• Question: how many times do you have to experiment on something before you know it works or if its safe

    Asked by choccybiscuit123 on 19 Jun 2019.
    • Photo: Matthew Bareford

      Matthew Bareford answered on 19 Jun 2019:

      Well this all depends… due to the nature of most things, whilst it may be safe 99 times, the 100th time, it may not be.

      I think this all depends upon what you are testing and what it is for. In terms of medicines, there are strict guidelines as set out by law as to how many times and when it is safe.

    • Photo: David Wilson

      David Wilson answered on 19 Jun 2019:

      When i design an experiment I include controls These prove that the experiment is working properly. If i’m looking to see if a protein is present in a cell I’ll check the experiment is working by also looking for the protein in a cell I KNOW it is in.
      I try and ensure all my experiments are safe, i use a lot of different chemical in my experiments. All chemicals have data sheets that tell you how to use them, what dangers they might pose and how to dispose of them properly. I usually wear a lab coat and gloves to protect me from any chemicals but sometimes I might need to use a fume hood if the chemical has noxious gases or a powder that i might inhale by accident. I also have written down instructions (sometimes called Standard Operating Procedures) that tell me how to do something step by step and can tell me if there might be any parts that i have to be careful with.

    • Photo: Marianne King

      Marianne King answered on 19 Jun 2019: last edited 19 Jun 2019 4:15 pm

      In what I do I guess I’m happy with a result if I get similar results after trying it three times. For example I work with patient cell lines, so if I wanted to know something I would do three of the same experiment per cell line, in three different cell lines. Different areas of science expect different numbers of repeats though. For things like drugs testing though there are much stricter guidelines because ultimately it’s a product that humans will be ingesting, so they have to be really really really sure they’re safe first! Discovering or creating a new drug and then taking it to a stage where humans can use it can take over a decade, and cost over $1 billion.

    • Photo: Kate Timms

      Kate Timms answered on 19 Jun 2019:

      This all depends on how variable the samples you’re looking at are. That means, how different the results are when you repeat the same thing. If they naturally vary quite a lot, you will have to do it more times than if its always quite similar.
      For example, I use a cancer cell line in the lab. It always grows very similarly because all of the cells are genetically identical to one another. So if I want to see if my treatment causes the cells to grow faster or slower, I would only need to repeat it 3-6 times.
      However, when I work on tissue from the placentas of pregnant women, they are all genetically different and grew in different conditions. This means that the cells will grow at a more variable rate. The cells from one placenta might grow much faster than from another for example. So to be able to know if it’s my treatment making the cells from the placentas grow faster, I would have to repeat it more times to make sure that it isn’t just ‘fluke’ caused by the placentas I used. For that, I’d have to do it 6-30 times depending on what I was looking at!

    • Photo: Kaitlin Wade

      Kaitlin Wade answered on 19 Jun 2019:

      Most of the ‘experiments’ that I do don’t involve manipulating cells or people in anyway so it’s always safe in the common sense of the word. One of the things that I have to deal with is safety of anonymisation of the people within my data (so I can’t identify anyone), so they just have numbers assigned to them rather than me knowing their names. And safety of data security so that no one else who shouldn’t have access to my data never gets hold of it. Especially given that most of the data that I work on is quite sensitive in that I have information from questionnaires and clinic measures, which people have given to me under the assumption that only I will look at it.

    • Photo: Nina Rzechorzek

      Nina Rzechorzek answered on 20 Jun 2019:

      Great question – it’s a critical part of statistical planning for an experiment. For most of the experiments I might do in the lab or the clinic, I use power calculations and samples size estimates to give me an idea of how many ‘samples’ or many ‘times’ I will need to measure something in order to properly test a hypothesis. To use a clinical study as an example, we would use power and sample size estimations to determine how many patients are needed to answer a given clinical research question like ‘is there a difference in response to a drug between these two groups of patients?’

      Factors that affect a power calculation include:
      (1) The precision and variance of measurements within any sample
      (2) The size of what we would consider to be a ‘clinically significant’ difference
      (3) How certain we want to be to avoid type 1 error (an error that occurs when we incorrectly reject the null hypothesis – in other words when we see a difference between two groups that isn’t truly there)
      (4) The type of statistical test we are performing

      When testing the safety of a drug for the first time in early clinical trials, a low dose is used in only a small number of patients before testing different doses in a larger group of patients. But at each stage of a clinical trial in humans or other animals, power and sample size calculations are performed to ensure that the minimum number of individuals are used to properly answer the question (so as to minimise harm and reduce waste of resources). All completely novel drugs must be first tested in animals for safety (in pre-clinical trials) before they can be tested in human clinical trials.

      If you want to learn more detail about power and sample size estimates for clinical studies here is a link to a free article:

      So in summary to answer your question – it depends on what type of research question it is, how variable you expect the data to be, what size of difference you think would be clinically or biologically relevant to see, and how confident you want to be in the results. Such information is often obtained from other published works that might have carried out a similar experiment in a different system, or a slightly different experiment in the same system, or from your own data obtained in a small pilot study.

    • Photo: Alex Blenkinsop

      Alex Blenkinsop answered on 20 Jun 2019:

      This is such a great question, and one which I think about every day for my work in testing new drugs in clinical trials. We have some fab answers already from the other scientists, so I’ll just explain how we do this in drug trials. We test drugs in phases, and each drug has to pass each phase to get through to the next.

      – Phase I is for safety (testing on healthy volunteers)
      – Phase II is for proof of concept and dose finding
      – Phase III is a large scale trial which is planned to be sure we have enough patients to be confident the treatment actually works. In this phase it’s crucial we do the sample size calculations Nina described so well to be sure we have enough power to detect the real effect of the drug and the probability we made a mistake and it actually doesn’t work is very small.

      After this, treatments will go to a regulator to decide whether or not there is enough evidence, and the trials have been conducted properly, to allow the drug to be licensed and go to patients. So it’s a loooong process and can take years and years.