375
subgroups of the
nervous system therapeutic group is perhaps inappropriate. The table shows a comparison of ganglioside prescriptions with drugs having a similar coadjuvant role in various
indications (eg, nasal calcitonins, often used in low back pain and also claiming usefulness in diabetic neuropathy pain;3 carnitine, often used as metabolic integrator; acetylcarnitine, used as cerebroactive drug and recently [February, 1992] approved also for diabetic neuropathies; and thymus extracts, with indications ranging from recurrent infectious dieseases such as chronic bronchitis to cancer). Those shown are only a fraction of the drugs used in Italy as coadjuvants or placebos. Other commonly prescribed substances include peripheral vasodilators, drugs for atherosclerosis, tonics, multivitamins, and topical non-steroidal anti-inflammatory drugs. The table indicates that a move towards more rational and cost-effective drug use in Italy is a long way off. Department of
Pharmacology, Interuniversity Research Centre on Pharmacoepidemiology, University of Bologna, 40126 Bologna, Italy
NICOLA MONTANARO ALBERTO VACCHERI NICOLA MAGRINI
including DDDs for plain substances. WHO Collaborating Centre for Drug Statistics Methodology. Oslo: WHO, 1991. 2. Montanaro N, Vaccheri A, Magrini N, Battilana M. Farmaguida: a databank for the analysis of the Italian drug market and drug utilization in general practice. Eur J 1. ATC Index
3.
Clin Pharmacol 1992; 42: 395-99. Quatraro A, Minei A, De Rosa N, Giugliano D. Calcitonin in painful diabetic neuropathy. Lancet 1992; 339: 746-47.
Complacency bias in clinical trials SIR,-Conscious or unconscious bias is one of the common limitations of clinical trials, sometimes resulting in invalid data. A long list of different biases has been compiled.l Most can be avoided by simple but careful design (randomisation and double-blinding, for example). However, an adequate design is not always enough to guarantee the avoidance of bias. Here we describe "complacency bias". Compared with "therapeutic personality bias", which only occurs in unblinded trials,2 complacency bias occurs despite a double-blind design. The desire for positive results causes the investigator to report abnormally high success in each treatment arm, probably because the investigator unconsciously feels that by providing a higher number of patients with positive results, the overall result will be better, not realising that there are two or more
research must be aware of this type of bias, particularly in those trials where subjective assessments by the clinician are involved. Clinical Research
Department,
Lilly SA; 28100 Alcobendas, Madrid; and Clinical Pharmacology Unit, Hospital Santa Cruz, Liencres, Spain
J. A. SACRISTÁN J. SOTO A. ALANÍS
1. Sackett DL. Bias in analytic research. J Chron Dis 1979; 32: 51-63. 2. Spilker B. Bias and confounding factors. In: Spilker B, ed. Guide to clinical trials. New York: Raven Press, 1991: 21-26. 3. Council on Scientific Affairs and Council on Ethical and Judicial Affairs. Conflicts of interest in medical center/industry research relationships. JAMA 1990; 263: 2790-03.
Randomised consent trials SIR,—There are three issues arising from your June 27 editorial (p 1574) that need further discussion. In our experience some investigators participating in a randomised consent design trial have chosen a treatment for their patient in advance of learning about the actual allocation specified by the randomisation. If the allocated treatment differs from the a priori chosen treatment, it is rejected. Clearly physicians who favour a particular treatment should not participate in randomised trials. To better acquaint participating physicians with the implementation of a randomised consent design it would be desirable to have a workshop before starting such a trial. One feature of the randomised consent design is that the patient knows in advance the identity of the treatment to be administered at the time of patient consent. The patient is more fully informed than in a conventional randomised trial, in which consent is given in advance of knowing the assigned treatment. As a result, some conventional randomised trials may have a significant number of declinations after an initial patient consent. These may not be fully disclosed in the reporting of the trial. Finally, some ethicists have raised the issue of the use of patient data if a patient declines to take part in a randomised consent trial. The use of such data is no different than the use of aggregate data in disease registries on the common practice of a clinical department routinely analysing their outcome data. It is not uncommon in the USA to have patients sign a consent form allowing the use of the data even when the patient is not on an
experimental protocol.
Harvard School of Public Health and the Dana-Farber Cancer Institute, Boston, MA 02115, USA
MARVIN ZELEN
treaments.
