IRB: Ethics & Human Research
The Evolution of Consent Forms for Research: A Quarter Century of Changes
The informed consent process in the research context typically relies heavily on information provided to potential research participants in a written consent form. Federal regulations governing research with humans list the elements of information that consent forms must contain and indicate that there may be additional elements relevant to particular types of studies.1Consent forms typically serve as the participants’ only written explanation of the studies in which they have enrolled and as important documentation for researchers and regulators of individuals’ consent to participate in a study. Institutional review boards (IRBs) often pay considerable—and some would say excessive—attention to the content and wording of these forms.2
Despite the critical role of consent forms in the research context, there is good reason to believe that, as currently formulated, they are often far from ideal.3Among the problems that have been documented are excessive length,4complexity of wording that exceeds average reading levels,5and problems with the accuracy with which key aspects of the studies are described.6Numerous proposals have been made for improving consent forms,7but there is little evidence that attempts at improvement have had a positive effect. Indeed, we have been unable to find sufficient data examining trends in consent forms over time to determine whether most of these concerns mentioned above have been adequately resolved. To our knowledge, the only available longitudinal data focus on changes in the length of forms, and the data suggest that things may be getting worse instead of better. For instance, Baker and Taub demonstrated that the mean length of consent forms nearly doubled between 1975 and 1982.8More recently, Beardsley and colleagues in Australia found that the median length of consent forms increased from seven to 11 pages between 2000 and 2005.9These findings are of particular concern because of data suggesting that greater consent form length is associated with less understanding by potential participants of key information related to the study in which they are being recruited to participate.10
Given the importance for researchers, IRBs, and policy-makers of understanding the extent to which the documented problems with consent forms have been ameliorated or exacerbated over time, the goal of this study was to explore changes over a quarter century in the accuracy and length of research consent forms used by one department in a major academic center.
Data spanning a 25-year period (1978–2002) were obtained from a sample of protocols and consent forms drawn from the files of the IRB serving a department in a major academic medical center. The sampling frame began with the earliest examples of consent forms and continued for the 25-year period covered by this study. Protocols were sampled in three-year increments, with a systematic selection of every third protocol in those years. Exclusion criteria were protocols submitted for periodic review after initial IRB approval, protocols that did not include consent forms, and protocols that never received final approval from the IRB whose files we examined. Of the 240 consent forms sampled, 215 met eligibility criteria (ranging from 19 to 29 protocols in each of the years sampled) and were included in this study.
From each consent form, we recorded the page length of the consent form, year of IRB approval, and risks to the participant. From the protocol associated with each consent form, we recorded statements of the specific risks and overall degree of risk for participants.
Descriptions of risks in consent forms and protocols were coded in the following categories: no mention of risk; no risk; social/legal/confidentiality risks; upset or discomfort upon questioning; minimal discomfort (e.g., pain from drawing blood, discomfort from electroencephalogram [EEG]); moderate discomfort (e.g., moderate nausea or vomiting, headache, slight elevation in blood pressure, drowsiness); significant discomfort that could require medical attention (e.g., fracture of bones, developing tardive dyskinesia); potentially life-threatening condition (e.g., heart attack, stroke); and safety profile unknown, but test drug or device potentially harmful. When multiple categories of risk were indicated in a consent form or protocol, only the highest applicable level of risk, in the order indicated above, was coded.
All data extraction and coding were performed by the first author. Variables were entered into an Excel database for analysis. Consistency of statements in consent forms was determined by comparing the coding for consent forms with the coding for the associated protocols. Statistical analyses were performed using SAS release 9 (SAS Institute Inc, Cary, NC), with P < 0.05 considered statistically significant.
Of the 215 protocols eligible for inclusion in this study, 102 (47.4%) were for clinical trials, 76 (35.3%) for observational studies, 23 (10.7%) for studies involving physical interventions, and 14 (6.5%) for studies of psychosocial or behavioral interventions. These protocols represented studies involving inpatients exclusively (n = 37, 17.2%), outpatients exclusively (n = 123, 57.2%), healthy volunteers (n = 30, 14.0%), and both inpatients and outpatients (n = 25, 11.6%). Thirty-five studies (16.3%) were conducted exclusively in pediatric populations, eight studies (3.7%) involved both minors and adults, and 172 studies (80.0%) involved adults exclusively.
