CTU Bern (Clinical Trials Unit)

Lessons learned

Welcome to Lessons Learned!

We have compiled questions that might arise in the course of your clinical research project.

We hope that you find these helpful.

The Wilcoxon rank-sum test which is also known as Mann-Whitney U test or Mann–Whitney–Wilcoxon test is arguable the most famous and frequently used non-parametric test to compare two groups. Very often, only a p-value is given as a result even though it is good practice and generally recommended to report a measure of effect size for an inferential test. The reason may be a widespread misconception about what the test actually tests and whether a usable measure of effect size is available.

It is a widely believed that the Wilcoxon rank-sum test is a test for equality of medians, which is only true under very specific circumstances—i.e. under the strong assumption that both of the two distributions are symmetrical about their respective medians or that the distributions are of the same shape but differ in location. In many situations, the median difference is therefore not a good measure of effect size.

What the test actually does was stated by Mann and Whitney in the title of their famous paper on the extension of Wilcoxon’s test: “On a test whether one of two random variables is stochastically larger than the other”. In other words, it tests whether an observation from one group is larger than an observation from the other group, or—in the context of a clinical trial—whether a patient in the treatment group has the better outcome than one in the control group.

The corresponding measure of effect size is the probability that a patient in the treatment group will have a better outcome than a patient in the control group—a quantity of practical interest that does not depend on the unit of the outcome. We suggest reporting this “Mann-Whitney statistic” with a confidence interval together with the p-value from the test.

Unfortunately, the Mann-Whitney statistic goes under various names—e.g. probabilistic index, measure of stochastic superiority, or common language effect size—and it is not always reported by statistical software packages. As a workaround, the well-known area under the receiving-operator curve (AUC or just C-statistic) can be reported—it is nothing else than the Mann-Whitney statistic.

According to Swiss law, a clinical the trial must be registered on the Swiss Clinical Trials Portal SNTCP (kofam.ch) before the official start of the trial. Moreover, the trial must be registered in a register that meets WHO registry criteria such as for example clinicaltrials.gov.

When filling out the application for the ethic commission (BASEC), it is recommended to agree to the automatic transmission of the trial’s info to the SNTCP portal. The info concerning the trial will be automatically transferred to the portal as soon as the trial is authorized and an external identification number is available (e.g. NCT12345678). This number is automatically generated by clinicaltrials.gov as soon as you start registering a trial.

It is possible to start adding info about the trial on clinicaltrials.gov at the time of your initial submission to the ethics commission but you may not yet have all the information to fill in the required fields. If you do so, please set the study to: “Not yet recruiting”. It is possible and actually advised to start and then add the information later, after receiving formal approval from the Ethics Committee.

Make sure that the information on kofam.ch and clinicaltrials.gov match the trial protocol and are up to date before including patients.

To be recognized as legally valid, or in other words, to be considered as fully equivalent to handwritten signatures, electronic signatures must comply with the requirements of a «qualified electronic signature» as stated in the Federal Code of Obligations (Art. 14 Abs. 2bis)

The “Federal Law regarding the electronic signature” regulates and defines the requirements for the qualified electronic signature:

  • it is allocated exclusively to its owner,
  • it enables the identification of the owner,
  • it is created by means which are exclusively controllable by the owner herself/himself
  • the link between the electronic signature and the data to which it refers, is implemented in a way that allows for the detection of subsequent modification of said data» (Swiss Federal Law regarding the electronic signature, Art. 2b, https://www.admin.ch/opc/de/classified-compilation/20131913/index.html)

In addition to the specifications regarding the signature itself, the Swiss Federal Law also stipulates how the creation process of the signature must be executed. The generation of electronic signatures must adhere to a specific authentication process, which can only be carried out by providers who have been validated by the federal administration.

In Switzerland, solely the „SuisseID“complies with all legal requirements as outlined above. It is offered for example by the Swiss Post, who is recognized as an accredited provider

The electronic signatures, which can be generated by the EDCSs used in clinical trials, do not comply with the legal requirements regarding validated electronic signatures effective in Switzerland. Consequently, the EDCS-generated signatures cannot be used on the Informed Consent Form, whether for the signature of the study participant nor for the signature by the Principal Investigator.

In a randomized-controlled trial (RCT), the standard device is compared to a newly developed device. The trial is designed as an open-label RCT: the participants as well as the researches (investigator, study nurse etc.) know which treatment (investigational or control intervention) is being administered to the participant; neither side is blinded to the allocation of treatment.
The statistician prepares the randomization list and forwards it to the data manager, who is responsible for the implementation of the list in the database. Additionally, the statistician sends the randomization list to the principal investigator for his/her information. Is this procedure acceptable?

No it is not - as a matter of fact, it is highly problematic. One of the key features of an RCT is the concealment of the allocation sequence: the person who is responsible for the inclusion of participants should have no knowledge about the type of intervention allocated to the next participant. This knowledge might influence the decision of the enrolling physician whether or not to include a specific prospective participant. In short, knowledge about the allocation sequence could lead to selection bias in RCTs. Consequently, it is essential that the randomization list is password-protected and concealed from study personnel involved in patient recruitment at all times. Only the collaborators that generate and implement the list (i.e. statistician, data manager) should have access.

