Impact of Restricting Enrollment in Stroke Genetics Research to Adults Able to Provide Informed Consent

Abstract

Background and Purpose

The extent of potential consent bias in observational studies elucidating genetic and environmental contributions to ischemic stroke is largely unknown. The purpose of this study was to assess differences in stroke cohort characteristics between those who provided informed consent and those whose enrollment was authorized by surrogate decision makers.

Methods

The Ischemic Stroke Genetics Study (ISGS) is a prospective, 5-center, case-control study of first-ever ischemic stroke. Demographic, clinical, and stroke characteristics were compared between cases enrolled by self versus by surrogate. Data from 1 site that limits enrollment only to those able to self-consent were also analyzed to compare those who enrolled with those not able to consent.

Results

Overall, 10% (54/517) were enrolled using surrogate authorization. Self-consented and surrogate-authorized cases did not differ significantly in age, sex, or conventional risk factors. Surrogate-authorized cases had significantly more severe stroke deficits, larger infarcts, and infarcts localizing to left supratentorial regions. Similarly, at the site restricting enrollment, stroke severity and characteristics differed between self-consented individuals and those otherwise eligible but unable to provide consent.

Conclusions

Failure to permit surrogate authorization in genetic studies of ischemic stroke may skew enrollment toward less severe strokes caused by smaller infarcts. This potential consent bias may undermine the ability to identify genetic determinants of risk and severity and suggests that surrogate enrollment in these studies can be ethically justifiable.

Introduction

Research elucidating genetic and environmental risk factors associated with complex diseases such as ischemic stroke has the potential to revolutionize prevention and treatment strategies. The ability to identify high-risk patients through genetic testing may make screening for treatable intermediate phenotypes more cost-effective. A clear and comprehensive understanding of genetic risk may promote advances in gene therapy and in development of novel pharmaceutical agents (1).

Like much of neuroscience research aimed at understanding etiology and pathophysiology rather than testing experimental treatments, research concerning genetic risk factors for stroke has limited direct medical benefit for the participant and poses unique ethical challenges (2-4). Stroke can impair patients’ capacity to make decisions, rendering them incapable of providing informed consent. Enrolling incompetent individuals in research challenges a fundamental norm of research ethics: that research subjects must give informed consent. Nevertheless, evidence-based improvements in treatment and prevention of an array of medical conditions, including stroke, depend on research involving individuals with, or at risk for, impaired decision-making capacity (5).

Federal regulations governing protection of human subjects in research are silent on the issue of enrolling adults who lack capacity to give informed consent outside clinical emergency research (6,7). Commonly used guides for institutional review boards (IRBs) generally provide little direction on whether, and how, adults who lack capacity should be enrolled in research, particularly if the research does not offer the prospect of direct medical benefit to the participant, but admonish IRBs to adhere to local laws (8,9). State and local laws are mostly silent on these issues as well, and institutional policies differ regarding permissible research and acceptable protections for research with decisionally impaired adults (10). National and state advisory committees constituted to address this issue have been unable to reach consensus on whether research may be ethically conducted on persons unable to give consent and unlikely to benefit from participation in the research, particularly if the research poses more than minimal risk (11-13).

The Ischemic Stroke Genetics Study (ISGS) is a multicenter, case-control study that compares allele frequencies of candidate gene polymorphisms between cases with first-ever ischemic stroke and stroke-free control subjects (1). ISGS is designed to enhance external validity and has broad eligibility criteria compared with clinical trials of drugs or devices. To avoid biasing the study toward discovery of risk factors for mild to moderate stroke but not severe stroke, individuals rendered incompetent by stroke are enrolled by surrogate decision makers to the extent permitted by local IRBs.

Because stroke genetics research does not offer the prospect of direct medical benefit, some institutions do not allow surrogate enrollment. Four of the 5 ISGS sites allow surrogate enrollment; 1 does not. We sought to assess the impact that restricting enrollment to individuals who provided their own informed consent would have had on the study population.

