Please call if we can help

1
RESULTS
BACKGROUND
Methotrexate (MTX), an antifol agent used in the
treatment of various pediatric cancers, ranks
second in oncology agents utilized at The
Childrens Hospital of Philadelphia (CHOP). As a
chemotherapeutic agent, high-dose MTX (HDMTX, gt 1
g/m2) can be life-threatening because it induces
nephrotoxicity at high or prolonged low
exposures. Leucovorin (LVR) is used as a rescue
therapy to manage this. Based on CHOP protocols,
Therapeutic Drug Monitoring (TDM) of MTX is
performed, which aids in pharmacokinetic (PK)
modeling since there is high inter- and intra-
patient variability. MTX Patient Experience
Methotrexate disposition is described using a
two-compartment model with first-order
elimination. Inter-subject variability is
described with an exponential error model and the
residual error is expressed by a proportional
error model. Two clearance distributions exist in
the model for a patient population with normal
renal function and for one with compromised renal
function. Based on a patients MTX plasma
concentrations, a patients is assigned to one of
the two populations. Bayesian prediction of MTX
concentrations is utilized along with the NONMEM
PRIOR subroutine, which incorporates the
population priors into the forecasting model.
Parameter estimates for the historical patients
were CLN (7.49 L/h, 6.37 CV), CLR (2.55 L/h,
75.37 CV), V1 (36 L, 19.03 CV), V2 (3.33 L,
52.15 CV) and Q (.0984 L/h, 12.25 CV).
FIGURE 7 Box and Whisker Representation of the
percent error of PRED with respect to DV against
run number. The 25th and 75th percentiles are
shown within the boxes. Values that fell out of
the range shown for Percent Error were included
in the calculation from the plot.
Figure 5 Relationship between the difference of
PRED and observation and the number of
observations and run number.
OBJECTIVES
The analysis of variance performed in SAS, showed
that both run number and the number of
observations had a P-value lt0.05, making them
statistically significant.
To perform a clinical validation of the
forecasting algorithm used for the MTX dashboard
of the PKB with actual retrospective data
assembled from our Electronic Medical Records
(EMR) system.
TABLE 3 Stability of Priors from Repeated Runs.
(CLN Clearance in patients with normal renal
function. CLR - Clearance in patients with
reduced renal function, V1 Volume of
distribution in central compartment, V2 - Volume
of distribution in peripheral compartment, Q
Inter-compartmental clearance)
Table 2 Tukey studentized range test for the
difference between PRED and observation based on
run number. Only comparisons made with the first
run were shown to be statistically significant.
METHODS
Under IRB approval from CHOP patient data for the
validation was obtained from the EMR system.
Patients were selected so that there would be
variation in age, sex, weight, gender and height.
28 patients visiting CHOP between September 29,
2004 and November 20, 2006 were used as the
validation set in this retrospective study.
FIGURE 1 A timeline representing the protocol
for an inpatient visit at CHOP where HDMTX is
administered.
FIGURE 2 An example of a protocol-specific
nomogram used by physicians for LVR dosing based
on patient plasma MTX concentrations.
Table 1 Demographics of historical and
validation patient populations.
CONCLUSIONS

• Based on this clinical validation, the MTX
  dashboard will be superior to the current
  nomogram-based criterion which now guides MTX
  pharmacotherapy in regards to reducing medication
  errors, earlier detection of nephrotoxicity and
  need for rescue therapy. Some limitations of the
  current analysis
• Unreliability of the EMR system being used as
  gold standard.
• Small sampling of actual patient population
  receiving HDMTX.
• Patients are assumed to be admitted
  simultaneously (naïve analysis).
• Prior to production version of the PKB
• Clustered validation of MTX model.
• Investigation of how often population parameters
  should update.
• Determination if the time LVR therapy is
  suggested to commence by the algorithm
  corresponds to the protocol-defined time.
• The dashboard will be integrated with the CHOP
  EMR systems.

REFERENCES
FIGURE 4 Validation Workflow
Analysis of variance was performed in SAS to
determine the affects of run number and the
nested effect of number of observations on the
difference of population prediction (PRED) and
observation. Statistical significance was
considered at p lt 0.05. A Tukey test was
utilized to compare the means for each run.
Barrett JS, Mondick JT, Narayan M, Vijayakumar K,
Vijayakumar S. Integration of Modeling and
Simulation into Hospital-based Decision Support
Systems Guiding Pediatric Pharmacotherapy. BMC
Medical Informatics and Decision Making 86,
2008.