As an example, in a double-blind randomised trial against placebo with response evaluated in part by the investigator’s subjective assessment, a planned interim analysis showed a high success rate (90%), which was surprising especially when considering that half the patients were being treated with placebo and that reported response rates to placebo in this disease are usually less than 60%. So we communicated our surprise for these overly positive results to the investigators. After completion of the study, during analysis of the final data, we decided to compare the global results before and after our comments to the investigators. We saw a statistically significant decrease in the total percentage of successes, mostly due to a decrease in the success rate in the group treated with
placebo. We think that the "desire" on the part of the investigator for reporting good results to the sponsor of a study has several implications. First, abnormally high success of the drugs compared during the study, higher than that in clinical practice, may be recorded. Second, type II errors, interfering with the demonstration of true differences, would increase. Third, complacency bias may, at least in part, be one of the causes contributing to high response rates to placebo in some clinical trials. In our experience as sponsors and monitors of trials, we have occasionally been informed by our investigators about good efficacy observed during a study that is still blinded. This tendency to reporting positive results may be related to the false perception by the investigator that all the compensations3 of a well-conducted trial must be necessarily associated with a "satisfactory" response. Although this may prove true especially when applied to trials sponsored by the pharmaceutical industry, all involved in clinical
Emergence of blinding malnutrition in Iraq SIR,—Levels of childhood morbidity and mortality have
strikingly during the past two years. Amid the overall desperate situation with respect to food shortages, contamination, increased
and increased disease prevalence, vitamin A deficiency is reemerging. In the spring of 1992 an outbreak of measles occurred, with a mortality rate in excess of 19% among children admitted to hospital. Complication rates were very high and at least 2 children became blind because of xerophthalmia. We have no supplies of vitamin A capsules available to us and I write this letter in
desperation. Ibn Al Baldi Hospital, Jamila, Baghdad, Iraq
Reversible writing in
NAIRA AWQATI
blocking of peripheral inputs and patients with cerebellar tremors
SIR,-In a 55-year-old patient with right-sided cerebellar lesions, hand-writing and electromyographs from forearm muscles were being recorded, with the expectation that writing would reveal intricate features of altered voluntary movements. At the end of a 2-3 h session, I observed that the patient could write clearly without tremor for a few minutes for the first time in six years. The figure shows his writing and that of another similar patient. I suspected a velcrostrap tied at midarm to fix the electrode holder was responsible for this effect by interference of transmission of neural impulses.
subgroups of the
nervous system therapeutic group is perhaps inappropriate. The table shows a comparison of ganglioside prescriptions with drugs having a similar coadjuvant role in various
indications (eg, nasal calcitonins, often used in low back pain and also claiming usefulness in diabetic neuropathy pain;3 carnitine, often used as metabolic integrator; acetylcarnitine, used as cerebroactive drug and recently [February, 1992] approved also for diabetic neuropathies; and thymus extracts, with indications ranging from recurrent infectious dieseases such as chronic bronchitis to cancer). Those shown are only a fraction of the drugs used in Italy as coadjuvants or placebos. Other commonly prescribed substances include peripheral vasodilators, drugs for atherosclerosis, tonics, multivitamins, and topical non-steroidal anti-inflammatory drugs. The table indicates that a move towards more rational and cost-effective drug use in Italy is a long way off. Department of
Pharmacology, Interuniversity Research Centre on Pharmacoepidemiology, University of Bologna, 40126 Bologna, Italy
NICOLA MONTANARO ALBERTO VACCHERI NICOLA MAGRINI
including DDDs for plain substances. WHO Collaborating Centre for Drug Statistics Methodology. Oslo: WHO, 1991. 2. Montanaro N, Vaccheri A, Magrini N, Battilana M. Farmaguida: a databank for the analysis of the Italian drug market and drug utilization in general practice. Eur J 1. ATC Index
3.
Clin Pharmacol 1992; 42: 395-99. Quatraro A, Minei A, De Rosa N, Giugliano D. Calcitonin in painful diabetic neuropathy. Lancet 1992; 339: 746-47.