Methods

The original aims and methods of ISGS have been published (1). All participants in ISGS have provided written informed consent or the consent form has been signed by a surrogate authorizing enrollment into the study. As part of a planned secondary analysis, we compared those who provided their own consent for participation with those enrolled via surrogate authorization. We compared baseline and outcome data collected as part of the primary study. The variables investigated include age, sex, National Institutes of Health Stroke Scale (NIHSS) score, infarct size and location, Oxford Handicap Scale (OHS) score (both before and after stroke), and 90-day follow-up data on mortality, morbidity, and measures of outcome (Glasgow Outcome Scale [GOS]) and functional status (Barthel Index [BI]). Stroke symptoms were assessed using the Questionnaire for Verifying Stroke-Free Status (QVSS) (14). Participants were scored as having visual deficits if they had positive responses to either QVSS item 5 or 6. Participants were scored as having symptoms of communication deficits if they had positive responses to either QVSS item 7 or 8. Neurologic signs were scored using NIHSS as previously described (15). Infarct size and location were determined by the magnetic resonance imaging scan, if available, or by the computed tomographic scan otherwise. We also compared stroke subtype between groups. As a stroke genetics association study, the ISGS protocol specifies prospective phenotyping and determination of stroke subtype using standardized criteria applied by a single physician adjudicator.

Self-consented and surrogate-authorized cases were compared using χ² or Fisher exact tests for categorical variables and Wilcoxon rank sum tests for ordinal and continuous measures. The ISGS protocol is approved by the IRBs at all participating centers.

At the site not allowing surrogate enrollment, nonidentifiable data items are recorded in screening logs on potentially eligible individuals who are not enrolled along with the reason for nonenrollment, including decisional incapacity. This allows enumeration and basic characterization of those otherwise eligible individuals not approached because of their decisional incapacity. In making comparisons between enrolled participants and those not able to be approached because of their lack of capacity, we also used χ² or Fisher exact test for categorical variables and the Wilcoxon rank sum test for ordinal and continuous measures. These analyses were approved by the IRB at this site.

Results

Between December 2002 and December 2005, a total of 517 cases with first-ever ischemic stroke were prospectively accrued to this study. Four hundred sixty-three (90%) provided self-consent, and 54 (10%) were enrolled with surrogate authorization.

Clinical Characteristics

Overall, 231 subjects (45%) were women, 150 (29%) were black, and 351 (68%) were white. The median time to enrollment was 4 days in both groups. Demographic and lifestyle characteristics for the 517 cases, stratified by consent status, are summarized in Table 1. These characteristics were similar between the groups except for race. Among the 150 black subjects, 129 (86%) self-consented and 21 (14%) were enrolled by surrogate; among the 351 white subjects, 322 (92%) self-consented and 29 (8%) were enrolled by surrogate. Medical history did not differ significantly between groups, aside from migraine headaches (Table 2).

Table 1.

Baseline Characteristics by Consent Status (N=517)^*

Characteristics	Self-consented	Surrogate-authorized	P value†
No. of patients	463	54
Age, y			.32‡
Mean (SD)	63.7 (15.2)	66.3 (12.5)
Median (IQR)	64.9 (51.9-75.7)	66.8 (56.5-75.1)
Sex			>.99
Male	256/463 (55)	30/54 (56)
Female	207/463 (45)	24/54 (44)
Race			.02
Black	129/463 (28)	21/54 (39)
White	322/463 (70)	29/54 (54)
Other	12/463 (3)	4/54 (7)
Weight by body mass index			.39
Underweight	9/461 (2)	1/54 (2)
Normal	119/461 (26)	17/54 (31)
Overweight	154/461 (33)	21/54 (39)
Obese	179/461 (39)	15/54 (28)
Smoking			.33
Never	144/463 (31)	18/54 (33)
Prolonged abstainer	171/463 (37)	14/54 (26)
Recent abstainer	29/463 (6)	3/54 (6)
Active smoker	119/463 (26)	19/54 (35)
Alcohol			>.99
Never/rarely	294/463 (63)	34/54 (63)
≥1/wk	169/463 (37)	20/54 (37)
Vigorous activity			.19
<1x/wk	235/463 (51)	34/54 (63)
1-3x/wk	135/463 (29)	10/54 (19)
≥4x/wk	93/463 (20)	10/54 (19)
Prestroke OHS score			.30
0	355/463 (77)	41/54 (76)
1	53/463 (11)	7/54 (13)
2	46/463 (10)	3/54 (6)
3	5/463 (1)	2/54 (4)
4	3/463 (1)	1/54 (2)
5	1/463 (<1)	0/54 (0)

