ARTICLES
`
`Quality of informed consent in cancer clinical trials: a cross-
`sectional survey
`
`Steven Joffe, E Francis Cook, Paul D Cleary, Jeffrey W Clark, Jane C Weeks
`
`Summary
`
`Background Investigators have to obtain informed consent
`before enrolling participants in clinical trials. We wanted to
`measure the quality of understanding among participants in
`clinical trials of cancer therapies, to identify correlates of
`increased understanding, and to assess providers’ beliefs
`about clinical research. We also sought evidence of
`therapeutic misconceptions in participants and providers.
`
`Methods We sent a standard questionnaire to 287 adult
`patients with cancer who had recently enrolled in a clinical
`trial at one of three affiliated institutions, and surveyed the
`provider who obtained each patient’s consent.
`
`Findings 207 of 287 (72%) patients responded. 90% (186)
`of these respondents were satisfied with the informed
`consent process and most considered themselves to be well
`informed. Nevertheless, many did not recognise non-
`standard treatment (74%), the potential for incremental risk
`from participation (63%), the unproven nature of the
`treatment (70%), the uncertainty of benefits to self (29%), or
`that trials are done mainly to benefit future patients (25%). In
`multivariate analysis, increased knowledge was associated
`with college education, speaking only English at home, use
`of the US National Cancer Institute consent form template,
`not signing the consent form at initial discussion, presence
`of a nurse, and careful reading of the consent form. Only 28
`of 61 providers (46%) recognised that the main reason for
`clinical trials is benefit to future patients.
`
`Interpretation Misconceptions about cancer clinical trials are
`frequent among trial participants, and physician/investigators
`might share some of these misconceptions. Efforts to
`educate providers and participants about the underlying goals
`of clinical trials are needed.
`
`Lancet 2001: 358: 1772–77
`See Commentary 1742
`
`Departments of Paediatric Oncology (S Joffe MD) and Adult
`Oncology (J C Weeks MD), Dana-Farber Cancer Institute, Boston,
`MA, USA; Department of Medicine, Children’s Hospital, Boston
`(S Joffe); Department of Medicine (J C Weeks) and Division of
`General Medicine (E F Cook DSc), Brigham and Women’s Hospital,
`Boston; Department of Health Care Policy, Harvard Medical
`School, Boston (Prof P D Cleary PhD); Division of Hematology/
`Oncology, Massachusetts General Hospital, Boston (J W Clark MD)
`Correspondence to: Dr Jane C Weeks, Center for Outcomes and
`Policy Research, 454 BRK Suite 21, Dana-Farber Cancer Institute,
`44 Binney St, Boston, MA, USA 02115
`(email: jane_weeks@dfci.harvard.edu).
`
`Introduction
`Ethical and legal doctrines mandate that, with rare
`exceptions, research participants or their surrogates give
`informed consent before enrolment in clinical research.1–7
`However, concerns have been raised as to the adequacy of
`such consent.8 Many studies have revealed difficulties
`with comprehension of clinical trials9–17 (eg, the right of
`patients
`to
`refuse
`to participate
`in
`research,
`misunderstandings of research procedures—including
`randomisation, under-reporting of potential risks, and
`inadequate information about alternatives).10,13–16 Perhaps
`most important, although trials are aimed to benefit future
`patients, many participants might have the therapeutic
`misconception
`that “every aspect of
`the research
`project . . . was designed to benefit [them] directly”.18
`Absence of simple, standard methods for assessment of
`outcomes of the consent process has restricted research on
`informed consent and prevented monitoring by
`Institutional Review Boards (IRBs).19 Investigators have
`used heterogeneous methods that are tailored to their
`specific protocols,
`thereby
`restricting meaningful
`comparisons of their work. One generic method exists
`(the Deaconess Informed Consent Comprehension
`Test20), but its validity has not been established, it is
`difficult to administer and score, and it has not been
`widely used.
