Research ArticleCOMPUTATIONAL MODELING

Trauma in silico: Individual-specific mathematical models and virtual clinical populations

See allHide authors and affiliations

Science Translational Medicine  29 Apr 2015:
Vol. 7, Issue 285, pp. 285ra61
DOI: 10.1126/scitranslmed.aaa3636
  • Fig. 1. Schematic of three-compartment mathematical model of human acute inflammation.

    (A) Human trauma model structure. The insult of trauma and pathogen/endotoxin induce systemic inflammation across the three compartments. (B) Progression of inflammation and positive feedback loops in the model after traumatic injury insult. Clinical outputs are shown in red.

  • Fig. 2. Patient-specific model fits.

    (A) Model time-course prediction and data for trauma patient with low insult (ISS = 5). (B) Model prediction and data for trauma patient with intermediate/high insult (ISS = 30). (C) Model prediction and data for trauma patient with severe insult (ISS = 50).

  • Fig. 3. Model-predicted and observed multiple organ dysfunction and ICU LOS.

    (A) Distributions of “Total damage” or susceptibility to multiple organ dysfunction for various levels of traumatic injury severity. (B) Distributions of “Damage recovery time” or ICU LOS for various levels of traumatic injury severity. (C) Predicted mean total LOS as a function of ISS. (D) Verification of model predictions shown in (C), in a separate cohort of 147 blunt trauma patients.

  • Fig. 4. Model-predicted impact of specific inflammatory mediators on simulated ICU LOS and multiple organ dysfunction.

    (A) Predicted responsiveness of “Damage recovery time” (that is, ICU LOS) to simulated degree of cytokine production. (B) Predicted responsiveness of “Total damage” (that is, total LOS) to simulated degree of cytokine production).

  • Fig. 5. Model-predicted and observed IL-6 dynamics.

    (A) Predicted IL-6 AUC as a function of injury severity. (B) Actual IL-6 AUC in a validation cohort of trauma patients. (C) Predicted IL-6 AUC as a function of ISS. (D) Qualitative concordance of IL-6 AUC as a function of ISS for the 147-patient validation cohort.

  • Fig. 6. Model-predicted and observed impact of IL-6 production on clinical outcomes.

    (A to C) Predicted circulating IL-6 concentrations, as a function of propensity to produce IL-6, in virtual blunt trauma patients subjected to low/intermediate (A), intermediate/high (B), or severe (C) injury severity. (D) In a 98-patient validation cohort, plasma IL-6 levels are shown as a function of IL-6 SNP. Plasma IL-6 levels in healthy, 11 noninjured volunteers of similar age and gender as the validation cohort. (E) Propensity to produce IL-6 is predicted by the mathematical model to have no statistically significant association with clinical outcomes. (F) In a validation cohort, blunt trauma patients with high IL-6 SNPs exhibit no statistically significant differences in clinical outcomes as compared to patients with low IL-6 SNPs.

  • Table 1. Summary of simulated 30-day survival rates after simulated blunt trauma.
    Traumatic ISSPercent survivors
    Low/intermediate (5–20)100.0
    ntermediate/high (20–35)99.8
    Severe (35–50)89.7
    Total96.5
  • Table 2. Summary of in silico randomized clinical trials in severe blunt trauma.

    Simulated clinical trials were carried out as described in the Materials and Methods. Percent survival was evaluated at 30 days of simulated time.

    InterventionAverage recovery time (days)Survivors (%)
    Placebo6.0889.7
    IL-1β neutralization6.0189.9*
    IL-6 neutralization5.8690.4*
    TNF-α neutralization8.7579.0*

    *P < 0.005 versus placebo.

    Supplementary Materials

    • www.sciencetranslationalmedicine.org/cgi/content/full/7/285/285ra61/DC1

      Fig. S1. Sensitivity of damage to IL-6 varies with ISS.

      Fig. S2. Model-predicted survival trajectories.

      Table S1. Parameters varied to reproduce patient-level trajectories.

      Table S2. Observed parameter ranges from individual-level fits.

      Table S3. All clinical, demographic, and inflammatory mediator data described in this publication (Excel file).

      Appendix 1. Model equations and parameter values.

    • Supplementary Material for:

      Trauma in silico: Individual-specific mathematical models and virtual clinical populations

      David Brown, Rami A. Namas, Khalid Almahmoud, Akram Zaaqoq, Joydeep Sarkar, Derek A. Barclay, Jinling Yin, Ali Ghuma, Andrew Abboud, Gregory Constantine, Gary Nieman, Ruben Zamora, Steven C. Chang, Timothy R. Billiar, Yoram Vodovotz*

      *Corresponding author. E-mail: vodovotzy{at}upmc.edu

      Published 29 April 2015, Sci. Transl. Med. 7, 285ra61 (2015)
      DOI: 10.1126/scitranslmed.aaa3636

      This PDF file includes:

      • Fig. S1. Sensitivity of damage to IL-6 varies with ISS.
      • Fig. S2. Model-predicted survival trajectories.
      • Table S1. Parameters varied to reproduce patient-level trajectories.
      • Table S2. Observed parameter ranges from individual-level fits.
      • Legend for table S3
      • Appendix 1. Model equations and parameter values.

      [Download PDF]

      Other Supplementary Material for this manuscript includes the following:

      • Table S3. All clinical, demographic, and inflammatory mediator data described in this publication (Excel file).

      [Download Table S3]

    Navigate This Article