This systematic review of human PRNT titers to DENV demonstrated highly variable laboratory methods, particularly among viral strains. Despite analyzing 1,689 reported titers, the effects of using different strains were difficult to ascertain given the heterogeneity in laboratory methods. The original report on the development of a PRNT for DENV by Russell et al. (1967) was referenced frequently in the articles reviewed here , sometimes as the sole description of PRNT methods. However, most articles reported alterations to this protocol due to the identification of techniques or materials enhancing the assay’s performance in their laboratories. A large number of articles did not report information such as neutralization percentage, cell lines, the use of complement, and virus concentrations, making it impossible to fully adjust for these factors. In addition to the diverse array of materials and methods, heterogeneity in PRNT titers can also be induced by volume of virus added, serum dilutions prior to addition of virus, plaque counting techniques such as accounting for plaque overlap, and titer calculations , which were rarely described in the articles reviewed here.
Several additional factors may have influenced the PRNT titers collected here and differences observed between strains after adjusting for other factors in the models. First, the study populations described in these reports were from several different geographical areas (Table 3), which may affect not only strains to which an individual was exposed, but also strains against which their serum was tested. Secondly, there is likely an association between the testing strains and primary or secondary exposure based on each study’s inclusion criteria and laboratory methods. Lastly, the antibody response elicited by different infecting strains may vary in quality, duration and magnitude, highlighting antigenic differences between strains that call for further exploration. For example, Asian dengue strains have been noted to produce different serological reactions than American strains . This may explain why DENV2 strain PR-159 and DENV4 strain Dominica/814669 produced relatively lower titers than Thai references strains.
Strain variation produced differences in median log titers within each serotype. After adjusting for several covariates, the choice of strain accounted for approximately 8% of the variation in titers, while variation between articles, which was used as a proxy for inter-laboratory variation, explained half of the variation in titers, suggesting comparability between laboratories is currently quite poor. The use of reference strains alone will not solve this issue, but the use of reference reagents or proficiency panels would lend validity to each laboratory’s results by offering a means of quality assurance and allow each laboratory to compare “in-house” materials and methods against a standard. Alternatively, algorithms that control for protocol variations would allow laboratories to continue with their preferred materials and methods but would adjust results for comparability.
Most often, PRNT titers are reported as the reciprocal of the serum dilution that shows a 50% reduction in the number of plaques produced by DENV exposure. While most studies use this neutralization level, higher neutralization percentages (e.g., PRNT60) have been used to increase specificity and prevent the detection of cross-neutralizing antibodies , but this must be weighed against decreased sensitivity . Upon inclusion of neutralization percentage in the hierarchical model, an increase in titer was observed with each 10% increase in neutralization percentage. While we would expect a decrease in titer with higher neutralization percentage (e.g., 1:80 in PRNT50 versus 1:10 in PRNT60), this relationship may have been highly influenced by the few low titers reported in studies using higher neutralization percentages (e.g. PRNT90) and the many low titers reported in studies using PRNT50.
As titers were abstracted directly from the literature, our results are limited by published data. Titers are more likely to be reported and published if the findings are unusual or significantly different from previous reports. Also, relatively few titers (n = 177) were reported from unexposed individuals. Prior exposure to DENV was determined based on authors’ report, which may result in misclassification, and the primary exposures were unknown among individuals with secondary exposures. Original antigenic sin and variability in cross-reactive responses may have large effects on neutralization responses that are impossible for us to quantify without knowledge of primary exposures . Furthermore, titers from individuals exposed to other flaviviruses were not included in this review. The exposure history of individuals to other flaviviruses was unknown in our sample and cross-reactions with non-dengue flaviviruses may be an additional factor to consider when analyzing PRNT data. The data were inadequate to estimate secondary exposure-test serotype pair effects in addition to effects for individual strains or articles. Lastly, the inclusion of reports published in English, Spanish and Portuguese reduced the potential for geographic selection bias, but may have excluded influential Asian studies.