IQR, interquartile range; OHS, Oxford Handicap Scale.

^*Values are number (percentage) unless specified otherwise. Data are incomplete for some characteristics. Total percentages may not equal 100% because of rounding.

^†Fisher exact tests unless specified otherwise.

^‡Wilcoxon rank sum test.

Table 2.

Medical History by Consent Status (N=517)^*

Characteristic	Self-consented	Surrogate-authorized	P value†
No. of cases	463	54
Transient ischemic attack	69/462 (15)	10/54 (19)	.55
Congestive heart failure	47/461 (10)	8/54 (15)	.35
Angina	49/462 (11)	4/54 (7)	.64
Myocardial infarction	69/462 (15)	12/54 (22)	.17
Atrial fibrillation
Chronic	27/460 (6)	5/53 (9)	.36
Intermittent	40/459 (9)	1/53 (2)	.11
Any	67/460 (15)	6/53 (11)	.68
Hypertension	327/461 (71)	40/54 (74)	.75
Deep vein thrombosis	30/462 (6)	2/54 (4)	.56
Pulmonary embolus	8/462 (2)	1/54 (2)	>.99
Peripheral vascular disease	20/461 (4)	3/54 (6)	.72
Migraine headaches
With aura	35/462 (8)	3/54 (6)	.79
Without aura	88/462 (19)	3/54 (6)	.01
Any	123/462 (27)	6/54 (11)	.01
Hyperlipidemia	211/461 (46)	24/53 (45)	>.99
Diabetes mellitus
On insulin	45/463 (10)	8/54 (15)	.24
Not on insulin	78/463 (17)	11/54 (20)	.57
Any	123/463 (27)	19/54 (35)	.20

^*Values are number (percentage) unless specified otherwise. Data are incomplete for some characteristics. Total percentages may not equal 100% because of rounding.

^†Fisher exact tests.

Stroke Symptoms and Severity

Stroke characteristics are summarized in Table 3. Reported QVSS symptoms of weakness, numbness, and visual deficits did not differ significantly between self-consented and surrogate-authorized cases. Reported communication problems occurred more frequently among the surrogate-authorized individuals (57% vs 39% of self-consented cases; P=.02). Signs of neurologic impairment detected by NIHSS occurred much more frequently among the surrogate-authorized cases: weakness (98% vs 60%; P<.001); numbness (52% vs 28%; P=.001); visual deficits (43% vs 17%; P<.001); and communication problems (78% vs 35%; P<.001).

Table 3.

