• Select the study group. Depending on the type of study, define the nature of the exposure and the most reliable way to measure it; identify clearly diagnostic criteria for the disease and if appropriate rank patients according to severity.

• Set inclusion/exclusion criteria: inclusion criteria could, for instance, refer to certain levels of exposure or stage of disease, or age-range or sex, in order to make the study more precisely defined; exclusion criteria are often imposed to eliminate confounding factors and subgroups of subjects not representative of the disease under evaluation.

• Define and select a control group, matched with the main study group in actual and potential confounding factors such as sex, age range, smoking habit (in other words, any factors which are associated with exposure and outcome and that could lead to false-positive or false-negative results). In an intervention trial randomisation is always recommended, and if practicable and ethically acceptable, control subjects should take a placebo.

• Decide how many subjects to include in the study: this requires a calculation, which takes account of the precision of the assay (standard deviation or coefficient of variation), and the expected effect of the factor evaluated on the biomarker measurement, given a certain level of statistical power (generally 80% or more), and likelihood of a false-positive or type 1 error (generally 5% or less, i.e., p<0.05). The power of a study indicates the likelihood that the test is correctly rejecting the null hypothesis; a high statistical power means that the test results are likely to be valid.

• Address all ethical aspects and apply to the competent ethics committee. Clearance from an ethics committee is required before starting any human study.

• Standardise the method and conditions of sampling for all subjects and controls. This could be achieved by organizing sampling of subjects and controls to take place in parallel, i.e. on the same days, and - ideally - analysing all samples for a given biomarker in the same laboratory. In multicentre studies, it is important to identify and normalise inter-laboratory differences to avoid variation in sampling or sample processing influencing the results.

• Perform a 'dry run', going through all practical aspects of the study, before embarking on the study proper.

• Be aware of the General Data Protection Regulation (EU) 2016/679; make sure all data are held securely (and anonymised, if appropriate).

• Follow, whenever appropriate, principles of good laboratory practice (GLP).

Regarding specific recommendations for human studies using the comet assay, here is a list of technical issues to be considered in the design of the study and included in the technical protocol:

• It is generally not possible to carry out the comet assay on samples on the same day that they are collected, and in any case it is a wise precaution to store samples (whole blood, buffy coats or isolated peripheral blood mononuclear cells, PBMCs) at -80ºC or in liquid nitrogen. Whole blood and buffy coats can be snap frozen in small aliquots, but PBMCs need to be frozen slowly, in culture medium with 10% serum and 10% dimethylsulphoxide.

• On the day the comet assay is performed, samples can be taken randomly from the freezer; or samples representing one time-point in a series, from several subjects and controls; or if several samples were taken from each subject (and controls) at different time-points, all these could be run in one experiment. The selection method will depend on the kind of analysis that will be done once the data are collected.

• Assays should be run in duplicate, i.e. two slides (or two gels) per sample and per endpoint (strand breaks, enzyme-sensitive sites, repair activity, etc.).

• Positive and negative control samples, also known as reference standards, or calibration standards, should be included in every experiment. These are aliquots of cells from a single batch (of cultured cells, or PBMCs - the type of cell is not so important), untreated, or treated with an appropriate DNA-damaging agent (at a concentration to give a comet score in the mid-range), and stored frozen. This provides a check on the performance of the assay; if the comet scores from the standard samples deviate significantly from the expected values, values for the test samples cannot be trusted and the experiment should be repeated. However, minor deviations are acceptable, and allow normalisation of sample results (Collins et al., 2014a). Reference standards are particularly important in biomonitoring studies where many samples are analysed over an extended time period.

• Scoring of comets must be done blind, i.e. with slides coded so that it is not possible to discern their source, and preferably by the same person throughout a series of experiments to avoid subjective bias. However, if different scorers are involved, or if for example a microscope lamp or filter is changed, variations will be evident from the reference standards, and corrections can be made to sample scores. Whenever different scorers are involved preliminary studies should be carried out to ensure interscorer agreement.

• By convention, (at least) 50 comets are scored per gel, giving 100 per sample. The mean or median of the 100 comets is calculated, and it is this overall value that represents the damage in each sample and that is used in statistical analysis at the population level. Do not be tempted to use individual comet scores in analysis (unless you are interested in the distribution of damage among a population of cells).

• After scoring, slides should be dried and then archived (stored in boxes at room temperature) in case repeat scoring is needed at a later date.A more detailed account of applications of the comet assay in human biomonitoring can be found in Azqueta et al. (2020, in press). See also Collins et al., 2014b, for an overview of comet assay studies of environmental/occupational exposure, nutrition and disease. Information on how to perform the comet assay, whether to measure strand breaks, damaged bases, or DNA repair, is not given here as detailed protocols are available in various publications (for example, Azqueta et al., 2019). For information on statistical analysis of comet data, the article by Lovell and Omori (2008) is recommended.

Azqueta, A., Muruzabal, D., Boutet-Robinet, E., Milic, M., Dusinska, M., Brunborg, G., Møller, P., Collins, A.R. (2019) Technical recommendations to perform the alkaline standard and enzyme-modified comet assay in human biomonitoring studies. Mutat Res - Genet Toxicol Env Mutag 843, 24-32.

Collins, A., Koppen, G., Valdiglesias, V., Dusinska, M., Kruszewski, M., Møller, P., Rojas, E., Dhawan, A., Benzie, I., Coskun, E., Moretti, M., Speit, G., Bonassi, S. (2014b) The comet assay as a tool for human biomonitoring studies: The ComNet Project. Mutat Res Rev 759, 27-39.

Lovell, D.P., Omori, T. (2008) Statistical issues in the use of the comet assay. Mutagenesis 23,171-182.