Length of Consent Forms. The length of the consent form increased roughly linearly by an average of 1.5 pages per decade. In the 1970s, the average consent form was less than one page long and often only a paragraph or two, but by the mid-1990s the average form had increased in length to over 4.5 pages (Figure 1). The mean and median page length of the consent forms over the entire course of the study was 3.03 and 3.00, respectively. No year sampled had consent forms with a mean length that was shorter than the forms from three years earlier.
A multivariate logistic regression analysis was performed, with length of forms dichotomized at two pages or less and greater than two pages, including those variables that had shown significant effects in bivariate analyses: type of study, degree of risk, and year. Significant effects were found for study year (odds ratio = 1.56 [C.I. 1.38, 1.78] for each additional year since the start of the window) and risk of the study (odds ratio = 10.98 [C.I. 3.64, 33.11] for moderate- to high-risk studies compared to no-risk and minimal-risk studies); type of study was not a significant predictor and was dropped from the model.
Descriptions of Risks. Discrepancies between the descriptions of the highest level of risks for participants in the consent form as compared with the protocol were found in 25.6% (n = 52) of studies. Fifty percent (n = 26) of discrepancies involved studies with minimal risks, while 50% (n = 26) related to studies with moderate or greater risks. In 45 of 52 studies with discrepancies, a higher level of risk was indicated in the protocol than in the consent form.
At the start of the study in 1978, more than 54% of protocols demonstrated discrepancies in the description of risks; a decided downward trend ended with 0% of protocols revealing such discrepancies in 1999 through 2002. Logistic regression revealed an odds ratio of 0.84 (p < 0.0001) for the presence of a discrepancy for each additional year beyond the beginning of the sampling frame. Thus, the odds of discrepancies in the description of risks in the consent forms decreased by an average of approximately 16% per year (Figure 1). Similarly, a statistically significant association existed between a length of less than two pages and risk discrepancy, with 43.27% of forms two pages or less having discrepancies as compared with 9.26% of forms that were more than two pages (p < 0.001). None of the independent variables—including study type, type of patient, protocol risk, and year—were associated with discrepancies, hence a multivariate analysis was not performed.
Of note, in studies from 1978, there was frequently no mention at all of risks in either the consent form or the protocol. However, 11 of the 23 studies sampled were studies of medications, which invariably are associated with risks of adverse effects, and one protocol described the study as involving a “pain-inducing procedure” but did not include this information in its consent form. One example of a discrepancy in the presentation of risks was seen in a 1978 drug study involving inpatients. The protocol stated, “The one potential risk is that [X drug] can be psychologically habit-forming,” but the consent form did not identify any risks. Another placebo-controlled medication study in 1978 stated in the protocol that since the study evaluates “various drug treatments” and since the population is depressed, there is a risk “that depressed patients will become more depressed and suicidal” during the placebo period. However, the consent form stated clearly, “there is no risk.”
Although the most striking examples of discrepancies in the presentation of risks were seen in 1978, such discrepancies continued to appear over the next two decades. A medication study in 1981 stated in the protocol that there was a risk of a “hypertensive crisis or lack of clinical improvement,” while the consent form, which noted some potential side effects of the medication, said nothing about these risks. The protocol of another 1981 medication study noted that “suicide is the greatest risk,” yet the consent form was silent regarding the risk of suicide. For a 1990 medication study, the protocol included risk information that stated the “most common side effect is dry mouth . . . rare side effects include low platelet count [and] seizures,” while the consent form stated, “there are no anticipated psychological or physical risks.”
Another discrepancy in the characterization of risks was seen in a 1993 drug study where the protocol listed three risks: “first of suicide, second risk is that patients will not improve (placebo), third is risk of side effects, including dry mouth, heart conduction problems.” The consent form stated that the “risks to me are the side effects (dry mouth, and heart conduction problems), possibility that symptoms of depression could worsen,” but not suicide. Another drug study in 1993 stated in the protocol that “risks of drug include: tremor, nausea, overdosing, risk of placebo, and development of suicidal ideation,” while the consent form noted “side effects include: nausea, tremor, headache,” but nothing about the development of suicidal ideation.