Stroke Symptoms and Severity by Consent Status^*

Characteristics	Self-consented	Surrogate-authorized	P value†
No. of cases	463	54
Symptoms/signs
Weakness
QVSS	310/461 (67)	31/49 (63)	.63
NIHSS	278/463 (60)	53/54 (98)	<.001
Numbness
QVSS	261/460 (57)	22/49 (45)	.13
NIHSS	130/462 (28)	28/54 (52)	.001
Visual deficits
QVSS	89/461 (19)	8/49 (16)	.70
NIHSS	78/463 (17)	23/54 (43)	<.001
Communication
QVSS	182/462 (39)	28/49 (57)	.02
NIHSS	163/463 (35)	42/54 (78)	<.001
No. of symptoms
QVSS			.94‡
No. of cases	462	49
Median (SD)	1.8 (1.1)	1.8 (1.3)
Median (IQR)	2.0 (1-3)	2.0 (1-3)
NIHSS			<.001‡
No. of cases	463	54
Mean (SD)	1.4 (1.0)	2.7 (0.9)
Median (IQR)	1.0 (1-2)	3.0 (2-3)
Stroke severity score
NIHSS			<.001‡
No. of cases	463	54
Mean (SD)	3.2 (3.7)	11.1 (7.2)
Median (IQR)	2.0 (1-4)	8.5 (5-16)
BI			<.001‡
No. of cases	463	54
Mean (SD)	81.6 (24.7)	47.2 (34.3)
Median (IQR)	95.0 (70-100)	42.5 (15-75)
OHS			<.001‡
No. of cases	463	54
Mean (SD)	1.8 (1.2)	3.0 (1.4)
Median (IQR)	2.0 (1-3)	3.0 (2-4)
GOS			<.001‡
No. of cases	463	54
Mean (SD)	1.6 (0.7)	2.4 (0.8)
Median (IQR)	1.0 (1-2)	3.0 (2-3)
Infarct diameter, cm			<.001
Not seen	40/463 (9)	2/54 (4)
<1.5	150/463 (32)	8/54 (15)
1.5-3.0	129/463 (28)	11/54 (20)
>3.0	144/463 (31)	33/54 (61)
Infarct location			<.001
Not seen	40/463 (9)	2/54 (4)
Supratentorial left	163/463 (35)	34/54 (63)
Supratentorial right	180/463 (39)	13/54 (24)
Infratentorial	77/463 (17)	3/54 (6)
Multiple locations	3/463 (1)	2/54 (4)

BI, Barthel index; GOS, Glasgow Outcome Scale; IQR, interquartile range; OHS, Oxford Handicap Scale; NIHSS, National Institutes of Health Stroke Scale; QVSS, Questionnaire for Verifying Stroke-Free Status.

^*Values are number (percentage) unless specified otherwise. Data are incomplete for some characteristics. Total percentages may not equal 100% because of rounding.

^†Fisher exact tests unless specified otherwise.

^‡Wilcoxon rank sum test.

Stroke severity was quantified by 4 measures—NIHSS, BI, OHS, and GOS (Table 3). Scores for each of these measures were significantly worse for the surrogate-authorized than for the self-consented cases (all P<.001).

Infarct size and location also differed between the groups (Table 3). The proportion with larger infarcts (>3.0 cm) was 31% for the self-consented cases and 61% for the surrogate-authorized cases (P<.001). In both groups, the majority of strokes were supratentorial. Surrogate-authorized cases had significantly higher rates of left supratentorial infarction (63% vs 35%; P<.001).

Subtyping classifications are summarized in Table 4. Trial of Org 10172 in Acute Stroke Treatment (TOAST) subtyping did not differ significantly between self-consented and surrogate-authorized cases (P=.75). Self-consented and surrogate-authorized cases did, however, differ significantly with respect to the OHS classification (P<.001). Notably, only 7% of the self-consented cases had total anterior circulation infarcts, whereas 33% of surrogate-authorized cases had total anterior circulation infarcts.

Table 4.

Stroke Subtype Classifications by Consent Status

Classifications	Self-consented	Surrogate-authorized	P value*
No. of cases	463	54
TOAST			.75
Large-artery atherosclerosis	83/463 (18)	13/54 (24)
Small-vessel occlusion	85/463 (18)	11/54 (20)
Cardioembolism	113/463 (24)	13/54 (24)
Other etiology	18/463 (4)	1/54 (2)
Undetermined etiology	164/463 (35)	16/54 (30)
OHS			<.001
Lacunar infarct	133/463 (29)	14/54 (26)
Total anterior circulation infarct	34/463 (7)	18/54 (33)
Partial anterior circulation infarct	205/463 (44)	19/54 (35)
Posterior circulation infarct	91/463 (20)	3/54 (6)

OHS, Oxford Handicap Scale; TOAST, Trial of Org 10172 in Acute Stroke Treatment.

