ELISA results

Complete sets of paired t0 and t3 sera were available of 781 patients with undifferentiated fever. A subset of 73 serum pairs was tested twice in the two different laboratories (IgGt0 was tested twice in 74 sera). Corresponding filter papers of 161 cases were also tested. The number of specimens tested, t-store, and the proportion of positive test results (index value ≥ 1) are shown in table 1. Based on the first test results, 428 (55%) of patients were classified as "past dengue", 54 (7%) as "acute primary dengue", 218 (28%) as "acute secondary dengue" and 80 (10%) as "no dengue".

Intra-individual variation of IgG, expressed as the correlation between serum IgGt0 and IgGt3 IVs, in cases classified as "past dengue" and "no dengue" was high (R^{2 }= 0.861). In four (0.7%) cases of "past dengue" IgGt0 was negative (index value

Inter-laboratory variation

The inter-laboratory variation was higher. Results of the repeated measurements in the two different laboratories were significantly different for IgGt0 (mean IV at first test: 3.0; at second test: 5.1; mean (95% CI) difference: 2.1 (1.5 to 2.6); p value

Comparison of serum and filter papers

Comparison between filter paper and serum showed that IVs tended to be lower and more variable in reconstituted filter paper than in the corresponding serum (Figure 2). The mean (95% CI) difference with the IV of the corresponding serum was 0.99 (0.67 to 1.31) for IgMt0, 0.82 (0.36 to 1.28) for IgMt3, 0.94 (0.51 to 1.37) for IgGt0 and 0.26 (-0.20 to 0.71) for IgGt3.

Table 1 shows that on t0 the detection of dengue virus IgM from filter paper eluate is not very sensitive, compared to serum. The agreement between filter paper eluates and serum was substantially higher. Of the 89 cases with a positive IgMt3 in serum, the corresponding filter paper eluates were positive in 54 (61%) cases. For dengue virus IgG the agreement was higher on t0 and also increased on t3. Of the 128 sera positive on t3, 113 (88%) corresponding filter paper eluates were positive.

The consequences for the diagnostic classification of cases are shown in Table 2. Differences in classification were in two directions but with filter papers patients were less frequently classified as acute dengue (Chi square 119, kappa value 0.43, p

The mean serum/filter paper IV ratios were 4.84 for IgMt0, 6.93 for IgMt3, 2.61 for IgGt0 and 1.70 for IgGt3.The distribution of the IVs of all filter papers was non-Normal and there was a poor linear relationship with their corresponding serum values. Consequently, no simple conversion factor could be determined and correction of the filter paper IVs by multiplying with the mean serum/filter paper IV-ratios did not improve the agreement between the serum and filter paper diagnostic classifications (overall kappa 0.294).

Effects of storage

The decay of serum Ig during storage was analyzed with mixed effects models, applied to the serum tests which were done two times. Four linear mixed effects models were constructed, with the repeated IgMt0, IgMt3, IgGt0 and IgGt3 index values as the dependent variables and t-store as fixed independent effect. The test identifier, indicating that the two tests were done on different occasions, was entered as cofactor. Also the natural logarithms of the index values were examined. No significant independent effect of the storage time on the index value was observed (mean increase of IV / day (95% CI) for IgM0: -0.0002 (-0.0009 to 0.0005, p = 0.62); IgM3: -0.0004 (-0.0014 to 0.0005, p = 0.39); IgG0: 0.0010 (-0.0009 to 0.0030, p = 0.28); IgG3: 0.0016 (-0.0007 to 0.0039, p = 0.17).

Similar to serum, there was no significant effect of the duration of storing filter papers (Change of ΔIV per day (95% CI) for IgM0: -0.001 (-0.003 to 0.001, p = 0.315); IgM3: -0.0005 (-0.004 to 0.003, p = 0.8); IgG0: 0.001 (-0.002 to 0.004, p = 0.4); IgG3: 0.003 (0.000–0.006, p = 0.052).