`We have designed a new questionnaire, the Quality of
`Informed Consent (QuIC), to assess the
`informed
`consent process for clinical research of cancer therapies.21
`In the present study, our main objectives were to measure
`how well newly enrolled trial participants understood the
`trials in which they were participating, and to ascertain
`what factors were associated with greater understanding.
`In particular, we investigated whether a consent form
`template that was recently published by the US National
`Cancer Institute22 resulted in a measurable improvement
`in participants’ knowledge. This template uses a question-
`and-answer format that is easy to read and understand, to
`structure
`and
`simplify disclosure of
`important
`information. Our secondary objectives were to describe
`the informed consent process from the participant’s point
`of view, and to describe how participants assessed their
`own comprehension. Finally, we assessed providers’
`understandings of central elements of informed consent.
`To account for the many types of clinical research in
`cancer, and to
`identify differences
`in participants’
`comprehension by phase, we questioned participants in
`phase I, II, and III trials.
`
`Methods
`Participants
`We included only participants in trials that assessed a
`cancer-directed treatment (ie, not supportive care) and
`were phase I safety and dose-escalation trials, safety trials,
`phase II single-group efficacy trials, or phase III
`randomised controlled trials. All open trials were reviewed
`in advance to see if they met these criteria. Potential
`patients were identified by the quality control centre at the
`Dana-Farber Cancer Institute, which registers all patients
`enrolled in clinical trials at its affiliated institutions.
`
`1772
`
`THE LANCET • Vol 358 • November 24, 2001
`
`For personal use. Only reproduce with permission from The Lancet Publishing Group.
`
`AVENTIS EXHIBIT 2113
`Mylan v. Aventis, IPR2016-00712
`
`

`
`Individuals were eligible if they were aged 18 years or
`older and had signed consent to a qualified cancer trial at
`Dana-Farber Cancer Institute, Brigham and Women’s
`Hospital, or Massachusetts General Hospital, within the
`previous 14 days. We excluded people if consent had been
`obtained in a language other than English or by an
`investigator of this study, if their mailing address was
`outside the USA, or if they had died (n=3) or had been
`removed from the clinical trial within 14 days of signing
`consent. Enrolment took place from June 28, 1999, to Jan
`1, 2000. We sent questionnaires to all eligible patients,
`and did not require permission from the respondent’s
`physician. The Dana-Farber Cancer Institute IRB, which
`oversees all cancer-related research at the institutions
`participating in this project, approved the study protocol.
`
`Survey methods
`Surveys were mailed to the participant’s home or
`delivered to their hospital room 3–14 days after consent to
`participation in the clinical trial. If the completed survey
`was not returned within 2 weeks, a second questionnaire
`was sent, together with a card on which the patient could
`decline participation. 2 weeks later, we telephoned non-
`respondents to ensure receipt of the questionnaire and to
`answer any questions. If requested, we mailed a third
`questionnaire. Follow-up questionnaires were not sent to
`patients who had discontinued participation in their
`clinical trial. Concurrent with the initial mailing, we sent a
`brief questionnaire to the provider who had signed the
`consent form. If necessary, a second questionnaire was
`sent to the provider 14 days later.
`The QuIC21 consists of two parts. Part A, which
`measures the knowledge of participants, has 20 questions
`on the basic elements of informed consent specified in US
`federal regulations.1 Responses to individual questions are
`combined in a knowledge score, which ranged from 0
`(least) to 100. Four items included in our QuIC were not
`analysed because of absence of expert validation (n=1) or
`inapplicability to some phase I participants with curable
`cancers (3).21 Part B has 14 questions,
`in which
`participants rated their understanding of
`important
`elements of the trial on a 5-point scale. Responses were
`averaged and normalised from 0–100 to generate a self-
`assessment score. The questionnaire also
`included
`questions about the consent process (eg, time spent, when
`they signed the form, who was present), supplemental
`information sources (eg, internet), the consent form (eg,
`care in reading, clarity), previous participation in research,
`and demographic characteristics (eg, age, race, sex,
`education, marital status, first language). Additionally, we
`assessed respondents’ preferences for information and
`decisional
`involvement using modified Autonomy
`Preference Index subscales.23
`The questionnaire sent to providers included questions
`about the participant’s Eastern Cooperative Oncology
`Group (ECOG) performance status,24 outlook (likelihood
`of 5-year disease-free survival), disease status (newly
`diagnosed, relapsed, or progressive), and time since
`original diagnosis. The provider was asked to rate the
`participant’s overall understanding of the trial on a
`5-point scale.
