A randomized comparison of two-stage versus traditional one-stage consent for a low-stakes randomized trial

Abstract

Background/Aims:

It has been proposed that informed consent for randomized trials should be split into two stages, with the purported advantage of decreased information overload and patient anxiety. We compared patient understanding, anxiety and decisional quality between two-stage and traditional one-stage consent.

Methods:

We approached patients at an academic cancer center for a low-stakes trial of a mind-body intervention for procedural distress during prostate biopsy. Patients were randomized to hear about the trial by either one- or two-stage consent (n=66 vs. n=59). Patient-reported outcomes included Quality of Informed Consent (QuIC, 0 – 100); general and consent-specific anxiety; decisional conflict, burden and regret.

Results:

QuIC scores were non-significantly higher for two-stage consent, by 0.9 points (95% CI - 2.3, 4.2, p=0.6) for objective and 1.1 points (95% CI −4.8, 7.0, p=0.7) for subjective understanding. Differences between groups for anxiety and decisional outcomes were similarly small. In a post-hoc analysis, consent related anxiety was lower among two-stage control patients, likely because scores were measured close to the time of biopsy in the two-stage patients receiving the experimental intervention.

Conclusion:

Two-stage consent maintains patient understanding of randomized trials, with some evidence of lowered patient anxiety. Further research is warranted on two-stage consent in higher stakes settings.

Introduction

In two-stage consent, originally termed “just-in-time” consent, informed consent for a randomized trial is split into consent for research procedures – randomization, questionnaires, use of data and the like – and separately, consent for allocated intervention.¹ Splitting the consent process is rendered unavoidable by some research designs, such as for “Trials Within Cohorts” (TWiCs),² where patients first join a research cohort and are then randomly offered experimental treatments. However, we argue that not only is two-stage consent ethically acceptable when logistically necessary, but ethically superior when logistically possible. In particular, we identified numerous problems with the traditional one-stage consent, where patients are informed about research procedures and then the possible harms and benefits of all interventions being compared in a trial, even those they are destined never to receive. These problems include information overload, confusion, undue decisional anxiety and decisional burden, as well as the “disappointment effect” of being randomized to control in a trial of a novel intervention. We emphasized that our arguments were empirically testable and called for appropriate trials.¹

Subsequently, we published the results of a single-arm study of two-stage consent.³ We implemented two-stage consent in a low-stakes trial, defined as one in which the decision whether or not to participate in the trial would make only a small difference to health. The trial concerned a brief mind-body exercise (guided mindfulness meditation) for procedural pain and distress before and during prostate biopsy (hereafter “mindfulness trial”). Accrual was very rapid with 98% of patients approached for the mindfulness trial agreeing to take part in the consent study, and scores on the standard “Quality of Informed Consent” (QuIC) questionnaire were close to those reported in studies of more traditional approaches to consent.⁴ We also found that some items on the QuIC were misleading in the context of two-stage consent. Our conclusion was that two-stage consent appeared promising and was worthy of further research in a randomized comparison with traditional, one-stage consent.

The prostate biopsy trial continued to accrue for the primary study question concerning the effectiveness of the mind-body intervention on procedural distress. We modified the mindfulness trial protocol such that patients being considered for participation were randomized to either two-stage or traditional one-stage consent (hereafter “consent trial”). Here we report the findings of this randomized comparison between different approaches to informed consent. The current trial was planned as part of a trio of studies on one vs. two-stage consent. Three trials are planned in three different settings, going from low stakes – this study on mindfulness meditation for procedural distress – to high stakes, such as a trial of chemotherapy for advanced cancer. We started with this low stakes trial on the grounds that, were two-stage consent be demonstrated to reduce patient understanding of the trial, this would not lead to important patient harm.

Methods

The mindfulness trial has been previously described.³ In brief, prostate biopsy is often uncomfortable, with approximately 1 in 3 men reporting at least moderate discomfort.^{5, 6} Guided mindfulness meditation has been shown to be of benefit for procedural distress for breast biopsy.⁷ We therefore considered whether this technique would be of benefit for prostate biopsy, and randomized patients scheduled for prostate biopsy to the meditation exercise or usual care control.

