Professional Documents
Culture Documents
Woitowich 2018
Woitowich 2018
Abstract
Objective: On January 25th, 2016, the National Institutes of Health (NIH) implemented a policy requiring
investigators to consider sex as a biological variable (SABV) within their grant submissions. We surveyed NIH
Downloaded by 193.56.67.114 from www.liebertpub.com at 02/06/19. For personal use only.
study section members in 2016 and 2017 to determine their attitudes toward the policy and their perceptions
regarding its implementation.
Materials and Methods: Members of standing study sections and special emphasis panels who met in May,
June, or July of 2016 and 2017, and had a publicly accessible e-mail address, were invited to participate in the
survey (n = 4376 and n = 4710, respectively). The survey assessed participant demographics, knowledge and
awareness of the SABV policy, and opinions regarding its utility and implementation.
Results: A combined total of 1161 study section members participated in the survey for a response rate of
10.2% in 2016 and 15.1% in 2017. Respondents thought it was important for NIH-funded research to consider
SABV (63% vs. 68%, p = 0.141) and that it will improve rigor and reproducibility (54% vs. 58%, p = 0.208). In
terms of implementation, respondents indicated that the percentage of grants, which have successfully ad-
dressed and incorporated the policy, has significantly increased over time ( p < 0.0001 for all endpoints).
However, open-ended comments revealed concern for federal research funding, the overuse of experimental
animals, and uncertainty surrounding grant scoring, as it relates to the SABV policy.
Conclusions: In this study, we show improving attitudes toward the sex-inclusive policy at NIH and that a
statistically significant number of grants are addressing sex as a biological variable appropriately in their
submissions. These data suggest the policy is becoming more well accepted, and it is thus anticipated that the
reproducibility of scientific reports will increase over time and new discoveries using sex as a biological
variable are on the horizon.
1
Women’s Health Research Institute, 2Department of Obstetrics and Gynecology, Feinberg School of Medicine, Northwestern University,
Chicago, Illinois.
9
10 WOITOWICH AND WOODRUFF
are themselves grantees, thus providing a proxy for investi- scoring. In addition, participants had the ability to include
gators across all disciplines of science. A survey was ad- open-ended comments at the completion of the survey.
ministered after the initial round of grant reviews in which
applicants were required to consider SABV and 1 year later,
thus providing the framework to assess change in compre- Study participants and recruitment
hension and compliance. In aggregate, the data indicate an Following procedural review of the online consent form and
increasing approval of the policy along and improved com- survey questions to be asked of intended participants, the
pliance by investigators. Northwestern University institutional review board deemed
this study exempt from further review. A comprehensive list of
Materials and Methods standing study section and special emphasis panel members
was compiled from meeting rosters published on the Center for
Survey
Scientific Review (CSR) website.13 Meeting rosters corre-
Questionnaires were administered through an online survey sponding to the initial review of proposals required to consider
platform (Qualtrics) in 2016 and 2017. Survey questions were sex as a biological variable (May–July 2016) and those 1 year
designed to determine study section members’ opinions and later (May–July 2017) were included. Mail-in reviewers were
attitudes toward the implementation and evaluation of the NIH excluded from this study.
policy, requiring the consideration of SABV in preclinical The 2016 and 2017 review cohorts contained 4686 and
research (Supplementary Data; Supplementary Data are 4841 individuals respectively. E-mail addresses of 4376
available online at www.liebertpub.com/jwh). Twenty-four (93.3%) individuals from the 2016 cohort and 4710 (97.3%)
Downloaded by 193.56.67.114 from www.liebertpub.com at 02/06/19. For personal use only.
questions were included. Major themes of the survey included individuals from the 2017 cohort were obtained through a
participant demographics, experience as a study section comprehensive web-based search utilizing Google search
member, previous experience conducting sex-based research, engines and institutional directories.
opinions on the overall policy as it benefits the research en- An e-mail outlining the purpose of the study, along with a
terprise, and evaluation of the policy as it relates to grant link to participate in the survey was sent to the 2016 review
Data analyses
Data analyses were primarily qualitative, with a small quan-
titative component. Descriptive statistics were used where
appropriate. Nominal data were described as n (%). Chi-
squared tests were used to compare categorical data as well as
to assess associations between participant sex and experience
conducting sex-based research with attitudes toward the
SABV policy (GraphPad Prism, version 7.0). p-values <0.05
were considered significant.
Open-ended survey comments were analyzed using a
thematic and semiquantitative approach. Comments were
coded using deductive methods by the authors. Both authors
Downloaded by 193.56.67.114 from www.liebertpub.com at 02/06/19. For personal use only.