`To clarify providers’ beliefs about the issues we had
`addressed with participants, at the conclusion of the study
`we sent the QuIC (Part A) to all providers whose patients
`had previously received our survey. We instructed them to
`complete the questionnaire as though they were fully
`informed patients on a clinical trial. We identified phase-
`specific questions and directed providers to answer them
`accordingly, whereas we instructed providers to answer
`
`ARTICLES
`
`generic questions without regard to phase. We calculated
`phase-specific knowledge scores for each provider with the
`same algorithm used for research participants.21
`We measured the length and Flesch reading ease25 of the
`consent form for each protocol included in our study, using
`the grammar function of Microsoft Word for Windows 97.
`The Flesch reading ease is a readability formula based on
`average sentence length and number of syllables per 100
`words; scores range from 0 (most difficult) to 100
`(easiest).25 We removed two pages of standard institutional
`language before analysis of each form. Also, we classified
`each consent form according to use of the NCI template.
`The Dana-Farber Cancer Institute IRB mandated that all
`newly submitted protocols use the NCI template as of
`March, 1999, but consent forms that had been previously
`submitted were not updated.
`
`Statistical analysis
`Bivariate associations with knowledge scores were
`assessed with t tests, ANOVA, or Pearson correlation
`coefficients. To ascertain which predictors were
`independently associated with respondents’ knowledge of
`their clinical trials, we developed a multiple linear
`regression model, with knowledge score as the dependent
`variable. Predictors that were significant (p<0·10) in
`bivariate analyses were entered into the original model.
`We then sequentially eliminated, in a backwards stepwise
`fashion, all variables for which p was 0·05 or greater.
`Because questionnaires were partly phase-specific, we
`kept indicator variables for phase in all models to control
`for potential confounding resulting from differences in the
`questionnaire itself. First, however, we calculated a
`summary
`score derived
`from phase-independent
`questions only and then verified that this generic score did
`not vary by phase (data not shown). Finally, we assessed
`the relation between providers’ responses to the QuIC and
`respondents’ summary scores using a separate linear
`regression model that accounted for within-provider
`clustering.
`A sample size of 200 participants was needed to achieve
`greater than 90% power (␣=0·05) to detect a difference of
`two-thirds SDs in knowledge score between respondents
`whose consent forms did and did not use the NCI
`template. We assumed that 15% of respondents would
`receive a template-based form. Analyses were done with
`Stata version 5.0.
`
`Results
`Questionnaires were mailed to 287 trial participants, of
`whom 207 (72%) responded. We received 240 of 287
`(84%) provider assessments. Respondents were from 73
`clinical trials and 77 providers. Respondents completed
`the questionnaire a median of 16 days after consent to
`their trials. The mean age of participants was 55·0 years
`(SD 12·7) with 23% (48 of 207) aged 65 years or older
`(table 1). Over half (92 of 175) for whom provider
`responses were available had relapsed or progressive
`cancer and almost two-thirds (111 of 171) had a 10% or
`less chance of 5-year disease-free survival (as recorded by
`their provider). A quarter were in phase I trials, half in
`phase II trials, and the remainder in phase III trials
`(table 1). More non-respondents than respondents were
`symptomatic (75% vs 55%, p=0·006), went off-protocol
`because of progression of disease or toxic effects within 60
`days of enrolling (25% vs 15%, p=0·05), and had newly
`diagnosed cancer (61% vs 47%, p=0·06). Non-
`respondents did not differ from respondents in age, sex,
`race, phase, provider-estimated prognosis, provider-
`estimated understanding, or use of NCI template.