The overall study schema is given in Figure 1. The study cohort consisted of patients who were on an active surveillance protocol at Memorial Sloan Kettering Cancer Center (MSKCC). Such patients have regular prostate biopsies to assess if their low-risk prostate cancer has progressed and therefore requires intervention. The date of the biopsy is known well in advance, allowing us to approach patients for consent about 6 – 8 weeks before their scheduled biopsy. Patients were telephoned by a Clinical Research Coordinator and asked if they would be prepared to be randomized to different types of consent and then answer a brief questionnaire about their experience. The script used is given in the Supplementary Material. For this minimal risk consent trial, the MSKCC IRB agreed that verbal consent was appropriate. Patients who did not consent were offered participation in the mindfulness trial using the two-stage approach; those giving such consent were randomized to one- or two-stage consent using the MSKCC Clinical Research Database, a secure database that ensures full allocation concealment.

Figure 1.

Schema of trial

Patients randomized to the traditional one-stage consent process were then approached for the mindfulness trial using a traditional informed consent. They were told both about research procedures, such as randomization, and the intervention, being informed that they had a 50:50 chance of receiving the experimental treatment. This was a mindfulness intervention, which involved being given headphones connected to an audio player in the clinic area, listening to a 10-minute pre-biopsy mindfulness exercise and then a second 10-minute exercise that guides the patient during the biopsy procedure. It was also explained that the outcome of the mindfulness trial was based on questionnaires about pain, discomfort, anxiety and tolerability that are given as a routine part of clinical care, such that participation in the trial did not require any additional forms, visits, tests or procedures.

Patients randomized to two-stage consent were similarly told that the trial was about the experience of prostate biopsy and that they were being asked to give consent to have routinely collected data made available to researchers. They were also told that if they agreed to take part, they may be selected at random to be offered “an experimental approach aimed at improving the experience of biopsy”. If thus selected, they would receive further information and could decide at that time whether to accept the experimental option or receive usual care. The mindfulness trial randomization took place shortly after this first stage consent. Patients randomized to usual biopsy care had no further conversations about research. Those randomized to the mindfulness arm were told of their allocation when presenting for biopsy. They were asked whether they wanted the mindfulness intervention or to receive biopsy without the mindfulness intervention. Henceforth, patients on two-stage consent will be described in “control” or “experimental” arms if randomized to usual care vs. mindfulness respectively. These consents for the mindfulness trial generally took place in person, although some were done by e-consent during the height of the COVID pandemic.

With the exception of decisional regret, which was administered at 3 months, all patient- reported outcomes were administered within 48 hours of the initial consent for those in the one-stage arm and control arm of two-stage consent, and the second consent for patients in the experimental arm of two-stage consent.

The primary endpoint for patient distress was anxiety measured using the six-item version of the Spielberger State Anxiety Inventory (STAI). This instrument has been extensively validated^{8, 9} and has been very widely used in the assessment of anxiety related to medical procedures or information. Additional endpoints were consent-related anxiety, decision burden, decisional conflict and decisional regret. Consent-related anxiety used a 0 – 10 numerical rating scale (NRS) anchored at “Not at all anxious” to “Extremely anxious” in answer to: “You recently met with researchers at MSKCC to discuss taking part in a clinical trial about prostate biopsy. How anxious did that discussion make you feel?”. Decisional conflict was measured using the support, uncertainty and decision effectiveness domains of the Decisional Conflict Scale¹⁰ and decision regret using the Decision Regret Scale.¹¹ For decisional burden, we used an adaptation of two questions from Cognitive Load Scale, scoring a patient as “high burden” if the patient answered “difficult” or “very difficult” to the item “Understanding what I was told during the informed consent discussion was...” or answered “agree” or “strongly agree” to the item “The topics covered in the informed consent discussion were very complex.”¹²

To measure patient understanding of consent, we used the validated “Quality of Informed Consent” (QuIC) instrument.⁴ This includes two domains, section A assessing objective knowledge of the trial (such as whether the trial would necessarily have direct medical benefit) and section B assessing subjective understanding (such as “how well did you understand … why the researchers are doing the clinical trial?”). Both sections are scored on a scale of 0 – 100, with higher scores indicating a better knowledge or greater subjective understanding of the consent process. As previously described,³ the QuIC was adapted to the special needs of a study with two-stage consent. The exact items given to patients on each arm of the trial are described in Supplemental Table 3, along with a description of how the QuIC was adapted to this study.

