Clinical trials that fail before they start. Practical view from a sponsor – Michael Herschel

There are trials where experienced clinical research experts have a bad gut feeling even before the first patient has been enrolled. There may be an excellent statistician, a famous group of key opinion leaders endorses the trial, regulatory agencies like it – but it becomes a never ending story finding patients for it.
The magic triangle of speed, quality, and cost needs to be changed into a square by adding human nature. Human traits here may have some consequences. Misanthropic Schopenhauer would see the dire consequences only:
Optimism – few centres will do it
Habits – keep those eligibility criteria that worked in phase II
Selective listening – but: they will try although they seem to fail to enrol
Grandeur – maximise it all at once, after all you are the leader
Linear thinking- if this is so I shall have to bow
Other traits, Rousseau might have remarked, create opportunities:
Curiosity – new ways of recruitment
Sports spirit – performance listings
Grace under pressure – last minute amendments
Community – investigator idea exchange
Sharing – financial support for study nurse
In preparing for a trial, feasibility is an often used phrase but is hardly done meticulously. It does start with a critical look at the protocol. Frequently, limiting patient eligibility starts with recruitment. Among the most frequent pitfalls are
– Excluding patients above a certain age, e.g. above 65 years of age
– Excluding women
– Excluding related disease even if the primary endpoint is applicable to them as well
More problems arise with exclusion criteria. Their choice is often based on obtaining a clinically pure population, with no concomitant disease and no other drugs taken regularly. Frequently used exclusion criteria whose wisdom may be questioned are
– Inability to co-operate (who knows ?)
– Intake of acetyl salicylic acid preparations
– Relevant cardiovascular, hepatic or renal disease (so what is “relevant”)
– Previous diagnosis of cancer (even if resolved…)
– Known allergies
The multitude is the more problematic as it not only limits the number of patients like a type of funnel, but has a negative impact on the drug approval labelling and prevents the sponsor from gaining knowledge in important patient populations before registration. In some cases the limitation, especially in phase II of clinical development, is reasoned for by reducing variability, however, there is no published evidence that “opening up” the patient population increases the variability of the primary endpoint and thus may lead to larger sample size.
Most often, feasibility is understood as the feasibility of the investigator site. According to ICH Good Clinical Practice, it is the task of the sponsor to ensure that the site is able to do the study. Even at renowned companies this meant to ask the investigator whether he would be able to come up with x randomised patients in y months. Temptingly, the answer was frequently “yes”, and this led to building, as the “yes” may have looked less credible, so-called contingency plans. Only recently it has been established practice in several companies to
– Ask for historical data, listing real patients from pre-specified past period, who would have been eligible, possibly with a disease characteristic that could even allow to calculate real variability of the endpoint (e.g. haemoglobin A1c)
– Set a time for the investigator to give feedback (as a knock-out criterion)
– Divide the number given by a figure between 2 and 6, depending on the difficulty of the study and the task to convince patients to participate
Yet, the mistakes of the past are still committed:
– Taking mostly key opinion leaders
– Taking the numbers given for granted
– Paying for feasibility data
– Believing in “centres of excellence” (who may have been good in the past…)
There are feasibility factors much less considered though their impact is high:
– Time to obtain an ethics committee vote
– Time to obtain trial permits (thanks to the European Directive, all of Europe but Italy will become slower now)
– Referral patterns or “who has got the patients for diagnosis, and who for treatment”
– Limited resources at the site even if not competing for the same patient group
– Investigator ready to move house or going on a sabbatical
– Time to completing the contract with the centre (mostly a problem in hospitals)
Finding the right investigator fee is yet another difficulty to overcome. The Scylla of too low fees, often bordering on “unfair” market value is facing, on the other side, the Charybdis of fees that may be spoiling investigators, or what is worse, be seen as an attempt to corrupt institutions. Clearly only those ,measures should be paid that exceed standard treatment cost. However, it may be necessary to use an database such as PICAS (Data Edge, Inc.) for comparisons. And, some studies may be necessary for approval, but boring enough so that only monetary inducement may lead to satisfying recruitment speed and quality.
