Using an in-country reference standard of culture and/or multiplex PCR, we show that dipstick RDT sensitivity at 70% (95%CI 55–82) was lower and specificity at 97% (95%CI 93–99) was substantially higher than found in an earlier dipstick RDT field evaluation using the same reference standard [89% (95%CI 80–95) and 62% (95%CI 48–75) respectively] 
. The dipstick RDT sensitivity found in this study was also lower than reported from prior evaluations conducted using ideal laboratory conditions (89%; 95%CI 84–93 
and 94%; 95%CI 92–96 
). Specificity, however, was comparable to that found by these studies [94% (95%CI 92–96) 
and 97% (95%CI 94–99) 
Variation in dipstick RDT results may be explained by differences in the test quality, which could be due either to batch-to-batch performance variations or poor stability of the test. The latter should be considered especially as dipstick RDT kits were stored at room temperature, which reached 40°C during the study. Such variations have been largely documented for other rapid diagnosis tests, e.g. the malaria RDT using similar mechanisms 
. Further development of the dipstick RDT should ensure better and more reproducible results, although the mass production of this test in the future is not yet assured 
. Importantly, these varying results (especially the poor sensitivity shown in the current study) indicate that this dipstick RDT should not be used at bedside for individual patient diagnosis. Conditional on sufficient batch-to-batch stability, this RDT may be useful to declare an outbreak, and to guide vaccination choice.
The variations in performance described here suggest that other rapid tests, such as the latex agglutination tests, are preferable for use at a peripheral laboratory level, once the necessary equipment is available. The latter is important, as earlier research emphasises that the latex agglutination tests must be used according to manufacturers' recommendations in order to achieve optimum sensitivity, requiring a certain level of human resources and infrastructure 
Although culture is known to have a low sensitivity, especially in cases where antibiotics have been administered to patients or if the sample collection and/or transportation are sub-optimal 
, this technique remains indispensable to determine the antibiotic sensitivity profile and genetically characterize the outbreak strain. PCR is of added value due to its greater yield of valid results in remote areas, and its capacity to detect and serogroup other agents of bacterial meningitis, namely pneumococci, Haemophilus influenzae
b and N. meningitidis
serogroup X 
. PCR is generally considered to be more sensitive than culture 
, since it can detect DNA of few dead or alive bacteria. However, previous studies in sub-Saharan Africa have shown comparable sensitivity and specificity of culture and PCR if performed on fresh CSF from untreated patients 
. This study, therefore, like previous evaluations of rapid tests for N. meningitidis 
, used a reference standard which combined the locally available techniques of culture and mutliplex PCR. A highly sensitive nested PCR technique was added (also in combination with in-country culture) as an absolute reference. In a nested PCR, the initial PCR reaction is performed on the extract and the product of this first reaction is then used as a template for a second PCR reaction. While a reference standard including this technique was thus expected to have a better sensitivity than one incorporating multiplex PCR (which involved two multiplex and one specific PCR step), the magnitude of this difference was surprising. As the laboratory performing the multiplex PCR in this study was a high-quality national reference laboratory, this difference could be even higher if it had been performed in regional or district laboratories. Comparative analysis of the two PCR techniques (data not shown) revealed that of 118 samples positive for N. Meningitidis
serogroup A by the more sensitive nested PCR technique, only 37 (31%) were positive by multiplex PCR (there were no additional samples positive by multiplex PCR, i.e. all samples negative by the nested technique were also negative by multiplex). Incorporating culture with the PCR results gave a similar result (48 culture-with-multiplex PCR out of 126 culture-with-nested PCR positive; 38%).
Although multiplex PCR is the current standard for bacterial meningitis surveillance in the meningitis belt, the more sensitive nested PCR technique would provide more accurate information for evaluation of a new diagnostic test. However, the greater risk of contamination during the nested PCR is much higher, so this technique may not be feasible as part of an in-country reference standard, although samples could be transported to a supranational reference laboratory certified for this procedure, as was done in our study. An alternative in-country method could be to use the PorA gene directly to indicate presence of N. meningitidis, followed by genogrouping (but should initially be compared with multiplex PCR).
A comparison study of PCR methods between several European laboratories has shown that the primers used for PCR identification of N. meningitidis
give equivalent results 
. However, this comparison study also demonstrated that results vary between laboratories, with one laboratory having a sensitivity of only 55% compared with the consensus results of all laboratories 
. This underlines the importance of the type of PCR technique used for the diagnosis of bacterial meningitis, and the potential for variation between laboratories.
Apart from these technical considerations, the choice of laboratory method ultimately depends on how the results are being used; these may be different if used for (a) clinical diagnosis at an individual level (bedside), (b) surveillance to detect outbreaks and to guide vaccine choice, (c) estimating burden of disease during or outside epidemics, (d) surveillance to describe the etiological range of bacterial meningitis, or (e) surveillance to evaluate vaccine impact and strain replacement. This evaluation shows the value of the dipstick RDT for use in (b).
The dipstick RDT is clearly easier to use and to store, and requires less training for its users, than any of the prior rapid tests (such as Pastorex® or Slidex®) or the two other tests used in-country for the reference standard, culture and multiplex PCR. This advantage for outbreak investigation in remote settings should be carefully weighed against limitations in performance identified in the present study. To limit potential harm by inappropriate treatment decision following a negative dipstick RDT result, it may be necessary to restrict its use to staff not involved in routine patient care. Further evaluation of the field performance of the dipstick RDT in identifying other serogroups is needed, in particular serogroup W135 (for which a different vaccine is required). In this analysis, we treated two dipstick RDT results positive for N. meningitis W135, as negative (i.e. negative for the outbreak strain). Interestingly, one of these samples was positive for N. meningitidis A by both PCR techniques, while the other sample was negative by multiplex PCR (there was insufficient CSF for nested PCR to be performed). This highlights the need to investigate the dipstick RDT further for potentially misleading cross-reactivities.
Finally, although results from evaluations performed in well-resourced, air-conditioned laboratories with fully trained personnel may well be excellent indicators of a new test's theoretical optimal performance, they cannot and should not be assumed to apply equally to the field without thorough ‘operational’ testing, especially when they are to be used in the hot, dry, dusty areas with limited resources such as the African meningitis belt. Our studies illustrate that there is no substitute for ‘real’ field testing in diagnostics, and prove the value of operational research.