Animal studies provide essential information in the evaluation of novel drugs and vaccines for human infectious diseases, and the robustness of the model impacts the quality and utility of the resulting data. Elevated temperature is a phenotype common to many infectious diseases and is noted as a correlate of pathogenesis 
. The current study applied statistical modeling to advance the use of temperature measurements such that a surrogate marker of death for F. tularensis
infection in mice was defined. Two surrogate endpoints for death based on temperature were identified. The first of these termed CP2 occurred consistently between the febrile (P2) and hypothermic phases (P3) of disease, but could only be defined by linear mixed-effect modeling. Thus, CP2 could not be determined in real-time during the course of an experiment and could not be applied as a surrogate of death. Inspection of temperature curves, however, indicated that CP2 closely correlated to the first temperature measurement at the end of the febrile phase (P2) falling below the mean temperature of the normal phase (P1), and thus was termed “drop point”. It was noted that mice at the drop point were responsive to stimuli and still taking food and water. Independent survival curves generated for F. tularensis
strains with varying virulence and based on CP2 or drop point were statistically indistinguishable from those obtained using observed time to death. The experimental feasibility of determining and using drop point as an accurate indicator of death was also shown through survival curve comparisons that showed no difference between mice infected with F. tularensis
Alb strain MA00-2987 euthanized at drop point and mice previously infected with the same strain for which drop point was calculated after they were allowed to expire. Together these results demonstrate that the easily determined drop point can be applied as an ethical, non-biased and accurate surrogate endpoint for death in mice infected with F. tularensis
. Given that fever is a characteristic symptom of tularemia 
, this approach can be further developed and modified for use in other animal models, and for different inoculation routes and doses. Further, the application of statistical modeling to temperature profiling should be applicable to the study of other acute infectious agents that cause febrile illness.
The data generated through our studies not only demonstrated that the drop point was a statistically valid surrogate endpoint for death, but also revealed that temperature profiling can be used to follow the variability in disease progression between infected animals to a level not previously accomplished. A previous assessment of the virulence of eight diverse F. tularensis
strains in the C57BL/6 mouse model based on time to death reveled that animals succumbed to infection between 82 and 242 h and the variance in time to death for a single strain was between 51 and 129 h 
. The evaluation of body temperature of C57BL/6 mice intradermally infected with the same eight strains of F. tularensis
revealed three temperature phases (normal, febrile and hypothermic) demarcated by two change points (CP1 and CP2) that occurred regardless of the strain. However, as was observed in the variance of time to death 
, mice advanced through the stages of infection as defined by temperature at dramatically different rates. This variability was not only between strains, but there was a notable variance for animals infected with the same F. tularensis
group, with less than 34% congruency with respect to the number of days in which mice were within the same temperature phase. Such variation among mice infected with F. tularensis
is not unique to this study 
. In fact, previous studies show that bacterial burdens analyzed for F. tularensis
infected mice that were euthanized based on a given day post infection differed by as much as 7 Log10
. Differences in disease progression as indicated by our subcutaneous temperature profiles would be consistent with variance in other measures of disease progression. Further, the majority of animal studies performed with F. tularensis
collect and test biological samples at specific intervals (24, 48, 72 h, etc.) 
. Thus, with an acute disease and one that is significantly influenced by the route of infection and minor variations in the infecting inoculums, the practice of collecting biological samples at specific time intervals rather than at defined points of disease likely contributes to the variability in the biological analyses of disease. Temperature monitoring therefore, provides a reproducible picture of infection progression; a feature critical for comparative animal studies within and between laboratories, and one that could be applied to normalize sample collection or biological data.
A non-biased temperature measurement of infection in animal models is advantageous in that it is minimally invasive, requiring the subcutaneous implantation of a probe via injection. A biological sample is not required from the animal and measurements are easily taken without disrupting or causing stress to the animals. Such a system has value in guiding the experimental administration of therapeutics as well as sample collection. As demonstrated by Bast et al. temperature provides a measure of pneumococcal pneumonia severity in mice, and when used in the evaluation of moxifloxacin and levofloxacin efficacy this group demonstrated that animals receiving antibiotics at temperatures ≥32°C fared better than those receiving antibiotics at temperatures of <30°C 
. When combined with appropriate statistical modeling, telemetry based measurements that include other physiological parameters such as blood pressure, biopotential (ECG, EEG, and EMG), heart rate and temperature offer an extraordinary tool to maximize animal resources and normalize experimental data for acute infectious diseases such as that caused by F. tularensis
. Additionally, as telemetry technology advances, automation of this system for use in mice will allow drop point and other measurements identified to be readily used as surrogate endpoints for death.