Impact of inherent variability and experimental parameters on the reliability of small animal PET data

Abstract

Background: Noninvasive preclinical imaging methodologies such as small animal positron emission tomography (PET) allow the repeated measurement of the same subject which is generally assumed to reduce the variability of the experimental outcome parameter and to produce more robust results. In this study, the variability of tracer uptake in the rodent brain was assessed within and between subjects using the established radiopharmaceuticals 18F-FDG and 18F-fallypride. Moreover, experimental factors with potential impact on study outcome were elicited, and the effect of their strict homogenization was assessed.

Methods: Brain standardized uptake values of rodents were compared between three PET scans of the same animal and scans of different individuals. 18F-FDG ex vivo tissue sampling was performed under variation of the following experimental parameters: gender, age, cage occupancy, anesthetic protocol, environmental temperature during uptake phase, and tracer formulation.

Results: No significant difference of variability in 18F-FDG or 18F-fallypride brain or striatal uptake was identified between scans of the same and scans of different animals (COV = 14 ± 7% vs. 21 ± 10% for 18F-FDG). 18F-FDG brain uptake was robust regarding a variety of experimental parameters; only anesthetic protocols showed a significant impact. In contrast to a heterogenization approach, homogenization of groups produced more false positive effects in 18F-FDG organ distribution showing a false positive rate of 9% vs. 6%.

Conclusions: Repeated measurements of the same animal may not reduce data variability compared with measurements on different animals. Controlled heterogenization of test groups with regard to experimental parameters is advisable as it decreases the generation of false positive results and thus increases external validity of study outcome.

Figure 1

Experimental design of homogenization vs. heterogenization. The scheme of the experiment is in analogy to the study of Richter et al. [7]; 8 × 8 animals underwent ¹⁸F-FDG scans according to PET protocol 1. Varying experimental parameters were gender, age, and cage occupancy of the animals (n = 2 or n = 13 per cage). Animal groups were divided according to gender. This setup allowed for four strictly homogenized test group comparisons with the only parameter varying being gender (one example of such a comparison pair is indicated in dark gray). On the other hand, by reordering the same data into 8 × 8 randomized groups, four comparisons of such pseudo-heterogenized groups were performed (indicated by the light grey squares).

Figure 2

Different variability types. (A) Different types of variability of ¹⁸F-FDG brain SUV in male Crl:CD(SD) rats (n = 6). Each data point corresponds to the ¹⁸F-FDG brain SUV of one individual animal. Intra-animal variability is the variability between three brain scans of one individual (indicated in red). (B) Column bar representation of the data from (A). Inter-animal variability is the variability between brain SUVs of different individuals acquired on the same test day (represented by the SD, indicated in blue). Inter-study variability is the variability of ¹⁸F-FDG brain SUVs between two independent test groups acquired under exactly the same experimental conditions (indicated in green); *P < 0.05.

Figure 3

¹⁸F-fallypride SUV in striatum, retina, and cerebellum. Black, striatum; red, retina; blue, cerebellum. NMRI mice (light colors): n = 6, C57Bl/6 J mice (dark colors): n = 7; ***P < 0.001.