Data sharing plan dents felt confident in their understanding of the SABV
To ensure respondent anonymity, data were not analyzed policy (83.5%), which increased significantly to 88.4% by
by study section or special emphasis panel. Deidentified data, 2017 ( p = 0.018, Fig. 1 and Supplementary Table S3).
which exclude this information, can be provided by request In both years, over 70% of respondents indicated that the
from the corresponding author. NIH provided adequate educational resources regarding the
SABV policy (70.8% vs. 74.5%; p = 0.165), whereas the ma-
jority reported that they were unaware of or were not provided
Results
any educational resources by their home institutions (75.2% vs.
A total of 1161 study section members, representing over 75.6%; p = 0.886, Supplementary Table S2). In both years,
175 standing study sections and continuing special emphasis more than 90% of the respondents indicated that they discussed
panels, participated in the 2016 and 2017 surveys (Table 1). the policy at their respective study section meetings (92.2% vs.
Briefly, the majority of respondents were white (80% vs. 92.3%; p = 0.951, Supplementary Table S2).
81%) males (51% vs. 56%) employed in academia (96% vs. To assess study section members’ attitudes toward the
97%), who served as permanent members of study section SABV policy, we asked if it was important to consider SABV
(58% vs. 56%) with 3 years or more experience on study in the experimental design of NIH-funded research, and if it
section (76% vs. 75%). would improve rigor and reproducibility (Fig. 1). These two
Next, we queried respondents on their awareness, under- questions highlight a direct application and goal of the policy,
standing, and attitudes toward the SABV policy (Fig. 1 and respectively. A majority of respondents thought that it was
Supplementary Tables S1–S3). The majority of respondents important to consider SABV in the experimental design
became aware of the SABV policy through the NIH (60% vs. (63.4% vs. 67.6%; p = 0.141) and that considering SABV
57%) and not through their home institution (13% vs. 12%, would improve rigor and reproducibility (54.0% vs. 5%;
Supplementary Table S1). In 2016, the majority of respon- p = 0.208, Supplementary Table S3). In both years,
12 WOITOWICH AND WOODRUFF
Table 2. Comparison of Respondent Attitudes Toward the Sex as a Biological Variable Policy
by Sex-Based Research Experience Between 2016 and 2017
2016 2017
Research Research
Research does not Research does not
involves involve involves involve
SSD SSD SSD SSD
(n = 156) (%) (n = 292) (%) p-value (n = 257) (%) (n = 456) (%) p-value
Importance of 0.0037 0.0001
considering SABV in
experimental design
Yes 113 72.4 171 58.6 197 76.7 285 62.5
No or unsure 43 27.6 121 41.4 60 23.3 171 37.5
Consideration of SABV 0.0002 0.0012
to improve rigor
and reproducibility
Yes 103 66.0 139 47.6 169 65.8 243 53.3
No or unsure 53 34.0 153 52.4 88 34.2 213 46.7
Downloaded by 193.56.67.114 from www.liebertpub.com at 02/06/19. For personal use only.
respondents whose research portfolios involve the study of tically significant increase in the proportion of grants that
sex differences were significantly more likely to think that the addressed the SABV policy between 2016 and 2017
consideration of SABV is important (73% vs. 59% in 2016; ( p < 0.0001 for each endpoint).
p = 0.0037, and 77% vs. 63% in 2017; p = 0.0001) and would A majority of study section members indicated that the
improve rigor and reproducibility compared to their peers consideration of SABV was consistently weighed into an
(66% vs. 48% in 2016; p = 0.0002, and 66% vs. 53% in 2017; applicant’s approach score (55% vs. 61%; p = 0.0518). In
p = 0.0012, Table 2). addition, there was a statistically significant increase in the
Self-reported biological sex of the respondent did not in- proportion of grants that provided consistent justification for
fluence these views in 2016, yet in 2017, females were sig- the use of a single sex between 2016 and 2017 (44% vs. 50%;
nificantly more likely to think that the consideration of SABV p = 0.0330).
is important and would improve rigor and reproducibility Respondents had the option to leave open-ended com-
compared to males, respectively (73% vs. 63%; p = 0.005, ments at the end of each survey (n = 130 in 2016 and n = 257
and 64% vs. 53%; p = 0.002, Table 3). in 2017, Supplementary Table S5). There were no differences
The NIH SABV policy only applies to grants that include in demographics between respondents who left comments in
the use of human or vertebrate subjects.7 As such, respon- 2016 and 2017, with the exception that there were signifi-
dents who reviewed grants meeting these criteria (99% vs. cantly more study section chairpersons who commented in
97%; p = 0.130, Table 1) were asked what proportion of 2017 compared to 2016 (8.6% vs. 2.3%, respectively;
grants they reviewed accounted for the consideration of p = 0.018). Following initial review, comments were cate-
SABV in the research strategy and adequately addressed the gorized into one or more of the following major themes:
incorporation of SABV into the experimental design, analy- budget, animal use, policy, research area, and grant review.
