Two hundred seventy-eight subjects were immunized, and 253 of these donated blood twice. Twenty-five subjects only donated once (between 10 and 35 days). A total of 15,333 vials were filled with 6.00 ml of serum, and the serum was lyophilized with a coefficient of variation (CV) of fill volume of <1% (gravimetric), moisture content of approximately 1%, and a bioburden below the level of detection. CBER confirmed the moisture content at 0.7% and the sterility of the vial contents.

Quality control of the data submitted for analysis revealed the data to be of high quality, and few were deemed outside QC specifications, leading to minimal reruns. The means of the log concentrations for each serotype were calculated for each laboratory and used to assess the level of agreement among the laboratories. displays scatter plots between all pairs of laboratories with the concordance correlation coefficient (CCC) listed within each plot. The solid diagonal line indicates perfect agreement (where slope=1 and intercept=0). These statistics indicate a high level of agreement, with *r*_{c} exceeding 0.97 for all plots. For the same data, the Pearson correlation coefficient was ≥0.98, and the accuracy coefficent (*C*_{a}) was ≥0.99 in each case. Analysis of variance (ANOVA) models were used to estimate antibody concentrations for each of the serotypes in 007sp, adjusting for laboratory. Variance components necessary for the calculation of confidence intervals were calculated, adjusting for the replicate values within a plate and between plates within a laboratory. Final point estimates and confidence intervals were obtained by back-transforming the estimated log-transformed concentrations and associated 95% confidence intervals (95% CI). These concentrations served as the “assigned” values for each serotype in 007sp and appear in . These values were derived following double adsorption of 007sp with cell wall polysaccharide (CPS) and polysaccharide 22F, and thus in the future, when used as a standard, both standard and unknown sera should be double adsorbed. (Lot 89SF values were derived following single adsorption, and thus the standard and unknown sera are dealt with differently in the current ELISA protocol.) The values assigned to 007sp compared to the original values assigned to 89SF are shown in .

| **Table 1.**Assigned IgG antibody concentrations for 007sp |

IgG ELISA concentrations were measured for a panel of 12 WHO calibration sera using both 89SF and 007sp as the reference standards. presents the assigned values for the 12 WHO calibration sera (*n*=25 for each estimate), while and display the scatter plots and box plots for the first seven serotypes (1, 3, 4, 5, 6A, 6B, and 7F) analyzed, and and show the same information for the remaining six serotypes (9V, 14, 18C, 19A, 19F, and 23F). These plots illustrate the agreement and precision of the five estimated assigned values for 007sp compared to lot 89SF for each WHO calibration serum and serotype.

| **Table 2.**Assigned values for 12 pneumococcal WHO calibration sera as determined using the new pneumococcal standard 007sp |

The scatter plots ( and ) show the high degree of agreement and correlation among the calculated concentrations for the panel of 12 WHO calibration sera using 007sp (vertical scale) versus lot 89SF (horizontal scale) as reference standards. A perfect level of agreement would yield a straight line with slope of 1 and intercept at 0. With rare exception, all data points cluster tightly about this line of identity.

The box plots ( and ) illustrate the deviation of the 007sp-based estimates from those obtained using lot 89SF as reference standard for the 12 WHO calibration sera. These plots offer more resolution than the scatter plots in that they relay more information regarding the deviation of the 007sp and lot 89SF estimates. In these plots, the box is defined by the 25th and 75th percentiles of the distribution; the horizontal line within the box represents the median or 50th percentile, and the asterisk signifies the mean. Vertical lines extend to the most extreme observation that is less than 1.5 times the interquartile range (75th to 25th percentiles), and the diamonds and boxes correspond to assay values which are progressively distant from the mean. The data above the dotted horizontal line indicate the 007sp-based estimates are greater than estimates using lot 89SF as reference standard. On the vertical axis, the number 2 indicates a point where the 007sp-based estimate was twice the lot 89SF-based estimate. A value of indicates the lot 89SF-based estimate was 4 times the 007sp-based estimate. Boxes centered on the horizontal dotted line indicate a good agreement between the 007sp-based and lot 89SF-based estimates. By and large, the concentrations calculated using 007sp as the reference standard are within 2-fold (1/2 to 2.0) of those calculated using lot 89SF as the reference standard. Notable exceptions include serotypes 6B and 19A for lab 5 and serotype 23F for lab 4.

presents accuracy (*C*_{a}), precision (*r*), and concordance (*r*_{c}) measures of agreement between pairs of laboratories and between laboratories and consensus ELISA concentrations for the WHO calibration sera. In order to form paired data between the labs for these comparisons, the five replicate concentration values were replaced by a single predicted value obtained from a mixed-model analysis of variance. There was an exceptionally high degree of agreement, with all values but one ≥0.90. In general, comparisons with lab 3 yielded the lowest measures of correlation and agreement. This is due to one concentration value for lab 3 (type 14) that is substantially greater than the values reported by the other four labs. If this value were removed, all statistics involving lab 3 improve. As an example, the *r*_{c} between labs 3 and 5 of 0.882 improves to 0.947 when this single value is deleted from the analysis.

| **Table 3.**Comparison of ELISA concentrations between laboratories and laboratory-to-consensus assigned values for WHO calibration sera |

In general there is a high degree of agreement between the 007sp-based and lot 89SF-based estimates. This inspires confidence in the validity of the 007sp assignments.