There is actually a pretty sizable debate going on

There is actually a pretty sizable debate going on in the research community about the validity of the xenografting process. Some scientists say it is so highly specific that lack of diversity/variability skew hypothetical application of results to humans - because the Hs/Mu mice have been genetically refined for decades and are too 'same'.. Others point to the fact that, once the grafted human tumor starts to grow, it gains many mouse cells and becomes a hybrid human/mouse tumor (cool and grisly, huh?) and this muddies ability to quantify drug agents effect on the human tumor cells, as the effect could be more mediated by the mice cells in the tumor.

excerpt: "A current limitation is that gene profiling products cannot be used reliably to characterize human cells grown in chimeric mice because of the similarities between the human and mouse genes...the question of whether human tumor data obtained from a xenograft model is impacted by gene expression changes within the mouse" and "Efforts to improve mouse models are underway by the introduction of human genes but these models currently remain a far cry from humans." http://www.bhbio.com/assets/pdf/XenoQ-Whitepaper.pdf).

And sometimes the xenografts from human tumors, when implanted on these immunosuppressed mice, have developed disease carried by the human donor that never flared in the human, like Epstein Barr virus. (http://www.plosone.org/artic/info%3Adoi%2F10.1371%2Fjournal.pone.0039294)

Researchers have been refining these lab mice for decades - creating human-mouse genetic chimeras - frankenmice - and they're still refining - now trying to bring mousemodels up to level of newest gene therapies in the biotech world. It's actually kind of fascinating - and if you want a good geek-read then read the bhbio.com paper - you'll be as up to speed on mouse xenographs as any researcher.

Here is another excerpt from that paper: "Drugs tend to fail late in clinical testing trials which is the most expensive part of the process, with investments of $500 million or more accumulating towards the end stages22. Only about 11% of potential anticancer drugs that yield promising data in mice are approved for use in humans46. This high rate of ineffective compounds that enter into the clinical testing phase indicates a need for predicting efficacy prior to clinical testing in humans. Better animal models and methods for analyzing them have the potential to increase success at the early pre-clinical stage. Towards this endeavor, the molecular profiling of Hs/Mu chimeras has the potential to accelerate the discovery of cancer targets, biomarkers and therapeutics; however, the current limitation is the lack of products designed to adequately discriminate between human and mouse genes."

But these techniques are highly applicable to human response. The technology is scary sci-fi stuff. Researchers just keep perfecting and if you took a culture of a wild mouse cell and compared it to a cell culture from the Hs/Mu chimera mouse (the one used for most xenografting), they would not immediately look like the same species...Bottom line is they're the best we have, unbelievably refined technology, not perfect and never satisfactory to the researchers...but pretty damn good.

This is a good read too: http://www.biomedcentral.com/1472-6750/11/124