`
`THE LANCET • Vol 358 • November 24, 2001
`
`1773
`
`For personal use. Only reproduce with permission from The Lancet Publishing Group.
`
`

`
`ARTICLES
`
`Patients
`(n=207)
`
`42 (20%)
`117 (57%)
`48 (23%)
`
`114 (55%)
`
`Characteristic
`Age (years)
`<45
`45–64
`⭓65
`Sex
`Women
`Race*
`White
`Hispanic
`African-Americans
`Asians
`Others
`College education†
`Married/living with partner‡
`Only English used at home§
`NCI template-based consent form
`Phase
`I
`II
`III
`Off-protocol for toxic effects, or progression within 60 days
`of consent
`Probability of 5-year disease-free survival ⭐10%¶
`Symptomatic from cancer (ECOG Performance Status >1)㛳
`Relapsed or progressive cancer㛳
`Physician estimate of understanding ⭐3 (5–point scale**)††
`*n=202 because of missing data. †n=203. ‡n=204. §n=204. ¶n=171 (of 175
`patients for whom provider questionnaires were returned). 㛳n=175. **1=didn’t
`understand trial at all, 5=understood trial very well. ††n=174.
`Table 1: Respondent characteristics
`
`184 (91%)
`6 (3%)
`6 (3%)
`4 (2%)
`2 (1%)
`107 (53%)
`156 (76%)
`194 (95%)
`64 (31%)
`
`50 (24%)
`103 (50%)
`54 (26%)
`31 (15%)
`
`111 (65%)
`97 (55)
`92 (53%)
`38 (22%)
`
`they were
`that
`Two (1%) respondents denied
`participating in a clinical trial. Table 2 summarises the
`remaining respondents’ descriptions of the consent
`process. Almost half of consent discussions lasted 1 h or
`longer. Participants signed the consent form a median of
`6 days (IQR 0–14) after their initial discussion; only 28%
`(56 of 200) reported signing at the first meeting. An adult
`friend or relative was present for 84% (170 of 202) of
`discussions, and a nurse for 39% (79 of 201). Few
`respondents gave consent as inpatients, and 14% (29 of
`203) reported previous participation in a clinical trial.
`Most felt that they, rather than their physician, had the
`main role in the enrolment decision. Many respondents
`sought additional information elsewhere. Most easily
`understood
`their consent
`forms, considered
`them
`important sources of information, and reported having
`read them carefully, but few judged the forms to be
`important to their decisions. Most reported having had
`adequate time to consider their decisions, having had
`sufficient opportunities to ask questions, and having
`received thorough answers. Most were satisfied with the
`informed consent process and reported that the decision
`to enrol was easy. Few respondents had felt pressure from
`their physician to participate, but most had felt urgency to
`begin treatment (table 2).
`The proportion of correct answers varied greatly across
`individual questions of the QuIC. Table 3 shows selected
`responses. A quarter of respondents did not agree that the
`main purpose of clinical trials is to benefit future patients.
`Many did not realise that the treatment being researched
`was not proven to be the best for their cancer, that the
`study used non-standard treatments or procedures, that
`participation might carry incremental risk, or that they
`might not
`receive direct medical benefit
`from
`participation. Most participants in phase III trials (48 of
`53) were aware that they were being randomly allocated to
`treatment, but fewer phase I participants (22 of 50) knew
`that their trial involved dose-escalation to assess toxic
`effects. Most respondents (170 of 204) recalled being
`
`offered alternatives to participation, and almost all knew
`that they could decline participation (202 of 204) or
`withdraw from the trial (183 of 204). More than half
`knew that outside parties might have access to their
`medical records because of trial participation. Appendix
`1, which shows complete responses to all questions, is
`available from The Lancet or from the authors on request.