Complacency bias in clinical trials SIR,-Conscious or unconscious bias is one of the common limitations of clinical trials, sometimes resulting in invalid data. A long list of different biases has been compiled.l Most can be avoided by simple but careful design (randomisation and double-blinding, for example). However, an adequate design is not always enough to guarantee the avoidance of bias. Here we describe "complacency bias". Compared with "therapeutic personality bias", which only occurs in unblinded trials,2 complacency bias occurs despite a double-blind design. The desire for positive results causes the investigator to report abnormally high success in each treatment arm, probably because the investigator unconsciously feels that by providing a higher number of patients with positive results, the overall result will be better, not realising that there are two or more
research must be aware of this type of bias, particularly in those trials where subjective assessments by the clinician are involved. Clinical Research
Department,
Lilly SA; 28100 Alcobendas, Madrid; and Clinical Pharmacology Unit, Hospital Santa Cruz, Liencres, Spain
J. A. SACRISTÁN J. SOTO A. ALANÍS
1. Sackett DL. Bias in analytic research. J Chron Dis 1979; 32: 51-63. 2. Spilker B. Bias and confounding factors. In: Spilker B, ed. Guide to clinical trials. New York: Raven Press, 1991: 21-26. 3. Council on Scientific Affairs and Council on Ethical and Judicial Affairs. Conflicts of interest in medical center/industry research relationships. JAMA 1990; 263: 2790-03.
Randomised consent trials SIR,—There are three issues arising from your June 27 editorial (p 1574) that need further discussion. In our experience some investigators participating in a randomised consent design trial have chosen a treatment for their patient in advance of learning about the actual allocation specified by the randomisation. If the allocated treatment differs from the a priori chosen treatment, it is rejected. Clearly physicians who favour a particular treatment should not participate in randomised trials. To better acquaint participating physicians with the implementation of a randomised consent design it would be desirable to have a workshop before starting such a trial. One feature of the randomised consent design is that the patient knows in advance the identity of the treatment to be administered at the time of patient consent. The patient is more fully informed than in a conventional randomised trial, in which consent is given in advance of knowing the assigned treatment. As a result, some conventional randomised trials may have a significant number of declinations after an initial patient consent. These may not be fully disclosed in the reporting of the trial. Finally, some ethicists have raised the issue of the use of patient data if a patient declines to take part in a randomised consent trial. The use of such data is no different than the use of aggregate data in disease registries on the common practice of a clinical department routinely analysing their outcome data. It is not uncommon in the USA to have patients sign a consent form allowing the use of the data even when the patient is not on an
experimental protocol.
Harvard School of Public Health and the Dana-Farber Cancer Institute, Boston, MA 02115, USA
MARVIN ZELEN
treaments.
As an example, in a double-blind randomised trial against placebo with response evaluated in part by the investigator’s subjective assessment, a planned interim analysis showed a high success rate (90%), which was surprising especially when considering that half the patients were being treated with placebo and that reported response rates to placebo in this disease are usually less than 60%. So we communicated our surprise for these overly positive results to the investigators. After completion of the study, during analysis of the final data, we decided to compare the global results before and after our comments to the investigators. We saw a statistically significant decrease in the total percentage of successes, mostly due to a decrease in the success rate in the group treated with
placebo. We think that the "desire" on the part of the investigator for reporting good results to the sponsor of a study has several implications. First, abnormally high success of the drugs compared during the study, higher than that in clinical practice, may be recorded. Second, type II errors, interfering with the demonstration of true differences, would increase. Third, complacency bias may, at least in part, be one of the causes contributing to high response rates to placebo in some clinical trials. In our experience as sponsors and monitors of trials, we have occasionally been informed by our investigators about good efficacy observed during a study that is still blinded. This tendency to reporting positive results may be related to the false perception by the investigator that all the compensations3 of a well-conducted trial must be necessarily associated with a "satisfactory" response. Although this may prove true especially when applied to trials sponsored by the pharmaceutical industry, all involved in clinical
Emergence of blinding malnutrition in Iraq SIR,—Levels of childhood morbidity and mortality have
strikingly during the past two years. Amid the overall desperate situation with respect to food shortages, contamination, increased
and increased disease prevalence, vitamin A deficiency is reemerging. In the spring of 1992 an outbreak of measles occurred, with a mortality rate in excess of 19% among children admitted to hospital. Complication rates were very high and at least 2 children became blind because of xerophthalmia. We have no supplies of vitamin A capsules available to us and I write this letter in
desperation. Ibn Al Baldi Hospital, Jamila, Baghdad, Iraq
Reversible writing in
NAIRA AWQATI
blocking of peripheral inputs and patients with cerebellar tremors
SIR,-In a 55-year-old patient with right-sided cerebellar lesions, hand-writing and electromyographs from forearm muscles were being recorded, with the expectation that writing would reveal intricate features of altered voluntary movements. At the end of a 2-3 h session, I observed that the patient could write clearly without tremor for a few minutes for the first time in six years. The figure shows his writing and that of another similar patient. I suspected a velcrostrap tied at midarm to fix the electrode holder was responsible for this effect by interference of transmission of neural impulses.