4 investigators in the UK, Germany, USA, and Switzerland plan to perform a drug trial with participating sites in all four countries, and all four investigators are named as co-sponsors in the protocol. However, according to the “AW-Information Sheet FAQ on clinical trials with medicinal products”, Swissmedic accepts only one sponsor for a clinical trial who must assume the overall responsibility for the clinical trial in Switzerland. How should this be handled?

To ensure that the responsibilities are clearly defined, only one investigator (or the institution he or she is associated with) in either one of the four countries should be appointed to the role of the sole sponsor, i.e. taking over the overall responsibility for study initiation and conduct and this should be displayed in the protocol accordingly.

In the other countries sponsor-representatives must be installed. If two or more countries are part of the EU, for those countries only one sponsor-representative in either one of those countries is required.

The responsibility of a sponsor-representative is usually taken over by a contract research organization, a country affiliate of the sponsor (if the sponsor is e.g. a pharma company with headquarters abroad) or another institution located in Switzerland and specialized in this field of work. It is recommended not to delegate this function to the individual investigator, since in the worst case, he or she might be personally confronted with liability and coverage claims, which may be difficult for the investigator to cope with. For studies conducted at Insel, an application can be made to the Departement für Lehre und Forschung (DLF) so the investigator can take over this function without taking personal risks.

In Switzerland the sponsor-representative takes over the role of the communication partner with the authorities during the approval process as well as other responsibilities, which are described in the Interpretation guide – Obligations of representatives of foreign sponsors, version 1.1 (link see below) from the Federal office of public health (FOPH).

Note that for non-drug / non-medical devices trials, co-sponsorship may be acceptable but this needs to be discussed with the responsible local authorities on a case by case basis since such trials may not be regulated in foreign countries.

A new device is compared against the standard device in a randomized controlled trial (RCT). A patient is randomized to the new device before the surgery. During the surgery, complications occur. Therefore, it is decided to use the standard device because the operation team feels more familiar with that device. Since the patient did not receive the randomized device, no data was collected and no further follow-up was performed for this patient. Is this correct?

No, this is not correct. A randomized patient may never be excluded from a study in retrospect. Every effort has to be taken to collect all baseline data, all procedural data, and all follow-up data, also of patients that did not receive the correct intervention or patients with major protocol violations. Only then, an intention-to-treat (ITT) analysis is possible, which is the gold standard in RCTs. This means that all randomized patients are evaluated in the group they were originally allocated to, regardless of the intervention actually received.

The case: CTU Bern had been asked to set-up a study database including electronic case report forms (eCRF) for a (national) multicenter RCT. The investigator asked whether it would be possible to also store names and contact details of participants within the database. She argued that this would make follow-up phone calls much easier. Based on a detailed assessment of the relevant regulations (Human Research Act, Human Research Ordinance), information from ethics committees, and personal discussions with the cantonal ethics committee the decision was taken not to store this identifying information centrally but rather only locally (paper or Excel code list, depending on the site).

Lessons learned: Generally, a research database should not contain any identifying data. Examples of identifying data are: names, date of birth, addresses including phone numbers, but also: initials, patient identification numbers (e.g. Insel PID),  case identification numbers (e.g. Insel FID), and e-mail addresses. Not identifying data according to swissethics are: Year of birth and gender.

If you are participating in an international study that is collecting identifying data, you should enter dummy information for the fields that are not allowed in Switzerland (e.g. ‘A.A.’ for the initials or ‘01.01.YYYY’ for the date and month of birth). Also refer to the guideline on coding of trial subjects accepted by swissethics which can be found in the template section of swissethics website (http://www.swissethics.ch).

In special cases the collection of identifying data might be needed (e.g. addresses for follow-up visits if these cannot be stored otherwise e.g. if reminders/forms are send by e-mail, month or even date of birth for studies with children). In this case, you must specify in the study protocol as well as in the informed consent that identifying data are being collected and you must add a justification why it is needed. The ethics committee will decide if they agree to the collection of identifying data.

Even if you are allowed to collect identifying data you should make sure that you protect them well, i.e. that you restrict access to this data to the people who need it e.g. addresses should only be accessible to persons who actually need to contact the patients.

A site delegation & signature log (also called authorization log, authorized signature log, or site delegation log) is an important document that should be implemented in all clinical trials. On this log, the principal investigator lists study staff who is authorized to participate in the clinical trial and documents which task each person is allowed to perform. Its aim is also to give a sample of the signatures and (hand-written) initials of study staff so it can be reconstructed which staff member signed or initialed a study document. This document should not be confused with the List of Study Staff submitted to the EC during the protocol submission process or other similar lists, since these may not fulfill the requirements of a site delegation log. A template can be downloaded here.

No, backdating is never allowed. You should always add the current date, ideally with a comment “signed for [date when you originally completed the document]” (e.g. “Lieschen Müller, 03.10.2016, signed for 02.03.2016”). The reason for this is that one should be able to reconstruct when documents were completed & signed & when later entries (in this case e.g. the signature) were added. If you simply backdate documents, it looks as if you dated the document at the time of completion, which is not the case.