`Mean knowledge score was 77·8 (SD 9·4). The
`knowledge score was not related either to time between
`signature of consent and completion of questionnaire, or
`to whether the respondent had already begun protocol
`treatment when he completed the questionnaire (data not
`shown). In bivariate analyses, higher scores were
`associated with college education, speaking only English
`at home, being white, receiving an NCI template-based
`consent form, not signing consent at the initial discussion,
`presence of a nurse at the consent discussion, and
`supplemental use of pamphlets, internet, magazines, or
`books (table 4). Respondents who were symptomfree, had
`reported that they had read the consent form carefully,
`had had adequate time to decide, and thought they had
`had sufficient opportunity for questions had higher scores
`than those who did not. Knowledge scores correlated
`weakly with consent form readability as measured by
`the Flesch Reading Ease (r=0·13, p=0·07). Sex, age,
`previous
`research participation,
`length of consent
`discussion, marital status, consent obtained by the
`trial’s principal investigator, phase, time since diagnosis,
`relapse status, and physician-estimated outlook, among
`other factors, were not significantly associated with
`knowledge scores.
`factors were
`six
`In
`the multivariate model,
`independently associated with
`improved knowledge
`scores—college education (␤=5·2, 95% CI 2·8–7·6), use
`of only English at home (10·0, 4·6–15·3), use of the NCI
`template (3·0, 0·2–5·8), not signing the consent form at
`the initial discussion (3·0, 0·3–5·7), presence of a nurse at
`the consent discussion (2·5, 0·1–5·0), and careful reading
`of the consent form (3·9, 0·7–7·2).
`Among participants in phase III trials, those randomly
`allocated to standard groups probably misinterpreted two
`questions in the QuIC to refer to their own groups rather
`
`Questions
`Consent discussions lasted ⭓1 h
`Signed consent form at first discussion
`Presence of adult friend or relative at consent discussion
`Presence of nurse at consent discussion
`Inpatient during consent
`Participation in previous clinical trial
`Enrolment decision made mainly by respondent
`Consulted pamphlets for supplemental information
`Consulted outside physician
`Sought information on internet
`Consulted books or magazines
`Discussed trial with other patients
`Consent form read carefully†
`Consent form important source of information†
`Consent form easy to understand†
`Consent form important to the decision†
`Enough time to learn about trial†
`Pressure from provider to sign consent form†
`Sufficient opportunity to ask questions†
`Questions answered thoroughly†
`Satisfied with informed consent process†
`Treatment needed to begin as soon as possible†
`Decision to participate easy or very easy
`
`Patients
`(n=205)*
`
`97 (48%)
`56 (28%)
`170 (84%)
`79 (39%)
`11 (5%)
`29 (14%)
`151 (74%)
`97 (49%)
`88 (44%)
`84 (42%)
`55 (28%)
`42 (21%)
`170 (84%)
`149 (73%)
`173 (86%)
`76 (37%)
`177 (87%)
`6 (3%)
`190 (93%)
`189 (92%)
`185 (90%)
`183 (90%)
`152 (75%)
`
`Because of missing data, denominators do not all equal 205. *Two
`respondents denied participation in clinical trials and are excluded from these
`data. †Proportion of subjects who responded agree or strongly agree.
`Table 2: Participants’ reports of the informed consent process
`
`1774
`
`THE LANCET • Vol 358 • November 24, 2001
`
`For personal use. Only reproduce with permission from The Lancet Publishing Group.
