A systematic review of the validity of patient derived xenograft (PDX) models: the implications for translational research and personalised medicine

Abstract

Patient-derived xenograft (PDX) models are increasingly being used in oncology drug development because they offer greater predictive value than traditional cell line models. Using novel tools to critique model validity and reliability we performed a systematic review to identify all original publications describing the derivation of PDX models of colon, prostate, breast and lung cancer. Validity was defined as the ability to recapitulate the disease of interest. The study protocol was registered with the Collaborative Approach to Meta-Analysis and Review of Animal Data from Experimental Studies (CAMARADES). Searches were performed in Embase, MEDLINE and Pubmed up to July 2017. A narrative data synthesis was performed. We identified 105 studies of model validations; 29 for breast, 29 for colon, 25 for lung, 23 for prostate and 4 for multiple tissues. 133 studies were excluded because they did not perform any validation experiments despite deriving a PDX. Only one study reported following the ARRIVE guidelines; developed to improve the standard of reporting for animal experimentation. Remarkably, half of all breast (52%) and prostate (50%) studies were judged to have high concern, in contrast to 16% of colon and 28% of lung studies. The validation criteria that most commonly failed (evidence to the contrary) were: tissue of origin not proven and histology of the xenograft not comparable to the parental tumour. Overall, most studies were categorized as unclear because one or more validation conditions were not reported, or researchers failed to provide data for a proportion of their models. For example, failure to demonstrate tissue of origin, response to standard of care agents and to exclude development of lymphoma. Validation tools have the potential to improve reproducibility, reduce waste in research and increase the success of translational studies.

Keywords: Cancer models; PDX models; Personalised medicine; Systematic review; Validity tools.

Figure 1. PRISMA flow diagram of the study selection process.

Figure 2. Overall validity ratings of PDX models.

Each study was assessed to determine if the reported PDX models were at high risk of concern for model validity. The graph indicates the percentage of studies per tissue: which had no concerns (white bars), high levels of concern (black bars), unclear levels (grey bars). 5% of prostate studies were not validated because they failed to derive a PDX.

Figure 3. Individual validity ratings of PDX models.

Each study was assessed to determine the level of concern for each signalling question eight criterion: a. was tissue of origin proven? b. confirmation that the PDX was derived from a given patient, c. was the cell lineage proven? d. confirmation that the PDX was derived from tumour and not normal cells, e. absence of murine overgrowth, f. was there comparable histopathology? g. concordance for standard of care agents? h. was the absence of lymphoma proven? A full description of signalling questions 8a–8h can also be found in Table S3 of the model validation tool. Each graph indicates the percentage of studies that were judged to be of low concern (white bars), high concern (black bars), unclear concern (grey bars). (A) Breast (n = 29 studies). (B) Colon (n = 31 studies). (C) Lung (n = 25 studies). (D) Prostate (n = 20 studies).

Figure 4. The proportion of PDX models from included studies that were validated.

Each study was assessed to determine if all reported PDX models were validated (any question 8 criterion). For each tissue type the percentage of studies is reported as: all models were validated (white bars), did not validate all reported models (black bars) or insufficient details reported to determine if all models were validated/not applicable (grey bars).