We are conducting three studies comparing one vs. two-stage consent. The power calculation called for 62 patients per group in each of the three trials in order to detect a difference of 0.3 standard deviations in STAI anxiety.

The analysis was by intent-to-treat so all patients were included regardless of whether they consented to proceed to the trial. The difference in mean scores between the one-stage group and the combined two-stage group (both experimental and control) was calculated along with a 95% confidence interval using a two-sample t-test. For the decisional burden outcome, which was dichotomous, we reported the number and proportion in each group, along with a difference in proportions using Fisher’s exact test. All analyses were conducted using R version 4.2.0 with the tidyverse (v1.3.1) and gtsummary (v1.6.1) packages.^13–15 The trial was approved by the MSKCC as IRB number 18 189. It is registered on ClinicalTrials.gov as NCT03507725. There are several differences between the registered protocol and the methodology described above. Specifically, the ClinicalTrials.gov record concerns the therapeutic trial of mindfulness for procedural pain and distress during prostate biopsy and so the registered sample size reflect that required for the pain and distress endpoints. Moreover, the original protocol involved two-stage consent for all patients. The primary outcome is stated as accrual rate. When this was found to be extremely high, as described in our prior publication,³ we changed to a randomized trial of one vs. two stage consent as part of a trio of studies comparing these approaches.

Results

Of 162 patients approached for the consent trial, 125 agreed to take part, with 66 patients randomized to the one-stage consent group and 59 patients randomized to two-stage consent. There were 2 one-stage patients and 1 two-stage patient who declined to participate in the study of the mind-body intervention. Among the remaining 58 two-stage patients, 29 were in the mindfulness group and 29 were in the control group. One patient randomized to the two-stage control group did not take any surveys at the time of consent and was therefore not included in the analysis. A summary of patient characteristics by randomization arm is presented in Table 1.

Table 1.

Patient characteristics, separately by randomization arm. Data are presented as median (IQR) and N (%).

Characteristic	One-stage, N = 66	Two-stage, N = 58
Age at consent	66 (61, 71)	68 (64, 72)
Race
White	58 (88%)	51 (89%)
Black or African American	7 (11%)	5 (8.8%)
Asian	1 (1.5%)	1 (1.8%)
Unknown	0	1
Ethnicity
Non-Hispanic	60 (94%)	56 (98%)
Hispanic	4 (6.2%)	1 (1.8%)
Unknown	2	1
Education level
Graduated high school/GED	7 (11%)	3 (5.4%)
Completed some college	7 (11%)	2 (3.6%)
Associate degree	2 (3.2%)	1 (1.8%)
Bachelor’s degree	14 (22%)	20 (36%)
Graduate school or professional degree	33 (52%)	30 (54%)
Unknown	3	2
Prior biopsy status
No prior biopsy	5 (7.6%)	6 (10%)
Prior negative biopsy only	3 (4.5%)	2 (3.4%)
Prior positive and prior negative biopsies	19 (29%)	22 (38%)
Prior positive biopsy only	39 (59%)	28 (48%)
Allocation
Experimental group	32 (48%)	29 (50%)
Control Group	32 (48%)	28 (48%)
Declined to participate	2 (3.0%)	1 (1.7%)

Table 2 gives between-group differences for total QuIC scores, with responses to each QuIC question reported in Supplemental Tables 1 and 2. QuIC scores were slightly and non- significantly higher in the two-stage group, by 0.9 points (95% CI −2.3, 4.2, p=0.6) for section A (objective understanding) and 1.1 points (95% CI −4.8, 7.0, p=0.7) for section B (subjective understanding). The lower bound of the 95% CI excludes the possibility that two-stage consent would lead to importantly poorer understanding in the two-stage group. Table 2 also shows differences between randomized groups for the anxiety and decisional endpoints. As expected in this low-stakes setting, scores are generally low, with only small and non-significant differences between groups.

Table 2.