^*Fisher exact test.

90-Day Outcomes

Stroke cases were contacted approximately 90 days after the initial stroke, and BI, OHS, and GOS scores were used to quantify the individuals’ functional status (Table 5). At 90 days, surrogate-authorized cases continued to do significantly worse than those who self-consented. The change in these measures from baseline to 90 days did not differ significantly between the 2 groups, with the surrogate-authorized group consistently performing worse on all measures.

Table 5.

90-Day Stroke Outcomes Scores by Consent Status

	Self-consented		Surrogate-authorized
Classification	No. of cases	Mean (SD) Median (IQR)	No. of cases	Mean (SD) Median (IQR)	P value*
No. of cases	463		54
BI	452	91.0 (20.9)	53	62.6 (38.8)	<.001
		100 (95-100)		80 (30-100)
OHS	454	1.4 (1.3)	54	2.8 (1.6)	<.001
		1.0 (0-2)		3.0 (2-4)
GOS	449	1.5 (0.8)	53	2.4 (1.3)	<.001
		1.0 (1-2)		2.0 (1-3)

BI, Barthel index; GOS, Glasgow Outcome Scale; OHS, Oxford Handicap Scale.

^*Wilcoxon rank sum tests.

Analysis From Site Restricting Enrollment

During the same study period, a total of 105 cases were enrolled from the 1 site that did not allow surrogate authorization (Table 6). An additional 56 individuals were approached but declined participation. A total of 85 otherwise eligible individuals were not approached regarding enrollment because they lacked capacity. Those lacking capacity were older (P=.002). Lack of capacity was most commonly caused by aphasia or communication difficulties (39/85 [46%]). The NIHSS scores were significantly worse for those lacking capacity than for those enrolled (P<.001). Distribution of stroke subtypes as determined by the TOAST classification showed a greater proportion of cases with cardioembolic stroke and a smaller proportion having a small-vessel mechanism (P=.01). The subtype classification using the OHS method also differed, with a greater proportion of those enrolled having lacunar infarcts or partial anterior circulation infarcts, whereas those lacking capacity were more likely to have total anterior circulation infarcts (P<.001).

Table 6.

Comparison of Enrolled Individuals and Those Otherwise Eligible Lacking Capacity to Consent at Site 4^*

Classification	Enrolled cases	Cases unable to consent	P value†
No. of cases	105	85
Age, y			.002‡
Mean (SD)	62.4 (14.7)	69.1 (14.7)
Median (IQR)	63.4 (51.9-73.6)	72.0 (60.0-81.0)	<.001§
Sex			.20
Male	58 (55)	39 (46)
Female	47 (45)	46 (54)
Race∥			.52
Black	20 (19)	22 (26)
White	82 (78)	63 (74)
Other	3 (3)	0 (0)
Stroke severity by NIHSS
Mean (SD)	3.4 (3.8)	14.1 (7.4)	<.001‡
Median (IQR)	2 (1-4)	15 (8-21)	<.001§
TOAST			.01
Large-artery atherosclerosis	30 (29)	14 (16)
Small-vessel occlusion	13 (12)	4 (5)
Cardioembolism	19 (18)	32 (38)
Other etiology	7 (7)	6 (7)
Undetermined etiology	36 (34)	29 (34)
OHS			<.001
Lacunar infarct	22 (21)	6 (7)
Total anterior circulation infarct	6 (6)	39 (46)
Partial anterior circulation infarct	52 (50)	27 (32)
Posterior circulation infarct	25 (24)	13 (15)

IQR, interquartile range; NIHSS, National Institutes of Health Stroke Scale; OHS, Oxford Handicap Scale; TOAST, Trial of Org 10172 in Acute Stroke Treatment.

