Linking profilers

Lots of news today.  If you haven't seen this note over at In the Pipeline, please check it out.  Allegedly, GSK China was involved in an epic falsification of data.  I was thinking about what would be required for this to be true: first off, the data integrity within the organization clearly failed.  The implications that raises are fearsome.  Next, the legal group clearly didn't bother to check the data.  Start pulling the threads on the failures, and this begins to look systemic.  Damn.

To quote Monty Python, "Now for something completely different..."

Of the many tricks I love in JMP is its ability to link profilers.  Continuing with the analysis of the formulation data in the last post, a respectable model is derived as demonstrated by the ANOVA results
ANOVA results for model fit to Viscosity

Source	DF	Sum of Squares	Mean Square	F Ratio
Model	13	2802.0550	215.543	34.1649
Error	66	416.3870	6.309	Prob > F
C. Total	79	3218.4420		<.0001*

With both profilers open (Th and viscosity), they can be linked to follow the simultaneous effect upon the outputs as the inputs are changed.
Linking Profilers in JMP

One of the many advantages of linking the two data sets is the results include the confidence intervals.  These provide some value in evaluating the process variations.

From a delivery point of view, the challenge of high concentration formulations is balancing stability and protein concentration against the viscosity.  The greater the viscosity, the greater the probability of discomfort for the patient.  In parallel with the cited paper, the upper limit (UL) to the viscosity is set to 6 cP and the lower limit (LL) of the Th to 50 C.  At the mid-point of the protein concentration and pH range studied, and in the absence of a cation or excipient the viscosity is its the upper limit.
Linked Profilers of Th and Viscosity

From the figures, it's readily apparent that the presence of cations will lower the viscosity but only the presence of Mg2+ (L3) will minimize the impact on the Th.

Impact of Magnesium Ion on the Prediction Profiler

For the pH, the plot shows viscosity is at a minimum at the coded value of 2 (pH 5.5).  Reducing the concentration improves the viscosity only marginally when the 95% confidence interval is taken into account.  If the concentration can be controlled to sufficient precision, the profiler suggests the viscosity could be maintained below its upper limit.  Based upon the results, a concentration, pH, and excipient can be defined to give a reliable formulation.  Furthermore, the intersection of the 95% confidence interval of the profiler with the limits to viscosity and Th defines one approach for establishing the acceptable operating range for the MAb concentration and pH.

Effect of Reduced MAb Concentration in the Presence of Mg2+