In a time of shortening recruitment lines, contingency plans are now mostly based on good feasibility data and sound planning. Recruiting extra centres many months after the start of a trial is no longer viable. It is needed to define for each trial the “last intervention point”, i.e. the date after which no “fire brigade work” would salvage the recruitment time any more.
Therefore we suggest the following practical advice:
– Plan each study with (more than) sufficient centres
– Do not think you save money by restricting drug supplies or other trial materials
– Train investigators how to convince a patient to participate in a trial
– Make every trial as representative of the real patient population as you can by
eliminating exclusion criteria and being liberal with inclusions
More problems arise with exclusion criteria. Their choice is often based on obtaining a clinically pure population, with no concomitant disease and no other drugs taken regularly. Frequently used exclusion criteria whose wisdom may be questioned are
– Inability to co-operate (who knows ?)
– Intake of acetyl salicylic acid preparations
– Relevant cardiovascular, hepatic or renal disease (so what is “relevant”)
– Previous diagnosis of cancer (even if resolved…)
– Known allergies
The multitude is the more problematic as it not only limits the number of patients like a type of funnel, but has a negative impact on the drug approval labelling and prevents the sponsor from gaining knowledge in important patient populations before registration. In some cases the limitation, especially in phase II of clinical development, is reasoned for by reducing variability, however, there is no published evidence that “opening up” the patient population increases the variability of the primary endpoint and thus may lead to larger sample size.
Most often, feasibility is understood as the feasibility of the investigator site. According to ICH Good Clinical Practice, it is the task of the sponsor to ensure that the site is able to do the study. Even at renowned companies this meant to ask the investigator whether he would be able to come up with x randomised patients in y months. Temptingly, the answer was frequently “yes”, and this led to building, as the “yes” may have looked less credible, so-called contingency plans. Only recently it has been established practice in several companies to
– Ask for historical data, listing real patients from pre-specified past period, who would have been eligible, possibly with a disease characteristic that could even allow to calculate real variability of the endpoint (e.g. haemoglobin A1c)
– Set a time for the investigator to give feedback (as a knock-out criterion)
– Divide the number given by a figure between 2 and 6, depending on the difficulty of the study and the task to convince patients to participate
Yet, the mistakes of the past are still committed:
– Taking mostly key opinion leaders
– Taking the numbers given for granted
– Paying for feasibility data
– Believing in “centres of excellence” (who may have been good in the past…)
There are feasibility factors much less considered though their impact is high:
– Time to obtain an ethics committee vote
– Time to obtain trial permits (thanks to the European Directive, all of Europe but Italy will become slower now)
– Referral patterns or “who has got the patients for diagnosis, and who for treatment”
– Limited resources at the site even if not competing for the same patient group
– Investigator ready to move house or going on a sabbatical
– Time to completing the contract with the centre (mostly a problem in hospitals)
Finding the right investigator fee is yet another difficulty to overcome. The Scylla of too low fees, often bordering on “unfair” market value is facing, on the other side, the Charybdis of fees that may be spoiling investigators, or what is worse, be seen as an attempt to corrupt institutions. Clearly only those ,measures should be paid that exceed standard treatment cost. However, it may be necessary to use an database such as PICAS (Data Edge, Inc.) for comparisons. And, some studies may be necessary for approval, but boring enough so that only monetary inducement may lead to satisfying recruitment speed and quality.
In a time of shortening recruitment lines, contingency plans are now mostly based on good feasibility data and sound planning. Recruiting extra centres many months after the start of a trial is no longer viable. It is needed to define for each trial the “last intervention point”, i.e. the date after which no “fire brigade work” would salvage the recruitment time any more.
Therefore we suggest the following practical advice:
– Plan each study with (more than) sufficient centres
– Do not think you save money by restricting drug supplies or other trial materials
– Train investigators how to convince a patient to participate in a trial
– Make every trial as representative of the real patient population as you can by
eliminating exclusion criteria and being liberal with inclusions
(C) Michael Herschel 2008 – Alle Rechte vorbehalten

Schreibe einen Kommentar

Deine E-Mail-Adresse wird nicht veröffentlicht. Erforderliche Felder sind mit * markiert.