Minimizing batch‐to‐batch variability of a live virus vaccine by process analytical technologies

Figures

• 

  FIGURE 1. The decay of Coxsackievirus A21 over time at several different temperature conditions. Production temperature (34°C, N = 8 batches) in black, approximate room temperature (20°C, N = 8 batches) in red, and a refrigerated temperature (8°C, N = 7 batches) in blue. (a) Average viral decay rates are plotted with one standard deviation. (b) Mean product loss over time from calculated decay rates, leveraging Equation (1). At Day 0, 100% of viral titer remains, meaning 0% of viral titer has been lost. Welch's ANOVA test determined a significant difference amongst the means with a p > 0.05. A post‐hoc Dunnett's T3 multiple comparisons test indicates a significant difference comparing the 34°C condition to both 20°C or 8°C (p < 0.001). There was not a significant difference between the decay rate at 20 and 8°C conditions (p = 0.99).

• 

  FIGURE 2. Theoretical model comparison to assay‐measured titer data. N = 25 batches are included in the overall dataset however, the timepoints have different numbers of samples. The timepoints from left to right include N = 25, N = 4, N = 25, N = 24, N = 17, N = 20 and N = 5 datapoints. The horizontal error bars represent the standard deviation of the time at which the sample was taken for the assay. The vertical error bars represent the standard deviation of the infectious titer value—assay or theoretical—respectively. (a) Includes only the timepoint of the sample from the continuous model. At data points with fewer than the total number of batches, only corresponding theoretical model values are represented. (b) Includes the entire continuous model of viral titer with the light blue lines representing the standard deviation over time. A Χ2 test was performed on the two discrete data sets in Figure 2a to determine if the theoretical predicted value is independent of the assay value. The p value of 0.73 supports that the theoretical predicted values are not statistically different from the assay values.

• 

  FIGURE 3. Online PAT parameters and viral titer measurements. (a–c) The blue data represents the theoretical continuous viral titer over the production phase averaged across N = 13 baseline batches. The dotted lines represent one standard deviation of the measurement at each time point. The black data represents average dissolved oxygen, viable cell volume, and oxygen uptake rate parameters across N = 13 batches. Dotted black lines represent one standard deviation of averaged parameters.

• 

  FIGURE 4. Harvest trigger evaluation for baseline condition batches. (a) The timing of peak viral titer for the N = 13 baseline batches is plotted – with one plus sign representing each batch. The mean is represented by the middle line with the standard deviation represented by the horizontal error bars and grayed region. (b) The three PAT harvest triggers are compared to the standard time based harvest–at the “optimal” time (average timing of peak viral titer). Each dot represents one batch with the horizontal lines representing the average and the vertical error bars representing the standard deviation. The four harvest triggers have significantly different standard deviations as assessed through a Brown‐Forsythe test of variance (p = 0.0128).

• 

  FIGURE 5. Case studies of deviant batches. (a) The average timing of peak viral titer is plotted with the error bars representing the standard deviation for each data set. The baseline group of N = 13 batches from Figure 4a (labeled “Baseline”) are plotted in black on the bottom and the gray region represents the range of the standard deviation showing where the timing of peak viral titer would occur for those batches. The other conditions represent N = 2 batches from an experiment investigating how varying cellular density at inoculation impacted cellular growth and subsequent viral progression. (b) Predicted percent yield of maximum titer for each condition is plotted for each harvest trigger in comparison with a time‐based harvest. Horizontal black lines represent the average of each condition. (c) The average timing of peak viral titer is plotted with the error bars representing the standard deviation for each data set. The deviant conditions represent two separate experiments investigating low DO (gassing N2 to force a low oxygen environment) for N = 2 bioreactors and tenfold increased multiplicity of infection (MOI) for N = 2 bioreactors. The same baseline bioreactors as Figure 5a…