Sources of Data Errors in Clinical Trials

The final outcome of any clinical trial is highly dependent on the data acquired from the study participants. Thus it is of at most importance that this data be of the highest quality as can be achieved and that it is credible, accurate and exactly in compliance with the respective source documents. In spite of endless efforts taken during this process of data acquisition, many data errors still occur, mainly in the form of human errors. Data validation thus becomes an essential process in all clinical trials, to ensure completeness and correctness of all data collected in order to preserve the integrity of trial data and the quality of the clinical investigation.

Data validation is a process undertaken to identify errors in the data collected during the clinical trial and to rectify them in a proper manner. Errors in the trial data should be corrected through the generation of proper discrepancies/queries, many trials use one or more software’s for the process of data validation and these software’s are usually used to generate queries.

The overall aim in any clinical trial would be to generate clean and accurate data. Thus the trial team would apply one or more ways to minimise errors generated and decrease the data validation and data correction time to save money, time and efforts.

Data errors can be classified into four main types:

Completely unintentional data errors: These errors could be generated by a wrongly calibrated or imprecise instrument.

Data errors resulting from negligence: These errors could be the result of wrong transcription either from source document to the CRFs or from CRFsto e-CRFs.

Fabricated data: These errors could arise from missing data or incomplete data which could be   replaced by similar values.

Data falsified to reach the desired objective: These errors arise when the data has been changed or created intentionally to meet the project target.

Completely unintentional data errors: These types of errors occur when an instrument is not calibrated properly or some types of system generated errors resulting from the spell checks, validation checks etc. These errors could be eliminated by proper calibration of the instrument daily according to the standard procedures. The data entry person should check twice if there was any failed validation check and must attach an annotation, so that change in the data can be noticed easily. Another approach to minimize such errors is by implementing double data entry procedure.

Data errors resulting from negligence: These types of errors occur as a result of improper transcription of data from the source documents to their respectiveCRFs or from the CRFs to the respective trial database. These errorsare usually validated by sending query forms to the corresponding site through the Data Clarification Form (DCF), and upon receipt of a proper response from the site investigator.This type of errors can also arise from misinterpretation of hand written data and notes, comments could lead to wrong data entry or sometimes due to wrong keying by the data entry technician, while incorrectly performed procedures could lead to inaccurate data as well as protocol violation. Missing data also fall in this category;these might be harmful for the trial result as it could be generally difficult to rely on such data.Normally these types of errors are typically detected and corrected by the data management team during the data validation process. Fortunately, missing data could be detected through   notes/new discrepancies being flagged.

Fabricated data: These types of data errors are considered as fraud and are very dangerous for the clinical trial. These errors are usually made by replacing missing datawith credible values obtained through interpolation from adjacent values or by intentional alteration or omission of undesired data. In some cases, data might berounded off to the nearly credible value. Masking of the data that could illustrate unreasonable risk to humans also falls under this category.These errors could be typically discovered through an abnormally small variability in the distribution of values andthrough multivariate tests, or through tests for the similarity of patterns in repeated measures.

Data falsifiedto reach a desired objective: This type of data error is also considered as fraud. Some examples of this type of error are; eligibility criteria has been fabricated in order to increase the number of subjects, data falsified to show a treatment effect, altering  the existing records to meet the desired objective. These errors, when present, could have a destructive effect on the trial credibility, especially if they were aimed at magnifying treatment efficacy. These types of errors could be detected through comparison of distributions.

The last two types of data errors are considered fraud and if this falsification gets proved, then thecredibility of the whole trialwill be damaged, and further legal actions would be taken eventuallyagainst the person/s /team responsible for the fraud. It probably might not be possible to completely prevent such cases if it was intentional,but definitely measures could be taken to reduce such kind of incidences to a great extent.

Thus, in order for the successful completion of any clinical trial, the data collected during the clinical trial must be accurate, credible and of high quality in terms of having minimal errors. Data validation plays a crucial role in generating such clean data. The use of stringent data validation processes along with proper data monitoring though monitors can prevent data errors, or minimize them that in turn would help generate accurate, reliable and credible-data.

Clinnovo is a clinical innovation company. It is pioneer CRO industry in India. Clinnovo offers professional Clinical Research Course, clinical data management course, SAS and imaging training. Clinnovo has been serving different bio-pharma industries across the world with excellence and high quality. For more information contact at +91 9912868928, 040 64635501