The CRESS checklist for reporting stability studies: on behalf of the European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) Working Group for the Preanalytical Phase (WG-PRE)

1. Title/keywords

The title is the first place many readers look and will determine whether or not they will invest further time in reading the main article. It is essential that the title clearly indicates that the article describes a stability study. It should include details of the analytes evaluated, specimen matrices investigated, temperature and the time period over which stability is assessed.

Keywords must be well chosen to ensure that other laboratories searching for stability studies on the same analyte should be able to easily locate the study in the majority of search media. As a minimum the keywords should include:

• Stability

• The name of the analyte(s) if <3 analytes

• Preanalytical

• Sample type

2. Abstract

The abstract needs to expand on the title and must state that the study is a stability study and what the aims of the study are. It needs to include an overview of how the study is designed, which defines which analytes are being examined, in which matrices, in which tube types, at which temperature and over what period of time. It should summarise the key findings with some headline data and draw appropriate conclusions, without going beyond the results of the study.

3. Introduction

The introduction should define why it is important to understand the stability of analytes under various different conditions and state where in the diagnostic pathology process this information is necessary. It must thoroughly review the current state of knowledge for the analyte(s) being studied in the manuscript. It should provide a summary of what previous studies show and then define the knowledge gap that will be filled by the proposed study.

4. Aims

The introduction should end by clearly defining what the study aims to show, along with brief details about how this will be achieved. The aim should include the analytes, specimen matrices, temperature and the time period over which the stability study is done.

5. Materials and methods

The methodology used in the study should be described in sufficient detail to allow other laboratories to clearly understand what has been done, enable them to reproduce the study if required and allow them to understand how it fits with their own healthcare setting. Authors should state when and where the study has been done (start date and the end date of the study) to allow other laboratories to understand the time of year and allow for any seasonal influences.

6. Measurand/analyte

The analyte(s) must be clearly defined and must remove any possible ambiguity. For example NT-proBNP must not be confused with BNP. Where possible these should follow standard international terminology. Where analytes are non-standard or not commonly used sufficient details should be provided to allow other laboratories to easily replicate the work.

7. Samples

Stability of an analyte largely depends on the sample matrix. For most analytes, plasma or serum is the matrix with the best stability, whilst most analytes are usually much less stable in other matrices like saliva, urine and cerebrospinal fluid [9]. That is why it is important to define the sample matrix, e.g. whether the stability of the analyte was studied in whole blood, serum, plasma, or other matrices (e.g. urine, faeces, saliva, cerebrospinal fluid, etc.).

The sample container type, manufacturer and additive for primary samples and all aliquots used in the study should be described.

It is most preferable to use real patient specimens with analyte levels covering the analytical range. If real patient samples are used, then it is important to state if the samples used were from individual participants or patient/subject pools. If stability was studied on patient pools, details of how the patient pools were prepared and how many samples were included should be stated, including details of any guidelines or protocols followed with references. Deviations from guidelines and/or protocols should be stated. Authors should also state whether the samples were specifically collected for the study in question or if leftover samples following routine analysis were used.

8. Participants

The population in which the analyte is being studied must be clearly defined as this may have an impact on transferability, e.g. if inpatient or outpatient, renal patients or intensive care patients etc. were included. Demographic information should also be stated, to include as a minimum gender, age, geographical location, comorbidities, relevant medications, health status and ethnicity.

Where applicable, details of how the health status, comorbidities and medications information was obtained should be included, i.e. were they self-declared by participants or independently verified. If verified, details of how this was done should be included.

9. Preanalytical conditions (conditions prior to the basal time point)

For each analyte, there should be a definition of the conditions the baseline sample was subjected to, from the time the sample was obtained until the baseline measurement is performed. This should follow, but not be limited to, the SPREC (Standard Preanalytical Code) criteria for defining the main pre-analytical factors that could impact on a sample’s integrity [10].