^*Values are number (percentage) unless indicated otherwise.

^†χ² test unless otherwise specified.

^‡Fisher exact tests.

^§Wilcoxon rank sum test.

^∥Comparing white versus nonwhite.

Discussion

Our study demonstrates and quantifies how stroke severity and subtype differ between those with and without capacity to provide informed consent for genetic research. Our study found few differences in baseline characteristics between self-consented and surrogate-authorized individuals. The difference in race is of uncertain importance and may be attributable to a number of factors, including racial differences in case severity. An earlier epidemiologic stroke study also demonstrated a significant effect of excluding cases unable to respond to interviews due to dementia, aphasia, or impaired consciousness (16). Together, these findings have implications for the design and conduct of stroke genetics research, particularly with regard to the ethical acceptability of enrolling individuals unable to provide informed consent for research participation.

Ethical clinical research requires balancing several ethical requirements, including the requirement for scientific validity and the requirement to respect individuals by treating them as autonomous agents through the process of informed consent (17,18). To justify enrolling any participant, a study must have the potential to advance knowledge through a scientifically valid study design. To protect against exploitation, a study proposing to enroll adults who lack the capacity to provide informed consent for research must at least satisfy the “necessity requirement”; that is, enrollment of such individuals is scientifically necessary (19). If a study can obtain scientifically valid results with participants who can provide informed consent, then persons who lack this capacity should not be enrolled even for research directed at conditions likely to produce cognitive impairment.

Some studies, however, may be able to obtain valid results only by including individuals who lack the capacity to provide informed consent. Allowing their enrollment with appropriate safeguards may be important for the study to provide generalizable knowledge. Such safeguards are delineated in the federal guidelines governing research with children, considered persons with diminished autonomy (6). Enrollment by surrogate decision makers and limiting permissible research to acceptable risk levels provide important safeguards and are built into the federal guidelines for children. However, there are currently no comparable federally regulated safeguards for adults who are unable to provide their own research informed consent.

We found that approximately one-third of otherwise eligible patients may have impaired capacity. However, the proportion of individuals enrolled into ISGS by surrogate decision makers is lower (13% at the 4 sites allowing surrogate enrollment). Our study was not designed to determine the reason for this. It may be that surrogate decision makers are reluctant to enroll individuals into this study, or it may be that investigators are reluctant to approach families whose family member has had a severe stroke. Further research should investigate potential reasons. Statistical and analytic innovations to address any residual bias due to incomplete ascertainment of the phenotypic spectrum must also be concurrently developed.

Conclusions

Our study supports the “necessity requirement” for ethical enrollment of decisionally impaired individuals in ischemic stroke genetics research. Enrollment by surrogate decision makers and limiting permissible research to acceptable risk levels provide important safeguards for these individuals.

Our findings further suggest that excluding individuals unable to provide informed consent from stroke genetics studies may diminish investigators’ ability to detect genetic associations with stroke severity and perhaps stroke subtypes more likely to lead to decisional impairment and may limit external validity. It is well established that in epidemiological research there is the potential for survival bias in general and in stroke research specifically (20). In many ways, survival bias can be considered as an extreme version of consent bias. Our findings raise the possibility of systematic bias in genetic research in any disease in which disease severity can negatively affect capacity to provide informed consent for research and suggest that there is an ethical necessity to consider allowing surrogate enrollment in these kinds of studies.

Abbreviations

BI

Barthel index

GOS

Glasgow Outcome Scale

IRB

institutional review board

ISGS

Ischemic Stroke Genetics Study

NIHSS

National Institutes of Health Stroke Scale

OHS

Oxford Handicap Scale

QVSS

Questionnaire for Verifying Stroke-Free Status

TOAST

Trial of Org 10172 in Acute Stroke Treatment