Home  >  Special Features
Msc_Apr23 .
you can get e-magazine links on WhatsApp. Click here
Special Features
+ Font Resize -

Drug discovery research data analysis & interpretation of efficacy studies

K. Sadasivan Pillai
Wednesday, September 6, 2023, 08:00 Hrs  [IST]

The Covid-19 episode has inspired several pharmaceutical companies to carry out drug discovery research in India for various indications. There was a boom in the synthesis of new drug entities (NDE), which are biotechnology or chemical based. A newly synthesized entity is initially subjected to throughput in vitro screenings, especially with cell lines, enzymes, etc. The data obtained from these screening studies are often analyzed using statistical software. In most in vitro studies, the dose-response is either agonist or antagonist. Often, the 4-parameter logistic curve fits in this dose-response relationship, and the statistical software determines the parameter of interest, the EC50/IC50. The EC50/IC50 is usually used to assess the efficacy of an NDE. Selection of an NDE for the next level studies based on EC50/IC50 should be made cautiously. A similar EC50/IC50 of two NDEs does not mean that the efficacy of these NDEs is the same. Similarly, different EC50/IC50 does not indicate that the efficacies are different. When comparing the EC50/IC50, always consider the confidence intervals of these parameters and the slope of the dose-response curve. In vitro studies of some NDE may show hormetic responses, that should not be ignored. In a hormetic response, either low dose stimulates and high dose inhibits, or low dose inhibits and high dose stimulates. According to E. J. Calabrese, Founder of the International Hormesis Society,  ‘hormetic dose–response model is the most common and fundamental in the biological and biomedical sciences, being highly generalizable across biological model and endpoint measured’. Hormetic dose–response has been reported in the Anti-tumour Drug Screening Programme of the National Cancer Institute, USA, where at small doses in yeast and human tumor cell databases, the response compared to the control increased, and at high doses it decreased. The author prefers to conduct in vivo efficacy studies with diseased models, especially rodents, where the animals show a slightly large variance in the efficacy parameter/s to be evaluated within the group. If the variance within the groups is very less, it is likely that small differences in the mean values will be statistically significant. This significant difference may not have any biological relevance. But for safety evaluation studies, it is recommended less within group variations, so that small differences in the mean values would become statistically significant.  While interpreting the results of animal efficacy studies, always consider whether the results are extrapolatable to patients. In an animal experiment, getting a significant difference compared to the control group does not mean that the NDE is efficacious. In recent years, assessing a significant difference based on P-values alone has lost its significance.   Though the P-value was initially introduced by Karl Pearson in 1900, it was Ronald A. Fisher who formalized the concept of it. But, Fisher never intended to use the P- value to classify the data dichotomously into significant and non-significant. According to him, it is up to the scientist to make the interpretation of the P -value. As Gertrude Stein said: ‘A difference, to be a difference, must make a difference.’  Now the question is how much should be that difference? Before designing an animal efficacy study, it is important to have an idea about the effect size expected from the study. If your NDE is an antipyretic drug to beat acetaminophen, the effect size of your NDE should be large, because acetaminophen is a widely accepted antipyretic drug across the globe. If your NDE is an anti-inflammatory drug to beat conventional NSAIDs, which cause gastrointestinal symptoms, you can think of a small effect size for your NDE, if it does not cause gastrointestinal symptoms like NSAIDs. Depending on the required effect size number of animals in the groups differs. It is not a good practice to design an experiment with a few animals in the control group and more in drug treatment groups. If the groups are not balanced with the number of animals, the sensitivity of the statistical analysis drops. In some occasions, the sensitivity of the test can be increased by placing more animals in the control group than the other treatment groups. If at all you are using the P-value to determine a significant difference, present 95% confidence limits of the difference along with the difference.
Always interpret the data obtained from in vivo studies biologically, rather than statistically. See the below-given example of an efficacy study with an antidiabetic NDE in rats.

Group 1 (Vehicle treated)`(Glucose mg/dl) Group 2 (NDE treated) `(Glucose mg/dl)
189, 195, 169, 206, 175 138, 161, 156, 171, 247
Mean ± SD= 186.8±14.9 (n=5) Mean ± SD=174.6 ± 42.2 (n=5)
Mean ± SD=156.5s ± 13.4 (n=4)

Note: SD-Standard Deviation; s-Significant; n- Number of observations.

The mean glucose value of NDE-treated rats (174.6 mg/dl) is not statistically significant compared to the glucose value of control rats (186.8 mg/dl). All the rats in the NDE-treated group showed a low glucose value, except one rat which showed a high glucose value (247 mg/dl). An outlier test (Grubb’s test) indicates that 247 is an outlier. Removing this value decreases the mean value to 156.5  and SD to 13.4. This mean value is statistically significant indicating that the NDE possesses antidiabetic/hypoglycemic properties. Suppose, in the NDE-treated group, the high glucose value that the rat showed was 246 mg/dl, instead of  247 mg/dl (a difference of 1 mg/dl), you will find the value 246 is not detected as an outlier according to Grubb’s test at 5% probability. That means the value 246 should be removed from the NDE-treated group for comparison with the control, which will show a non-significant result on the comparison. The NDE has failed because of the difference of 1 mg/dl glucose. In such a situation, always interpret the data biologically. In this context, it may be emphasized that do not use Grubb’s test to remove data that you do not like in order to obtain a result that you like. Keep in mind that sometimes an outlier could be a genuine data point and others are not. Another important aspect to be considered in the conduct of animal experiments is randomizing animals into groups. Being, randomization is tedious and laborious work, scientists perform this work casually.  In an article published in Cancer Research in 2011, Kenneth R Hess stated that ‘the key deficiencies that are seen in animal experiments are failure to randomly allocate animals to treatments and failure to blind observers to treatment assignment during outcome assessments’. The chances of getting positive results are very high if animals are not properly randomized.
Never use a statistical tool without understanding the underlying principles of it. The OECD Guidance Notes (ENV/JM/MONO(2002) state that: ‘where statistical analyses are used to reach a judgment, an awareness of the validity of the tests employed and the degree of certainty (i.e. confidence) pertaining within the context of the study should be demonstrated’. For example, if the number of groups to be compared is more than two, one-way analysis of variance (ANOVA) is the statistical tool of the choice. In some cases, though the high-dose group shows a higher value than that of the control and other dosed groups, ANOVA may not show a significant difference, which means there is no significant difference among the groups.  Sometimes you may get a significant difference when you compare the high-dose group with the control using Dunnett's test, ignoring the results of ANOVA.  Interestingly, Dunnett never recommended ANOVA for comparing a dosed group with the control. The ICH S2 R1 Guidance on Genotoxicity Testing and Data Interpretation states that ‘the application of statistical methods can aid in data interpretation; however, adequate biological interpretation is of critical importance’. Assessing the efficacy of an NDE should be based on its biological relevance. Statistical analysis should never override biological relevance.

(Author is Director - Toxicology, PNB Vesper Life Science Kochi, Kerala-682011)


Post Your commentsPOST YOUR COMMENT
* Name :     
* Email :    
  Website :  
Pharma live expo
Close Open
Copyright © 2023 Saffron Media Pvt. Ltd |