Use of historical controls

Haseman, J.K., Huff, J. and Boorman, G.A. (1984b). Use of historical control data in carcinogenicity studies in rodents. Toxicol. Pathol. 12 126-135. 

Tarone, R.E. (1982). The use of historical control information in testing for a trend in proportions. Biometrics 38 215-220. 

To understand the context of these other techniques, one should put them in perspective. One can rank methods for obtaining information in order of increasing confidence that the conclusions are valid. The order would be a single memorable case, a series of memorable cases, and a series of consecutive cases. The control observations for these would be historical controls, either articulated by the observer or merely understood. The assumption with the use of historical controls is that the controls are comparable to the patients and that the outcome of the treated patients if not given the new treatment would be identical to the historical controls. Often initial therapeutic trials of new cancer drugs are done in a consecutive series of treated patients and compared to historical controls. In these studies, the controls are usually not articulated. The authors assume that the natural course of the disease is so predictable in these patients that change from the predicted course is due to the drug. The validity of this assumption must be carefully examined when one interprets any study that is a case series. 

The wording cited above from the Guidance Document is selected very carefully and there is not too much to add. Generally, historical control data should be collected and adequately documented in each laboratory, and their usefulness is selftevident. For exceptional cases like very rare malformations even the use of historical control data from other laboratories may be appropriate (these are most unlikely to be caused by specific housing conditions). However, a clear and unequivocal study will always be one which exclusively refers to current control data. If this is not the case, then the use of historical control data may help clarifying open questions, but the study will normally be classified as grey zone. ... 

Haseman, J. K. (1995). Data analysis Statistical analysis and use of historical control data. Regul Toxicol Pharmacol 21, 52-59 discussion 81-86. 

S.5.5.2.3 Historical controls Since many clinical trials are conducted in the same diseases, with the same control treatments there is an obvious desire to make the most use of this potentially valuable information. Can we compare the results of a new treatment in a group of patients with a group of control patients extracted from a historical database For example, suppose we are testing a new treatment for migraine headache and 60% of patients improve in the first 2 h post-treatment, compared to 30% in a group of historical control patients treated who had been treated with the current gold standard. Are we able to conclude that the new treatment is preferable to the gold standard ... 

When limited historical data is available from the aforementioned sources, there are external sources of historical control data that can be used with the appropriate qualifiers. The leading source of historical control data for developmental toxicity remains the MARTA Historical Control Database. This data originated from industry surveys conducted by the Middle Atlantic Reproduction and Teratology Association (MARTA) and the Midwest Teratology Association (MTA) (4-6). It is the most comprehensive compilation of control data from developmental toxicity studies in CD rats and... 

Animal suppliers can be a source of historical control data, although it is usually limited, since they do not typically perform cesarean sections or fetal or pup evaluations, but they may compile some natural delivery data such as litter size and fertility rates in breeding colonies. This type of data can be useful when a laboratory is attempting to gain a basic understanding of the reproductive performance of a species/strain, or gather information about a particular supplier or supplier site with which they are not familiar. 

EOQC (1980) European Organization for Quality Control, 4th European Seminar (1980) Validation of Manufacturing Processes (Geneva) [5] Definitions, installation and operational qualification, development and manufacturing phase, responsibilities and organization, use of historical data, change control and revalidation... 

The choice of strains is highly influenced by the availability of historical controls (data from parameters in control animals). Although some authorities do not rely on comparisons of controls from other studies, the investigator should try to use a strain for which a profound basis for comparisons with results from other studies is available. In case of the availability of data from historical controls , the investigator can identify trends due to genetic shifts and use this information for the interpretation of the actual study. 

The third problem concerns the construction of the yardstick itself. We need to know what would have happened to the patients had we treated them otherwise, but whatever approach we use will have its difficulties. For example, we may use the method of historical controls by which we compare our results to results obtained previously with different patients using the alternative treatment (which may be another treatment or no treatment at all), but the problem is that not only are the treatments different so are the patients and they may differ in important ways we cannot even measure. Another alternative that is possible with some diseases is the method of baseline comparisons, whereby patients are compared with their own baseline values. This method is subject to very many possible biases, among which regression to the mean (see below) and time trends (see discussion about multiple sclerosis above) are important. 

It turns out that there is another branch of mathematics, closely related to tire calculus of variations, although historically the two fields grew up somewhat separately, known as optimal control theory (OCT). Although the boundary between these two fields is somewhat blurred, in practice one may view optimal control theory as the application of the calculus of variations to problems with differential equation constraints. OCT is used in chemical, electrical, and aeronautical engineering where the differential equation constraints may be chemical kinetic equations, electrical circuit equations, the Navier-Stokes equations for air flow, or Newton s equations. In our case, the differential equation constraint is the TDSE in the presence of the control, which is the electric field interacting with the dipole (pemianent or transition dipole moment) of the molecule [53, 54, 55 and 56]. From the point of view of control theory, this application presents many new features relative to conventional applications perhaps most interesting mathematically is the admission of a complex state variable and a complex control conceptually, the application of control teclmiques to steer the microscopic equations of motion is both a novel and potentially very important new direction. 

Reflectance Spectrophotometry. Because of discrepancies that can occur between strength and shade evaluations in solution and on textile substrates, the latter is often the preferred evaluation technique. In the case of dye manufacture, many dyes are standardized in solution but there is always a final control step where dyeings are prepared. Historically, such dyeings have been evaluated visually for the relative strength and the shade of the dye under test on the substrate, compared to the standard. More and more attempts are being made to do such evaluations objectively. Guidelines for the use of this technique have been pubflshed (43). 

---