Few previous studies have examined changes in the nature of consent forms over time, and none over a period of 25 years. Two trends were notable in our study: greater consistency in the description of risks, and an increase in the length of consent forms. The odds of discrepancies between the descriptions of risks in protocols and consent forms decreased by an average of 16% per year and finally disappeared after 1996. Consent forms grew in length by an average of 1.5 pages per decade.
The decrease in discrepancies and increase in page length suggest a tendency over time toward a greater focus by researchers and/or IRBs on the accuracy and completeness of the information provided to prospective research participants. That is, in more recent years information in the consent form has become more detailed and precise. However, by providing more detail and precision, consent forms became longer. The entire informed consent document evolved from a paragraph with information that frequently appeared to have been inaccurate and/or missing, to multiple pages with thorough descriptions of the risks and other details of the study. These changes occurred slowly and steadily over time, representing a gradual evolution of the type and amount of information contained in consent documents. The increase in page length suggests that additional language was added to clarify risks, though exactly where the additional wording was added was not determined in this study. Although discrepancies in the descriptions of risks between the consent form and the protocol had ceased before the end of our study window, page length continued to increase, suggesting that greater attention to risks has not been the sole factor responsible for the increase in page length.
We can only speculate about the drivers of these changes, but it is interesting to note that the relevant research regulations in 1974 included the same basic eight elements of informed consent that are in the current version, including the requirement that potential participants be provided with “a description of any reasonably foreseeable risks or discomforts to the subject.” Why, then, were there so many discrepancies in the disclosure of risks in 1978 and 1981, and why did it take so many years for the discrepancies to disappear? The slow implementation of the regulations is consistent with the initial low level of oversight of the decentralized system for review of human subjects research created by the regulations. Incentives for strict compliance by institutions and investigators were minimal and increased only as federal oversight became more extensive and as regulatory staff and IRB members became more experienced. Increased public attention to human subjects research—including allegations of failure to disclose risks in research1112 and more frequent litigation over alleged negligence in the conduct of research—are other possible explanations for increased accuracy in consent forms, especially during the latter part of the study window.
One of the limitations of this study is that the data are drawn from a single IRB, yet the research enterprise is characterized by variation across IRBs regarding approaches to implementing the federal regulations.13In addition, since the records of the IRB we examined deal specifically with studies conducted by members of a single department, they may not be characteristic of other kinds of medical research. By focusing on the highest level of risk disclosed, the data do not allow us to comment on the extent of discrepancies in other study risks, although it is our impression that they diminished in a similar way over time. Moreover, although the coding categories were straightforward, coding was performed by a single rater without a test of reliability. Finally, while this study illustrates the evolution of the consent form itself, it cannot capture the oral portion of the informed consent process. Hence, we cannot comment on whether the total package of information received by a potential research participant was more or less accurate and comprehensive than is suggested by these data.
Our findings highlight the inherent paradox in attempting to use consent forms to convey ever-more-complete information to potential research subjects. Greater information is associated with increased length of consent forms, and studies have shown an inverse relationship between length and individuals’ comprehension of the information provided.14For instance, Sharp has pointed to data suggesting that consent forms longer than 1,000 words (four double-spaced pages) are unlikely to be read, perhaps in part because of the time involved.15He recommends that length be limited to no more than 1,250 words (five pages), which would take an average high school graduate five to seven minutes to read. In many cases for that to be accomplished, other means would have to be used to serve the informational and legal interests of potential participants, investigators, and institutions. These may include supplementary informational booklets that potential participants can read at their leisure, video or computer-based disclosures that would supplement written documents, and disclosing in consent forms only the most important information.
This study’s findings—particularly the increasing length of consent forms—are problematic given other data showing that the longer a form, the less likely subjects are to read and understand it. Thus, continuing to use current approaches—which have led to steadily larger amounts of information being included in consent forms—is unlikely to be effective in achieving informed decision-making by research subjects. Innovative approaches are likely needed as we continue to strive for genuinely informed consent.
We would like to thank Cathy Gere for her insightful comments.
nIlene Albala, AB,is a student at the University of Pennsylvania School of Law, Philadelphia, PA;Margaret Doyle, MPH, is a statistician in the Department of Neurology, Columbia University, New York, NY; andPaul S. Appelbaum, MD, is the Elizabeth K. Dollard Professor of Psychiatry, Medicine and Law, and Director of the Division of Law, Ethics, and Psychiatry, Department of Psychiatry, Columbia University, New York, NY.