Total questionnaire scores, along with the difference between all one-stage and all two-stage patients with 95% confidence interval. Data are presented as mean (SD) or N (%). Quality of Informed Consent: 0 – 100 with higher scores showing greater understanding. Anxiety: 0 – 10 for consent-related and 20 – 80 for State-Trait Anxiety Inventory with higher scores indicated higher anxiety. Decisional Conflict and Decisional Regret: 0– 100 with higher scores indicating greater decisional conflict or regret.Decisional burden is dichotomized as high nor not.

Characteristic	One-stage consent N = 66	Two-stage consent (Control) N = 28	Two-stage consent (Experimental) N = 29	Two-stage consent (all) N = 584	Difference between one-stage consent and combined twostage consent	95% CI	p
Quality of Informed Consent *
Section A	83 (9)	88 (9)	81 (6)	84 (9)	0.9	−2.3, 4.2	0.6
Section B	87 (16)	86 (18)	91 (14)	88 (16)	1.1	−4.8, 7.0	0.7
Anxiety
Consent Related Anxiety	1.3 (2.4)	0.3 (0.7)1	1.3 (2.3)	0.8 (1.7)	0.5	−0.2, 1.3	0.2
State-Trait Anxiety Inventory	36 (12)	33 (10)	36 (12)	34 (11)	1.9	−2.1, 6.0	0.3
Decisional measures
Decisional conflict
Support Subscale	17 (24)	16 (25)	7 (11)	12 (20)	4.8	−3.1, 13	0.2
Uncertainty Subscale	16 (24)	24 (25)	10 (14)	17 (21)	−0.9	−8.9, 7.2	0.8
Effective Decision Subscale	16 (24)	22 (25)	6 (12)	14 (21)	1.6	−6.5, 9.6	0.7
Decisional Burden	6 (9.1%)	2 (7.1%)	2 (6.9%)	4 (6.9%)	2.2%	−8.9%, 13%	0.7
Decisional Regret	18 (16)3	13 (16)2	21 (20)1	18 (19)	−0.1	−6.7, 6.6	>0.9

^1:One missing observation.

^2:6 missing observations.

^3.7 missing observations.

^4:One patient from the two-stage arm who declined to participate in the trial was excluded from the “Two-stage consent (Experimental)” and “Two-stage consent (Control)” columns but included in the “Two-stage consent (All)” column.

^*As previously described³, the QuIC was adapted to the special needs of a study with two-stage consent.

For the QuIC A and anxiety-related endpoints, scores were similar between one-stage and two-stage experimental groups, but slightly superior in the two-stage control group; conversely, for decisional conflict, scores were similar between one-stage and two-stage control but somewhat better in the two-stage experimental group. Accordingly, we conducted an analysis that was not planned in the original study protocol, testing for a difference in scores between two-stage control and two-stage experimental arm patients using a two-sample t-test. The results are shown in Table 3. Scores on QuIC Section A (7.4; 95% CI 3.1, 12; p=0.001) and Consent Related Anxiety (difference of −1.0; 95% CI −1.9, −0.12, p=0.028) were significantly superior in the two-stage control group. QuIC section A and all three Decisional Conflict subscale scores were superior in the two-stage experimental group, a difference of 7.4 (95% CI 3.1, 12, p=0.001) for QuIC section A and 9.5 (95% CI −1.1, 20, p=0.078), 14 (95% CI 2.7, 25, p=0.016), and 16 (95% CI 5.1, 26, p=0.005) for decisional support, uncertainty and effective decision subscales, respectively. Responses for each question on QuIC Section A and Decisional Conflict Scores separately for one-stage, two-stage control group and two-stage experimental group patients, are given in Supplemental Tables 3 and 4.

Table 3.

QuIC Section A, Decisional Conflict Scale, and Consent Related Anxiety score for two-stage control group and two-stage experimental group patients, with mean difference and 95% confidence interval. Data are presented as mean (SD). One two-stage patient who declined to participate in the trial was excluded from this table. Quality of Informed Consent: 0 – 100 with higher scores showing greater understanding. Consent-related anxiety: 0 – 10 with higher scores indicated higher anxiety. Decisional Conflict:0–100, with higher scores indicating greater decisional conflict.