`
`

`
`ARTICLES
`
`Paticipants’ responses (n=205) Providers’ responses* (n=61)
`
`Disagree Unsure
`
`Agree
`
`Disagree Unsure
`
`Agree
`
`31 (15%) 20 (10%) 153 (75%) 15 (25%) 18 (30%) 28 (46%)
`
`Questions
`The main reason cancer clinical trials are done is to improve the treatment of future
`cancer patients†
`All the treatments and procedures in my clinical trial are standard for my type of
`cancer‡
`The treatment being researched in my clinical trial has been proven to be the best
`treatment for my type of cancer‡
`Compared with standard treatments for my type of cancer, my clinical trial does not
`carry any additional risks or discomforts‡
`There may not be direct medical benefit to me from my participation in this clinical
`trial†
`Values are number (%). Participants’ and providers’ responses to all items on the QuIC are available from The Lancet or from the author. *Providers were asked to
`answer what they believed was right—ie, what a fully and accurately informed patient would answer. Four providers left questions 1–5 blank and wrote in comments
`suggesting that the appropriate response was context-dependent. These were coded as unsure. †Correct answer=agree. ‡Correct answer=disagree.
`Table 3: Participants’ and providers’ responses to selected questions on the quality of informed consent (QuIC)
`
`52 (26%) 53 (27%)
`
`95 (48%) 49 (80%) 11 (18%) 1 (2%)
`
`62 (30%) 82 (40%)
`
`60 (29%) 50 (82%) 10 (16%) 1 (2%)
`
`75 (37%) 52 (26%)
`
`77 (38%) 43 (71%) 17 (28%) 1 (2%)
`
`33 (16%) 25 (12%) 145 (71%)
`
`1 (2%)
`
`2 (3%) 58 (95%)
`
`than to the trial as a whole. Responses to “all the
`treatments and procedures in my clinical trial are standard
`for my type of cancer” varied by group, with those in the
`intervention group (disagree=10, unsure=9, agree=10)
`more likely to be correct than those in the standard group
`(disagree=10, unsure=1, agree=18; p=0·005). Similarly,
`responses to “compared with standard treatments for my
`type of cancer, my clinical trial does not have any
`additional risks or discomforts”, with those in the
`intervention group (disagree=10, unsure=12, agree=8)
`more likely to be correct than those in the standard group
`(disagree=6, unsure=3, agree=14; p=0·03). We therefore
`
`Knowledge score p value
`(n=204)*
`
`Answers
`Ethnic origin
`White
`Other
`Education
`Not College-educated
`College-educated
`Language
`English
`Other (with or without English)
`NCI Template
`No
`Yes
`Signed at first discussion?
`No
`Yes
`Presence of nurse at consent discussion
`No
`Yes
`Pamphlets
`No
`Yes
`Use of Internet
`No
`Yes
`Magazines/books
`No
`Yes
`Symptomatic from cancer
`No
`Yes
`Consent form read carefully†
`No
`Yes
`Enough time to decide†
`No
`Yes
`Sufficient opportunity to ask questions†
`No
`Yes
`
`78·2 (9·4)
`72·2 (7·8)
`
`74·9 (8·3)
`80·2 (9·6)
`
`78·2 (9·3)
`69·3 (7·6)
`
`76·7 (9·4)
`80·1 (9·0)
`
`79·2 (9·5)
`74·2 (8·1)
`
`76·2 (9·5)
`80·0 (8·7)
`
`76·5 (9·5)
`79·4 (9·1)
`
`76·2 (9·3)
`79·6 (9·2)
`
`76·6 (9·7)
`80·9 (7·9)
`
`79·3 (9·5)
`76·6 (9·2)
`
`73·4 (8·4)
`78·7 (9·3)
`
`73·8 (9·1)
`78·5 (9·3)
`
`73·8 (8·9)
`78·1 (9·4)
`
`0·01
`
`<0·0001
`
`0·003
`
`0·01
`
`0·0005
`
`0·005
`
`0·03
`
`0·01
`
`0·004
`
`0·07
`
`0·003
`
`0·02
`
`0·09
`
`Data are means (SDs). *Could not be calculated for one patient because of
`missing data. †Subjects who responded agree or strongly agree.
`Table 4: Correlates of improved knowledge score in cancer
`clinical trial participants
`
`reanalysed the data after adjusting for this difference
`between groups. Neither bivariate nor multivariate
`analyses were substantially changed.