1. 45 CFR 46.101-119.
2. May T, Craig JM, Spellecy R. Viewpoint: IRBs, hospital ethics committees, and the need for “translational informed consent.”Academic Medicine 2007;82(7):670-674; Paasche-Orlow MK, Taylor HA, Brancati FL. Readability standards for informed consent forms as compared with actual readability.NEJM 2003;348(8):721-726; Whitney SN, Alcser K, Schneider C, et al. Principal investigator views of the IRB system.International Journal of Medical Sciences 2008;5(2):68-72.
3. Annas GJ. Reforming informed consent to genetic research.JAMA 2001;286(18):2326-2328; Flory J, Emanuel E. Interventions to improve research participants’ understanding in informed consent for research: A systematic review.JAMA 2004;292(13):1593-1601; Kahn JP, Mastroianni AC. Moving from compliance to conscience: Why we can and should improve on the ethics of clinical research.Archives of Internal Medicine 2001;161(7):925-928.
4. Beardsley E, Jefford M, Mileskin L. Longer consent forms for clinical trials compromise patient understanding: So why are they lengthening?Journal of Clinical Oncology 2007;25(9):e13-e14; LoVerde ME, Prochazka AV, Byyny R. Research consent forms: Continued unreadability and increasing length.Journal of General Internal Medicine 1989;4:410-412.
5. Christopher PP, Foti ME, Roy-Bujnowski K, et al. Consent form readability and educational levels of potential participants in mental health research.Psychiatric Services 2007;58(2):227-232; Jackson RH, Davis TC, Bairnsfather LE, et al. Patient reading ability: An overlooked problem in health care.Southern Medical Journal 1991;84:1172-1175; see ref. 2, Paasche-Orlow et al. 2003; Williams BF, French JK, White HD, et al. Informed consent during the clinical emergency of acute myocardial infarction (HERO-2 consent substudy): A prospective observational study.Lancet 2003;361:918-922.
6. Resnik DB, Peddada S, Altilio J, et al. Oncology consent forms: Failure to disclose off-site treatment availability.IRB: Ethics & Human Research 2008;30(6):7-11; Burman W, Breese P, Weis S, et al. The effects of local review on informed consent documents from a multicenter clinical trials consortium.Controlled Clinical Trials 2003;24:245-255; see ref. 4, Beardsley 2007.
7. See ref. 3, Flory 2004; Ryan R, Prictor M, McLaughlin KJ, Hill S. Audio-visual presentation of information for informed consent for participation in clinical trials.Cochrane Database of Systematic Reviews 2008;(1):1-52.
8. Baker NT, Taub HA. Readability of informed consent forms for research in a Veterans Administration medical center.JAMA 1983;250:2646-2648.
9. See ref. 4, Beardsley 2007.
10. Epstein LC, Lasagna L. Obtaining informed consent.Archives of Internal Medicine 1969;123:682-688; Mann T. Informed consent for psychological research: Do subjects comprehend consent forms and understand their legal rights?Psychological Science 1994; 5(3):140-143.
11. Hilts PJ. Agency faults a U.C.L.A. study for suffering of mental patients.New York Times , March 10, 1994.
12. Mello MM, Studdert D, Brennan T. The rise of litigation in human subjects research.Annals of Internal Medicine 2003;139:40-45.
13. DeRenzo EG. The ethics of involving psychiatrically impaired persons in research.IRB: Ethics & Human Research 1994;16(6):7-11; Goldman KJ. Inconsistency and institutional review boards.JAMA 1982;248(2):197-202; Sachs GA, Hougham GW, Sugarman J, et al. Conducting empirical research on informed consent: Challenges and questions.IRB: A Review of Human Subjects Research 2003;25(5);S4-S10; Williams PC. Success in spite of failure: Why IRBs falter in reviewing risks and benefits.IRB: A Review of Human Subjects Research 1984;6(3):1-4.
14. See ref. 10, Mann 1994.
15. Sharp M. Consent documents for oncology trials: Does anybody read these things?American Journal of Clinical Oncology 2004;27(6):570-575.
Ilene Albala, Margaret Doyle, and Paul S. Appelbaum, “The Evolution of Consent Forms for Research: A Quarter Century of Changes,”IRB: Ethics & Human Research 32, no 3 (2010): 7-11.