2

I'm writing a systematic review on endothelial cell cultures undergoing a certain treatment. In all studies included, the cultures come from an immortalized cell line. I'm assessing bias in the studies, following the Cochrane tool for critical appraisal, and noticed that many of the protocols do not describe random allocation of treatments. Furthermore, the Cochrane tool is specifically developed for RCTs, and a few of the questions feel only loosely applicable to the specific conditions of in vitro studies. As far as I know, there's no critical appraisal tool specifically developed for in vitro studies of cells, only human or animal studies.

I'm probably getting caught in the weeds here, but I have a few questions. Would random allocation of treatment vs no treatment potentially harm validity, if all the cells come from the same cell line and are essentially identical? If we deem random allocation non-applicable to the specific context of in vitro cell culture experiments, is it kosher to modify the Cochrane tool and omit random allocation as a criterion?

I posted this to the academia stackexchange first, but I was recommended to post it here. Hope this makes sense. Thanks for your help.

springbok
  • 21
  • 1
  • Welcome to the Medical Sciences Stack. Please take a [tour] and visit the [help] for more information on this Stack and how we work. Please note that we have a requirement for showing your prior research into your question. – bob1 Sep 26 '23 at 22:50
  • Might be a better fit on Bio.SE, this is a cell/tissue culture question, and less Medical. Also note that that many labs, including all the ones I've been in, do reorganize plate layouts to avoid bias from order or location (e.g., outer rim of a plate will dehydrate/evaporate faster than the middle generally), but this is rarely noted in the protocols. Often just in the notes of how you set something up that specific run or in a batch production record, etc. – Atl LED Oct 07 '23 at 04:27

1 Answers1

3

The main purposes of randomization in this context would be to avoid experimenter/observer bias and any temporal biases in the experimentation, rather than primarily trying to control for differences in the individual cell samples.

An example of a temporal bias might be that a filter in the lab is dirty, or a new reagent bottle is opened that is somehow different from the first, atmospheric/environmental conditions in the lab are different (seasonal changes in humidity or temperature, for example), or experimenter practice, or changing personnel. If a lab runs all the control experiments first, and then all the test experiments, they may have systematic differences due to these environmental changes. Or, the result of an assay might depend on precision of pipetting or measuring reagents. Or, a new student who joins the lab to run some test samples may be doing things slightly different than a previous one who did the controls.

Some of these problems might be solvable not by randomization but by pairing test and control samples in a non-randomized way, but that could also introduce biases. For example, if all the test samples are always run on the left, and the sun shines in from a window on the left, that could affect the test samples differently than the control.

Experimenter/observer bias might involve expectations of certain results that change how they are evaluated. Even if the measurements themselves are objective in some way, like an output from a machine, experimenters can bias the results if they, say, feel like an output is "too low" and therefore decide something is wrong and they re-run the sample, but they only do this when they're expecting a high result (e.g., with the test group).

It's hard to come up with any sort of comprehensive list of possible problems, but yeah, in general, lacking randomization even in a fairly standardized in vitro setting can harm validity.

Bryan Krause
  • 14,047
  • 4
  • 34
  • 54
  • 2
    I don't disagree with this at all, but on the other hand, in terms of practicality randomly assigning samples to wells in a 96-well plate for example, will lead to significant experimental error (set-up and interpretation issues), but assigning blocks of wells randomly can overcome this. For example don't always assign sample 1 to row A, sometimes have it in row C or F. Edge effects in cell culture plates are a real problem, so much so that if doing critical experiments I'll leave them out of the experimental design and fill with medium or PBS instead. – bob1 Sep 26 '23 at 19:26
  • 2
    @bob1 Agreed to all. Recommending an appropriate procedure to protect against biases while also mitigating other concerns requires a lot more specifics and details than available here. But, I'd also say that people doing in vitro work tend to be a lot sloppier about these things and seem to not be coached in avoiding bias from non-randomization than in clinical research. – Bryan Krause Sep 26 '23 at 20:11
  • 2
    I totally agree with the in vitro sloppiness - even down to the old n=3 problem! – bob1 Sep 26 '23 at 21:15