, 2], little information is available on the performance of RIDTs in tropical areas. Compared to temperate developed countries [3], the seasonality of influenza viruses is less predictable in sub-Saharan Africa (SSA). The populations of SSA also tend to be younger [4, 5], with more variable healthcare access and utilization [6], and have a greater prevalence of some chronic co-infections [7, 8]. Influenza has a clear marked seasonality in temperate regions [9‐11]. This is less the case in the tropics where influenza circulation may occur throughout the year. RIDTs have been shown to perform well during the high influenza activity [12, 13]. However there are fewer data pertaining to their performance in clinical settings in tropical Africa where it is possible that RIDTs could facilitate greater awareness of influenza; improve the practical relevance of influenza surveillance in locations where laboratory confirmation may take days to weeks; and possibly inform clinical management and infection prevention and control practices at the clinic or hospital level [14]. The Becton Dickinson (BD) Veritor™ System (Becton, Dickinson and Company, Franklin Lakes, New Jersey) RIDT is intended for use in clinical settings. We implemented this test to detect seasonal influenza virus infections in both outpatient and inpatients reporting to existing respiratory disease surveillance systems in Kenya. We compared the performance of the BD Veritor test to detect influenza A and B infections, against the performance of real time reverse transcription polymerase chain reaction (rRT-PCR) that is used in the surveillance system.

During the study period (July 1, 2013 through July 31, 2014) 3,569 patients were enrolled (median age 2.1 years; range 2 months – 91 years). Of these, 3,199 patients provided NS specimens that were tested by BD RIDT and rRT-PCR; and 3,119 patients provided both NS and NP/OP specimens for testing by BD RIDT and rRT-PCR (Fig. 3). Patient characteristics are summarized in Table 1. The majority (78.5%) of the patients were aged <5 years. The median age and interquartile range for those aged <5 years was 1.4 years (0.72, 2.57) and for those ≥5 years was 10.5 years (6.6, 26.5). Overall, 27.3% (874 out of 3,199) of the patients presented to Nakuru, 22.6% (722 out of 3,199) to Kibera, 18.6% (595 out of 3,199) to Mombasa, 15.9% (509 out of 3,199) to Nyeri and 15.7% (501 out of 3,199) to Lwak. Overall, 64.4% of patients had SARI and 35.6% had ILI. Most (85.2%) patients had specimens collected ≤7 days after the onset of symptoms, of which 346 (12.7%) were collected in the first 48 h. The median time of illness of inpatients was 5.0 days and for outpatients was 3.0 days. Overall, 20% had specimens collected during low influenza periods 27.4% during moderate influenza periods and 52.6% during high influenza periods (Table 1).

In this study we undertook a large point of care evaluation of the BD Veritor™ RIDT in clinical surveillance settings at multiple geographic localities in Kenya. When compared to NS specimens tested by rRT-PCR, the overall sensitivity of the BD RIDT to detect influenza A or B viruses was 77%, and specificity was 95%, similar to the performance of RIDTs in clinical care settings reported in temperate countries [22, 23]. Although RIDTs can never completely replace the need for viral culture to characterize influenza viruses, they may have a role to play in sub-Saharan Africa as they have the potential to reduce laboratory transport and diagnostic costs in routine influenza surveillance. RIDTs also have potential to inform clinicians and public health departments about when influenza viruses are circulating and relative levels of influenza activity.

The per-specimen rRT-PCR cost, including reagent purchases and RNA extractions at the KEMRI laboratory in Kenya (not including specimen collection and transport) is ~ $13.3 USD. This can conservatively be compared to a per-test cost of $10.6 USD of the BD-Veritor system. In the context of many competing priorities and funding challenges, the sustainability of influenza sentinel surveillance is a serious consideration in many African countries [21, 24, 25]. As a result, innovative methods to reduce the cost of surveillance and to generate timely feedback of test results to clinicians have to be considered, perhaps with a known subset of specimens (e.g. those of greater severity, those in severe illness clusters, or those with epidemiologic-linkages to high risk exposures etc.) sent for rRT-PCR and viral culture analyses.

The moderately high sensitivity observed in this study may relate to the fact that 80% of the ILI and SARI case-patients tested were aged <5 years, and RIDTs have been shown to have relative high sensitivity among young children because of increased influenza viral load compared to older children and adults [26, 27]. The sensitivity of RIDTs can be substantially diminished among adults (~18%), [28] and some studies have shown even lower sensitivity among those ≥65 years (8%) [29‐31], leading to underestimation of disease burden [32, 33].

There were no significant differences in the sensitivity and specificity of the BD RIDT compared to rRT-PCR by age, surveillance site, and duration from illness onset to sample collection. However, as expected, PPV was substantially lower during low influenza periods [13, 34]. The PPV was ~30% for specimens tested by the RIDT during low influenza activity (<5% influenza-positive samples), and ~50% when the prevalence of positive samples for influenza was between 5–10%.

While RIDTs have some promise for surveillance purposes, the relatively unpredictable and variable seasonality of influenza in Kenya also presents some challenges to the practical utility of RIDTs [4, 35] for clinical decision-making. In the U.S., empiric antiviral treatment is recommended as soon as possible for patients with confirmed or suspected influenza who are at increased risk for complications from influenza, without waiting for the results of influenza testing. This is possible because negative test results (especially RIDTs) do not exclude a diagnosis of influenza during influenza season [36]. This problem would only be exacerbated in the Kenyan context where there is continuous but variable annual circulation. In low resource settings there may be a desire to base use of antiviral therapy on positive RIDT results; however, the high frequency of false positive results during periods when influenza prevalence is <10% would result in inappropriate use of antiviral treatment for many RIDT positive patients.

Our study had several limitations. Even though our intent was to evaluate the performance of the RIDT in patients of all ages, our study population was primarily children <5 years, with median age of 1.4 years, and therefore our findings cannot be generalized to all adults. Also, these results reflect a single year of influenza virus circulation in Kenya and could vary in other years when influenza activity has different patterns and levels. Although almost half of the cases presented within 72 h of illness onset, most samples (88.6%) were tested >7 days from the time they were collected, and this could have reduced the sensitivities of both the RIDT and rRT-PCR to detect influenza viruses. The prolonged time in the freezer, before testing, could have affected the sensitivity of the rRT-PCR, hence the lower rate of positivity by rRT-PCR when tested >7 days compared to RIDT. However, Caselton, et al. [37], there was no statistically significant difference in influenza positivity of specimens stored up to five days when compared to zero to one day, but they described a reduction in positivity rates after this period. Finally, nasopharyngeal washes typically yield higher viral titers than nasopharyngeal swabs [38], and the BD Veritor ^TM test was only indicated for use in nasal and nasopharyngeal swabs [20] and not oral pharyngeal swabs at the time of our study. Given the improved influenza viral yield from the use of NP/OP relative to NS swabs, future validation of the RIDT assay using NP/OP specimens could be warranted. Indeed, when compared to rRT-PCR undertaken on NP/OP swabs, the use of the RIDT on NS specimens resulted in our missing ~30% of true influenza cases. This was due in part to the reduced performance of the RIDT against rRT-PCR, and in part to the use of NS vs NP/OP swabs.

In conclusion, the BD Veritor^TM System RIDT demonstrated a moderately high sensitivity and high specificity in NS specimens when used predominately in young children at pediatric clinical sites in Kenya. While they are not sufficient on their own for clinical decision-making related to influenza in these contexts, RIDTs may have a role in promoting sustainable and timely influenza surveillance, particularly in remote locations where transportation of specimens to laboratories for rRT-PCR testing is difficult.