Every model system has its advantages and disadvantages. It is up to the researcher not to over interpret their results and make conclusions that are not supported by the data. Even in vivo models have limitations when making conclusions about clinical significance, such as differences in drug metabolism rates, drug diffusion, dosing or biological half-life that may vary in different organisms. One example is convection enhanced delivery (CED) of chemotherapeutic agents to brain tumours. Although this method has potential to be advantageous one has to remember that brain tumours in mice are smaller than in humans, therefore the drug diffusion rate in a mouse might deliver higher concentrations of drug to the periphery of the tumour, where in humans the distance for the drug to diffuse is greater. Direct correlation of results in murine models have to be adjusted for when planning clinical trials or stating conclusions about clinical efficacy.

Every model system has its advantages and disadvantages. It is up to the researcher not to over interpret their results and make conclusions that are not supported by the data. Even in vivo models have limitations when making conclusions about clinical significance, such as differences in drug metabolism rates, drug diffusion, dosing or biological half-life that may vary in different organisms. One example is convection enhanced delivery (CED) of chemotherapeutic agents to brain tumours. Although this method has potential to be advantageous one has to remember that brain tumours in mice are smaller than in humans, therefore the drug diffusion rate in a mouse might deliver higher concentrations of drug to the periphery of the tumour, where in humans the distance for the drug to diffuse is greater. Direct correlation of results in murine models have to be adjusted for when planning clinical trials or stating conclusions about clinical efficacy.

As Pawel said, every model system has its advantages and disavantages. The main problem for me is if working with in vitro system prevents you to think globally. The main goal in cancer research is to increase the therapeutic index.

http://onlinelibrary.wiley.com/doi/10.1002/cam4.91/abstract

From System biology point of view, definitely in vitro experiments look very far from reality as it is almost impossible to produce exact physiological conditions in vitro (e.g. Cytokine milieu which could change altogether receptivity of different receptors). But, still to check various therapy agents it is always cost effective and logical too to try in vitro conditions first and see the efficacy in defined set of conditions and then going ahead and checking efficacy in more complex physiologically relevant systems.

What you are asking is an age-old question for which there exists various answers. The best way to look at this scenario from 'big picture' and 'reductionist' view. Both are necessary to address certain questions. It is only through simple in vitro models, we have learned so much about developing and targeting a pathway or a molecule. Along the way, we have failed, Those failures helped in the long run to correct our pathways. Specific examples can be found in IL-2, IFN-gamma, TGF-beta and other studies. Similarly, drug effect on Tregs, macrophage polarization, T-cell vaccines etc. Now, the very biology of these isolated cells in isolated conditions have provided us with wealth of information that is being used in modern immunotherapy. Thus, those who work using in vitro models understand the limitations and strengths. Hope this answer helps a bit.

Well friends, are we able to recreate truly the micro-environment in vitro that is conductive to real and interactive physiological and pathological life processes at organismal level? Then how far can we depend on the outcome?