`The median self-assessment score was 89·3 (IQR
`82·1–96·4). A weak but
`significant
`correlation
`(Spearman’s r=0·25, p=0·0004) was recorded between
`respondents’ knowledge and self-assessment scores. Of a
`maximum possible rating of 5, the median provider rating
`of respondent understanding was 4 (IQR 4–5); one-fifth
`of respondents were rated 3 or less (table 1). Providers’
`ratings correlated weakly with respondents’ knowledge
`scores (␳=0·23, p=0·003).
`the QuIC after
`Of 91 providers who received
`completion of the patient survey, 61 (67%) responded.
`Table 3 shows selected answers for comparison with
`participants’ responses. Fewer than half agreed that the
`main reason trials are done is to improve treatment of
`future patients. Additionally, up to a third of providers
`were uncertain whether clinical trials always use non-
`standard treatments or procedures, whether treatments
`assessed in clinical trials are by definition unproven, and
`whether trials involved some, however minor, incremental
`risk or discomfort. Most agreed that research participants
`might not benefit from participation. For phase I, II, and
`III trials, respectively, the mean providers’ summary
`scores were 92·6 (SD 6·2), 91·9 (7·1), and 92·5 (6·5).
`Providers’ scores did not predict the knowledge scores of
`individual respondents whose consent they had obtained
`(data not shown). Appendix 2 (available from the Lancet
`or from the author) shows complete provider responses to
`all questions.
`
`Discussion
`We investigated informed consent in clinical trials of
`cancer, using a questionnaire that was designed to assess
`the elements required by US federal regulations.1 On
`average, providers spent much time with participants, and
`most participants took several days to consider their
`decisions. Few found their decisions difficult, almost none
`reported coercion, most were satisfied with the consent
`process, and most felt that they understood their trials
`well. Most respondents had used additional sources of
`information and had support from family or friends.
`Nonetheless, knowledge varied widely and there were
`important misunderstandings. Major
`deficiencies
`included not being aware of non-standard treatment, the
`potential for incremental risk or discomfort, the unproven
`nature of treatment, and the uncertainty of benefits to self.
`These problems characterise what Appelbaum and
`colleagues18 have
`referred
`to as “the
`therapeutic
`misconception”.
`Lower knowledge scores were associated with absence
`of college education and use of languages other than
`
`THE LANCET • Vol 358 • November 24, 2001
`
`1775
`
`For personal use. Only reproduce with permission from The Lancet Publishing Group.
`
`

`
`ARTICLES
`
`English in the home.9,11 Although there were few non-
`native English speakers, their reduced scores are of great
`concern. Efforts are needed to ensure that their consent is
`adequately informed, perhaps by expanding the use of
`interpreters and translated consent forms.
`Participants who were given consent forms that used the
`NCI template had higher knowledge scores than those who
`received other forms, suggesting that structure and
`simplification of such consent forms might be effective.
`Respondents who had time to consider participation and
`those who had a nurse present at their consent discussions
`were also more knowledgeable than those who did not.
`Previous studies9,26,27 lend support to these results. Finally,
`respondents who reported reading their consent forms
`carefully achieved higher knowledge scores than those who
`did not. The quality of informed consent might thus be
`substantially enhanced by addressing these deficiencies.
`Contrary
`to expectations,
`several
`factors were
`unassociated with
`respondents’
`knowledge
`after
`adjustment for confounders in the multivariate model.
`Knowledge scores were not lower among elderly patients,
`those who were sicker, or those who had poorer outlooks.
`Neither readability of the consent form nor relapse status
`affected the knowledge scores. Despite differences in the
`circumstances of enrolled individuals, knowledge scores
`did not differ by phase of trial, although participants of
`phase III trials were more aware of purposes and
`procedures than those in phase I or II trials, and phase I
`participants were more likely to recognise the possibility of
`incremental risk than other participants. These findings
`contrast with those of previous investigators, possibly
`because of differences in study populations or outcome
`measures.9–11 In particular, we emphasised conceptual
`issues rather than trial details, a strategy that should
`facilitate cross-setting comparisons in the future.