The method of phlebotomy is an important preanalytical step which needs to be described and properly referenced in the manuscript. Authors should state which guidelines were followed and if there was any deviation from this or manufacturer guidelines. EFLM WG-PRE recommends that venous blood sampling is done in full compliance with the EFLM-COLABIOCLI Recommendations for venous blood sampling [11]. Whether the blood sample was venous, collected via a vascular catheter or another form of sampling should also be included, as well as how and for how long the tubes were mixed. Authors should also provide information on the length of sample storage before the baseline sample was analysed, how long the contact time with cellular components was, and when the sample was separated prior to storage as this can significantly affect sample stability for some analytes [10], [12], [13], [14]. Additional information including whether the tubes were kept open or closed during the experiment and whether the sample was exposed to light should be recorded and reported in the manuscript.

Sample transport conditions should be described and include as a minimum, details on the transit time, temperature, exposure to mechanical forces e.g. if samples have been transported via pneumatic tube systems [15].

Authors should provide details about centrifugation (e.g. time, g force, temperature) and storage conditions (temperature, air, light) throughout the process, as sample exposure to air is detrimental to the stability of some analytes (e.g. bicarbonate) [16]. If samples were stored frozen for a certain period of time, the thawing procedure may affect sample integrity, at least for some analytes (e.g. coagulation parameters) [17]. For this reason, if samples were stored frozen, authors should provide the detailed description about how samples were thawed and the number of freeze thaw cycles.

Additionally, temperature for a particular storage or handling period or any other part of the process throughout the study should be reported as mean, minimum, and maximum temperature. How temperatures were monitored throughout the process should be described in detail and include, but not be limited to, the following: measurement device, measurement method, measurement time points, device calibration status.

Any exclusion criteria and potential sources of bias introduced should be stated. This should include whether haemolysis, icterus, and lipaemia were excluded and if so, how the measurement of these indices was performed and what cut-offs were used, together with evidence for these thresholds [18], [19].

10. Analytical procedure

The methods must clearly define the platform on which each analysis was performed and which exact methodology (assay) was used. Any deviations from manufacturer recommendations should be explicitly stated and the rationale behind this provided. The traceability of the method to international standards should be defined, and within and between batch performance (i.e. analytical coefficient of variation, CV_A%) of the assay for the duration of the study should be stated at relevant levels including how this information was derived, i.e. was it from the manufacturer kit inserts or in house experiments. If it was derived from in-house experiments, details of the protocol followed should be stated and the protocol referenced. Details of the reagent, calibrators, and IQC (internal quality control) used including lot numbers for all and a statement on recent external quality performance should be stated.

Methods should also state whether analysis was completed in a single analytical batch, or in multiple batches. Single batch studies reduce the analytical bias due to method drift or recalibration but add additional time to the storage of samples, which could impact on the results [4]. It also requires the introduction of a preservation process without deterioration (freezing at −70 °C for example), which does not always exist or is possible for a specific analyte/sample matrix and may be itself a stability variable. If samples were run in multiple batches, the authors should state all relevant conditions between batches that may have influenced the results of the experiment and how bias between these batches was controlled.

Finally, in this section, the authors should state the number of replicate analyses for each individual sample and the number of samples at each time point.

11. Spiking studies

Where spiking studies are used, there should be a clear description of how these were performed, which – if any – guidelines were followed and how results were calculated [20]. Spiking studies add additional potential sources of error and bias due to the need to open and manipulate the sample. Moreover, the spiking material may behave differently to the naturally occurring analyte in a given matrix. To avoid any potential source of bias due to sample spiking, it is preferable to use patient specimens with high analyte levels, rather than spiked samples wherever possible.

12. Length of study

The actual duration over which stability of the analytes in each sample will be measured should be documented along with the time points at which analysis will occur. Justification for time points at which analytes will be measured must be stated.

13. Storage conditions

As for the storage of samples prior to measurement of analyte concentration at the basal time point, sample storage conditions for the duration of the stability experiment should be documented in line with all requirements listed under points 7 and 9.

14. Statistical analysis

The methodology behind data analysis must be stated and justified. Provide a justification for the sample size and numbers of replicates required, and confirm that this minimizes the effect of analytical imprecision [21].

The instability equation should be expressed as follows:

This complex equation will fit to non-linear behaviour observed for some analytes [4], [22], but is difficult to implement. A first-order equation could fit properly for many equations. Expression of simplified equation should be:

where PD% is percentage difference between the baseline result and a result measured at a particular time point and given by ‘a’, the slope of the equation, multiplied by the time. The instability equation does not have an intercept, as there is no difference at the starting point (time zero) i.e. the intercept would be zero. Reporting an instability equation in addition to a graph facilitates the interchangeability of information between laboratories to calculate the stability limit in relation to their MPD specifications based on their performance characteristics to meet their clinical requirements.