sis, and reporting, as instructed by NIH guidelines (Fig. 2 and In the budget theme, comments highlighted the need for
Supplementary Table S4). In both cases, there was a statis- increased federal funding for research citing that the SABV
Table 3. Comparison of Respondent Attitudes Toward the Sex as a Biological Variable Policy
by Sex Between 2016 and 2017
2016 2017
Male Female Male Female
(n = 228) (%) (n = 220) (%) p-value (n = 397) (%) (n = 316) (%) p-value
Importance of 0.3736 0.0051
considering SABV in
experimental design
Yes 140 61.4 144 65.5 251 63.2 231 73.1
No or unsure 88 38.6 76 34.5 146 36.8 85 26.9
Consideration of SABV 0.3278 0.0018
to improve rigor and
reproducibility
Yes 118 51.8 124 56.4 209 52.6 203 64.2
No or unsure 110 48.2 96 43.6 188 47.4 113 35.8
IMPLEMENTATION OF THE NIH SEX-INCLUSION POLICY 13
Downloaded by 193.56.67.114 from www.liebertpub.com at 02/06/19. For personal use only.
FIG. 2. Comparison of respondents’ perception of SABV policy implementation between 2016 and 2017.
policy places, ‘‘[an] unreasonable demand given the limited to determine whether the individual reviewers modified their
budgets allowed for NIH grants,’’ while comments pertaining score on Approach based on whether [sic] the criteria for sex
to animal use indicated that the SABV policy ‘‘requires in- as a biological variable,’’ and ‘‘Reviewers have different
creasing the number of experimental animals.’’ Comments that opinions on how much weight consideration of SABV should
addressed the implementation of the policy itself were mixed, have on the score of the application.,’’ While another re-
as some lauded the policy, as ‘‘a simple and needed criterion,’’ spondent shared that, ‘‘Consideration of SABV was not
while others were more critical citing it as, ‘‘[a] waste of time, usually a score-driving consideration.’’
not at all getting to the root of reproducibility problems.’’ In addition to our qualitative analysis, semiquantitative
Finally, respondents provided comments directly related to in silico sentiment analysis was utilized to interpret the re-
study section and the grant review process. Comments spanned spondents’ comments. In general, the overall comment sen-
multiple topics such as reviewer education (‘‘Our review of- timent analysis scores for both years were slightly negative
ficer [redacted] was really good in explaining the new policy to [median (quartile 1–quartile 3): -0.262 (-0.828–0) vs.
us. That is how I learned about how to incorporate it into my -0.181 (-0.726–0); p = 0.358, Table 4]. When analyzed by
proposals and my reviews of others proposals,’’), policy im- theme, comments pertaining to animal use or budget/funding
plementation (‘‘The applicants have gotten the message and issues received the lowest sentiment analysis scores, fol-
now include sex as a biological variable in their proposals.’’), lowed by comments related to the SABV policy itself.
and discussion of the policy itself (‘‘SABV was strongly em- Because we identified biological sex and experience con-
phasized in the premeeting discussions, and it was mentioned ducting sex-based research as two potential factors that in-
for every discussed application.’’). fluence views toward the SABV policy, we analyzed
However, respondent comments revealed variability in the comments by these variables. While there were no significant
way the policy impacts grant scores citing that, ‘‘It is difficult differences in 2016, sentiment analysis scores significantly
Table 4. Comparison of Respondents’ Comments Regarding the Sex as a Biological Variable Policy
by Sentiment Analysis Score Between 2016 and 2017
2016 2017
SA score median SA score median
N (quartile 1–quartile 3) N (quartile 1–quartile 3) p-value
Total 130 -0.262 (-0.828–0) 257 -0.181 (-0.726–0) 0.358
Comment themes
Animal use 14 -0.681 (-0.95–0.271) 30 -0.641 (-0.945–0.235) 0.915
Budget 32 -0.660 (-0.95–0.471) 72 -0.631 (-0.948–0.156) 0.268
Policy 63 -0.576 (-0.95–0) 133 -0.468 (-0.878–0) 0.337
Research area 41 0 (-0.802–0) 72 -0.001 (-0.559–0) 0.868
Review 58 0 (-0.414–0) 134 0 (-0.556–0) 0.188
14 WOITOWICH AND WOODRUFF
Table 5. Examples of Respondents’ Comments in the Policy Theme by Sentiment Analysis Score
Comment
Positive I think it is an important issue and am glad that the NIH is stressing its importance.
sentiment Sex consideration and incorporation in grant applications greatly improve scientific rigor
(SA > 0) and reproducibility, hence should be promoted!
Sex as a biological variable is important beyond rigor and reproducibility. It informs the fundamental
validity of the premise and observation. It is high time this was treated seriously and I hope NIH will
continue to increase its emphasis on this point.