`Many providers seemed to share the same therapeutic
`misconceptions as participants. Despite all providers
`surveyed being on the academic staff at research
`institutions, fewer than half recognised that the main
`purpose of clinical trials is to benefit future patients.
`Although individual providers’ uncertainties did not
`correlate with lower scores in their patients, our findings
`support Miller’s28 assertion that physician/investigators
`often deal with the moral tensions inherent in their role by
`adopting the perspective of the pure clinician. Research
`ethics rest on the realisation that the goals of advancing
`science or treatment, however noble, could conflict with
`the interests of present patients.29 That providers and
`investigators appreciate this conflict is essential if they are
`to help participants distinguish research goals from
`therapeutic intentions.7,30 Thus, education about the
`dilemmas of clinical research, and efforts to provide a
`coherent professional identity for physician/investigators,
`are needed.28
`Although our results suggest the need for improvements
`in informed consent to research, they also point to its
`complexity
`in the setting of cancer clinical trials.
`Physicians often recommend that patients with cancer
`enrol in trials because they feel that trials offer the best
`therapeutic
`option
`under
`the
`circumstances.
`Furthermore, patients are increasingly demanding access
`to research for similar reasons.31 Thus, beliefs about best
`interests could explain why, in our survey, many patients
`were entered in trials despite providers’ concerns that
`patients might not fully understand the implications of
`participation. How best to reconcile legitimate hopes for
`benefit with the need to help research participants
`understand central concepts of clinical research remains
`an essential unanswered question in research ethics.
`
`Several limitations to this study merit discussion. First,
`the study investigated three affiliated institutions, the
`characteristics of which could differ from other sites.
`However, in view of the evidence of attention to process
`and the high satisfaction in respondents, the problems we
`identified are likely to be widespread. Second, response
`bias could have been introduced. The few differences we
`detected between respondents and non-respondents
`suggest that non-respondents were more acutely ill. Any
`resulting bias is therefore likely to be conservative, with
`respondents both being more knowledgeable and having a
`more positive attitude than non-respondents. Third, our
`questionnaire addressed mainly conceptual
`issues
`associated with clinical research. We did not ask
`respondents to reiterate details of risks, procedures, or
`other technical issues. Also, there was little ethnic
`diversity in our sample, and we recommend further study
`in more varied populations.
`We did not have enough respondents in each phase to
`draw phase-specific conclusions. Also, there might be
`disagreement about whether some items on the QuIC fit
`the special circumstance of a randomised trial that
`compares two accepted therapies. In our study, however,
`all phase III trials compared one or more investigational
`groups to a standard group. Our data also suggest that
`some respondents who were enrolled in the standard
`group of a phase III trial might have misinterpreted two
`questions to refer to their own group rather than to the
`trial as a whole. Because participants in cancer clinical
`trials are generally aware of treatment assignment, we
`intend to change those questions to reduce the likelihood
`of such misinterpretation.
`Respondents filled out the questionnaire a median of 16
`days after consenting to their clinical trials and their
`understanding of the concepts we measured might have
`been better at the time of consent than at this later time.
`Furthermore, information that contradicts the therapeutic
`misconception might be difficult for patients to assimilate
`in the context of their natural hopes and anxieties. Finally,
`because we did not directly observe the consent process,
`we cannot address whether deficiencies were due to
`providers’ failures to discuss certain
`issues, or to
`respondents’ lack of recall. We suspect that the quality of
`written and verbal communication about clinical trials
`varied greatly. Additional studies that directly monitor the
`consent process and then assess how variability affects
`participants’ comprehension would be valuable.
`We recorded important flaws in research participants’
`understandings of their cancer clinical trials, despi