There should be sufficient data to estimate confidence intervals for the instability equation or the instability estimation in a given time for each analyte. Any statistical tools used to calculate statistical significance should be justified with details as to why that tool was selected.

15. Outliers

If outliers were observed, the statistical tools used to identify these should be described including justification for their use. Outliers should be looked for in both the samples and in the replicates of the samples. They should not be discarded but should be recorded in the final document. If outliers were removed or not used in any data produced from the study this should also be stated.

16. Performance acceptability criteria

The establishment of the MPD must be very clearly defined. It is critical that future readers are able to clearly understand how the MPD was determined and what the justification was for its value. The MPD should ideally be traceable to published data and should follow the Milan hierarchy for analytical performance. Therefore, ideally the MPD should be derived, in order of preference, from outcome data, biological variation, or state-of-the-art in line with the Milan hierarchy [23]. Whatever the source, justification of why the selected MPD is fit for purpose for the analyte and population in question should be given.

17. Results

Results should be presented textually, in tabulated forms and graphically. The instability equation should always be shown in conjunction with the results from which it was derived. Data should be consistently presented throughout the manuscript using consistent terminology and units (e.g. S.I.). Data interpretation should be easy, allowing key results to be quickly identified and understood. Where replicates have been performed, averages should be calculated using means for duplicates or medians in the case of triplicates, which should then be used for calculation of PD%.

Results should be plotted on a scatter graph with PD% on the ordinate and the time with appropriate units on the abscissa. Examples of data presentation are shown in Table 2 and Figure 1. As an indication of inter-individual variability, Figure 1A represents each patient studied with a different marker or line whereas Figure 1B gives confidence intervals of the mean PD%. Preferably, both types of graph should be provided, given the slightly different information they display and therefore the easy visual information they provide discussed below. Representation of individual plots are useful to identify possible cases of inter-individual variability, since an individual equation of loss of stability could be very different from the overall trend. The relationship of analytical variation with analyte concentration can be assessed by looking at individual patient data. Other inter-individual variation factors, such as cell number or genetic differences in metabolic pathways must be suspected on participants with different relationship patterns. These may require manual adjustment of the instability equation on the stability graph or, preferably, use of a least squares adjustment methodology. The regression equation must pass through the origin of coordinates since at time 0 the specimen is by definition unchanged. Note that the best fit to the measured concentration change over time may not be linear. Confidence intervals for defined time points should be plotted. From the MPD it is possible to extrapolate the SL from the instability equation and graphs. Raw data should also be available, ideally as an Excel file presented as supplemental data. An example format for the raw data is shown in Table 3, which provides percentage deviations (PD%) for each patient and average PD% and its confidence interval (1.96 × standard error of the mean [SEM]).

Figure 1:

Example of stability data presentation: stability of glucose in whole blood at room temperature (22 °C) in closed serum tube (manufacturer XX), protected from light. (A) Instability equation calculation (black line) using the least squares adjustment with confidence intervals (dotted lines) for the slope. All patient data is also shown. (B) Stability equation calculation using the point to point estimation with confidence intervals for the mean of patients at every study time. Blue line presents the MPD (maximum permissible difference) for serum glucose based on biological variation (which was 2.34% at the time when the study was performed).

The time point defined as the SL must be the last time point at which the analyte has been proven to be stable. As its behaviour is not known beyond measured time points, the instability equation should not be extrapolated outside the range of data included in the stability experiment.

18. Discussion

The discussion must discuss not just what is shown by the data but how it addresses the original aims of the study. It should also discuss how the data can be applied to other healthcare settings. Perhaps most importantly it should state what the limitations are and how these will affect transferability to other healthcare systems.

19. Funding

Details of how the study was financially supported should be included. This should incorporate details of any support in terms of provided consumables such as blood collection tubes, instruments, reagents, controls and calibrators.

20. Ethics

Details of any ethical approval obtained should be included. This should state who this was obtained from and whether patient consent was required/obtained.