Characteristic	Two-stage Control Group, N = 28	Two-stage Experimental Group, N = 29	Difference	95% CI	p-value
Quality of Informed Consent Section A*	88 (9)	81 (6)	−7.4	−12, −3.1,	0.001
Decisional Conflict Scale
Support Subscale	16 (25)	7 (11)	9.5	−1.1, 20	0.078
Uncertainty Subscale	24 (25)	10 (14)	14	2.7, 25	0.016
Effective Decision Subscale	22 (25)	6 (12)	16	5.1, 26	0.005
Consent Related Anxiety	0.3 (0.7)	1.3 (2.2)	−1.0	−1.9, −0.1	0.028

^*As previously described,³ the QuIC was adapted to the special needs of a study with two-stage consent.

The therapeutic comparison of mindfulness meditation versus control continued after closure of the comparison between one and two-stage consent. The investigator preferred two-stage consent on the grounds that the simpler presentation was less confusing to patients, and continued the trial using that approach.

Discussion

The rationale for two-stage consent involves two hypotheses, maintaining patient understanding and lowering consent-related anxiety and distress. We clearly found evidence supporting the first hypothesis: understanding of consent as measured by a standardized measure was very slightly higher in the two-stage consent group with the lower bound of the 95% CI excluding an importantly poorer score. With respect to the second hypothesis, we did not expect to see important differences in this low-stakes trial. As expected, overall and consent-related anxiety were very low, inducing a “floor effect”. Even then, we did see some evidence of a signal, with lower scores in patients at the first stage of two-stage consent.

For consent-related anxiety, and the understanding domain of the QuIC, there were no statistically significant differences between one and two-stage consent. Yet patients in the first stage versus second stage (mindfulness arm) two-stage consent had significantly higher scores for understanding and lower scores for consent-related anxiety (p=0.021 / 0.033 vs. one-stage consent and 0.007 / 0.046 vs. second stage of two-stage consent, respectively). With respect to understanding, it is true that the QuIC instruments given to each group varied, as this is inevitable given that patients were given different information, and the relevant QuIC does test what patients were told: patients in the control arm in a two-stage consent design have less information to digest, and fewer potentially ambiguous questions to answer, so it is not surprising that they understand it better. With respect to both STAI and consent-related anxiety, one important difference between the groups is that patients in the experimental arm of two-stage consent were assessed shortly after their actual biopsy procedure. The protocol called for all patients to be assessed within 48 hours of their consent, as this would lessen the problem of recall. The second stage of two-stage consent took place shortly before biopsy and it is likely that patients’ general anxiety about the procedure – and that it may find aggressive prostate cancer – influenced their responses to questions about consent-related anxiety and distress. It might well be seen as a “win” for two-stage consent that patients completing consent at a particularly stressful time – just after a cancer biopsy – did not report on average higher anxiety than patients completing one-stage consent at a routine appointment. Decisional conflict was lower at the second stage. We hypothesize that this is because it was clearer what patients are being asked to do at second stage consent.

There has been some debate in the literature about the ethics of two-stage consent. Although this has sometimes been based on misapprehensions of the methodology,¹⁶ some substantive concerns have been raised. Weijer et al¹⁷ claim patients in staged-consent trials are unable to provide “autonomous authorization” because, in order to do so, “they must adopt the ends of the study as their own” and this is not possible if the description of the study is “so vague as to encompass any study involving the disease in question.” In the current study, the study aims were described in terms of “making small changes to the [biopsy] approach” to find ways to improve the patient’s experience. This hardly “encompasses any study involving [prostate cancer]”. We also disagree with Weijer et al.’s argument that patients may only participate in a randomized trial if they are doing so because of their interest in finding out whether experimental treatment is superior to control. This conflates understanding of trial information with motives for participating.

In conclusion, we have found no evidence that, in the setting of a low-stakes trial, two-stage consent has any adverse impact on the consent process compared to the one-stage approach. At least for low stakes trials, trialists who prefer the two-stage approach – as pointed out, the PI of the prostate biopsy trial immediately opted for two-stage after completion of randomized comparison of consent approaches – can be confident that doing so maintains patient understanding of research and hence fulfills the ethical goal of patient autonomy. We also saw some signal of decreased anxiety associated with the two-stage consent, an effect attributable to the experience of patients in the two-stage control group. Further research is warranted on two-stage consent in higher stakes settings.