The RAINBOW workshop addressed the integration of in vivo, in vitro and computer based methods (in silico) to study toxic properties of chemicals. About 50 participants from 14 countries attended the workshop, organised within the EC funded project RAINBOW.




There is an urgent need for test systems that can fill the enormous data gap for untested or insufficiently tested substances as efficiently as possible. The recent approval of the REACH regulation by the European Parliament is an important step towards solving this society request. An efficient test strategy must take into account the limitations in economic resources and testing capacity. It also has to be in line with the aim to reduce the use of animals in toxicological testing. By the year 2009 the 7th amendment to the Cosmetics Directive prohibits the use of animals for pre-marketing toxicity testing for cosmetic products in the in the EU (Directive 2003/15/EC, Official Journal L66:26-35).




Individual toxicological and ecotoxicological tests can be described in terms of their cost, validity, reliability, and sensitivity, and there is no such thing as the perfect test. If we had, for all important endpoints, tests that fulfil the criteria of low cost, sufficient sensitivity and high validity and reliability, then the scientific uncertainties inherent in testing and risk assessment could be substantially reduced. In reality every test is a trade-off between these requirements. Since every test represents a trade-off between these aspects, we face the challenge to combine tests with different strengths and weaknesses into scientifically well-founded and resource-efficient test systems in which the tests compensate for each other's weaknesses as far as possible. The integration of the various possibilities offered by in vivo, in vitro and in silico methods is the most mature solution to the knowledge gap which is huge; integration can take advantage of the possibilities of each approach.

In regulatory applications, toxicological tests are combined into test systems. A test system contains rules for when and in what order the different tests should be performed. With the resources presently available it will be necessary to use tiered systems in which relatively simple tests are performed for all chemicals that are up for assessment, and the outcomes of these simple tests are used to prioritize substances for further, more resource-intensive testing. Furthermore, efforts toward integration should address a more flexible scheme beyond the classical tiered approach, in which in silico screening is the first step, followed by in vitro and finally in vivo. A feed-back mechanism may offer advantages, and multiple parallel inputs may increase the understanding.

In the future a large number of data, from in vivo, in vitro and in silico studies, will became available. A strategy to integrate the different methods has to keep into account the evolving situation.




The discussion at the workshop addressed the barriers to a more efficient integration and the needs to solve current limitations.
These are some of the problems identified:

· Language: experts in different fields use terms which are typical of a given field, and in some cases the same term has different meanings in the different fields.

· Concepts: there are cultural barriers related to concepts typical of one methodology.

· Formalism: some common practices should be codified in a clear sequence of steps.

· Subjectivism: in some cases the human expert covers a central role, and the final decision is based on the individual experience, which however is not clearly codified.




The discussion identified areas to be evaluated in order to improve the integration between in vivo, in vitro and in silico methods. Some of these efforts should clarify certain aspects of individual methods, allowing a better dialogue between the methods; other efforts are required to better plan the overall integrating scheme.

· To qualify and quantify uncertainties. Each method (in vivo, in vitro and in silico) should produce a result (for instance a toxicity value or class) with a given defined uncertainty. If such an uncertainty is not defined it will be very difficult to use the result of a given method for the final chemical assessment. Also the current practice is not satisfactory in this aspect.

· To improve transparency.

· To improve standardisation, achieving a common metric.

· To evaluate advantages and disadvantages of each method, in vivo, in vitro and in silico. It has not been possible, within the workshop, to explicitly list a series of features characterising the different approaches. Some common features apply to the same category of methods, but this is not always the case. Pros and cons of individual methods are related to the target of the integrated system, which may be different.

· In order to determine the predictive value of any test, we should know how results obtained from this test relate to effects on the target system, i.e. humans, and the ecosystems that the regulation intends to protect. This, however, is seldom possible. At best, we can compare the simple test to a more advanced one, and even this is a far from perfect approximation. Even a state-of-the-art test, such as a long-term animal test, provides in its turn only an estimate of effects in humans.

· To codify in mathematical terms the common scheme for risk assessment.

· To define the purposes of the integrated system.

· To define a clearer and more formally correct decision analysis, identify criteria and acceptability of decision.

· To identify inputs, outputs and scope of the overall scheme. Different strategies are necessary for the different scopes. For instance, if the target of the integration is for regulatory purposes, attention has to be given to false negatives, and thresholds for them. If the purpose of the integration is drug development, false positives have to be minimised.

· To identifying a reference for an integrated system. Bodies, such as OECD and ECVAM, address individual methods, but the evaluation of the integrated system requires further efforts.

· To identify features for evaluation of the integrated system, such as cost, friendliness, possibility of automatism.

· To develop integrated in silico tools, which will address the overall chemical assessment, targeting ecosystems, and improving the current approach in which in silico models mimic the existing tests focussed on specific species.