Neutral Sex is a reasonable biological variable to require in many instances; however, it is just one.
sentiment The requirement should be broadened to reasonable biological variables like sex, age, environment, etc.
(SA = 0) Why is sex being singled out as the key biological variable to be considered for review purposes?
The idea of focusing on sex as a biological variable is important and necessary. However, like many
regulations, the implementation of a single rule to all areas of research is difficult and may eventually
undermine the utility of the rule.
Negative For some areas of research, this is nonsense. It’s not applicable and requiring it stands in the way
sentiment of making progress.
(SA < 0) It seems simplistic and blunt to have a policy requiring that sex be included/considered biological
variable in all NIH studies.
This policy ignores the complexity of sex as a biological variable and the cost and time that would
Downloaded by 193.56.67.114 from www.liebertpub.com at 02/06/19. For personal use only.
be required to adequately address this in every grant. The policy introduces a variable that cannot
be adequately controlled for with typical budget and capacity constraints and weakens rigor
of the majority of applications.
differed for males and females (median = -0.2193 vs. 0; in the 2017 cohort were significantly more likely to view the
p = 0.032), and those whose work involves the study sex dif- SABV policy as favorable compared to men. While women
ferences compared to those whose work does not (median = 0 are in the minority of study section membership (44% of our
vs. -0.2429; p = 0.007) in 2017 (Supplementary Table S6). respondents in 2017), they may see value of the policy at a
Table 5 provides representative examples of positive, negative, personal level, as sex-based research is directly implicated in
and neutral comments in the policy category theme. women’s health.6,12,15
The ability to leave open-ended comments provided study
section members with the opportunity to discuss issues not
Discussion
directly assessed in the survey. Sentiment analysis, utilized in
Sex, as it relates to the biological determination of males an attempt to remove potential author bias, revealed negative
and females, XX or XY, has largely been left out of biological feeling toward the policy. However, open-ended responses
studies not only in terms of inclusion but also in the analyses were entirely voluntary, and it may be that those with po-
and reporting of data.5,14,15 The NIH SABV policy was cre- larized feelings toward the policy were more likely to reply.
ated to address these inequities through the promotion of sex- Nevertheless, these comments provide a contextual frame-
inclusive research studies, in an effort to enhance the rigor work in which the policy can be evaluated.
and reproducibility of the biomedical research enterprise.16 Budgetary and/or funding concerns were woven through-
As the policy is relatively new, it is difficult to determine out many of the comments, citing that increased funds were
its direct impact on research output. Instead, we chose to needed to appropriately consider SABV. Also, many com-
survey the attitudes and opinions of NIH study section ments voiced concerns for animal welfare, stating that the
members who serve both as policy adjudicators and repre- SABV policy requires an increase in experimental animals.
sentatives of the biomedical research community. The results This is likely derived from the assumption that to consider
presented in this study suggest that awareness, perceived SABV, sample sizes must be doubled and female animals at
comprehension, and implementation of the NIH SABV pol- varying stages of the estrous cycle should be utilized. While
icy are improving over time, but that negative attitudes to- estrous cycle staging may be required to conduct experiments
ward the policy persist and additional training and/or specifically designed to explore hormonal influences on a
guidance may be warranted. given trait or condition, several studies reported that there is
Our data indicate that the SABV policy is accepted and no increase in intragroup variability between males and fe-
understood by a majority, but not all, of study section members. males, regardless of estrous cycle stage,17–19 thus allowing
We identified potential biases that may influence attitudes to- for the inclusion of both male and female subjects, without an
ward the SABV policy, such as field of study and sex. Re- increase in sample size (for a more detailed discussion of this
searchers who currently explore the study of sex differences topic, consult a recent review by Beery20).
were significantly more likely to agree that it is important to In terms of policy implementation, study section members
consider SABV and that it improves rigor and reproducibility. expressed differing viewpoints as to how the SABV policy
This may indicate a potential knowledge gap for investigators was to be evaluated and scored within the context of grant
who are less familiar with the role of sex in health and disease. review. This highlights an opportunity for increased reviewer
Also, sex of the study section member appears to influence education, although many scientific review officers and sec-
attitudes toward the policy as female study section members tion chairs were praised for their efforts to educate their
IMPLEMENTATION OF THE NIH SEX-INCLUSION POLICY 15
members on the policy. Finally, it came to our attention ately in their submissions. Yet, respondent comments re-
during review of the comments that numerous individuals vealed that negative attitudes toward the policy do exist.
misused the terms sex and gender, referring to gender in a As women are only now approaching parity in clinical
specific scientific context (i.e., cell or animal models) when research,5 25 years after the implementation of the NIH Re-
sex is the appropriate term.21 This could suggest that re- vitalization Act, we encourage the research community to
searchers may benefit from additional training and/or edu- view policies that examine the influences of both sex and
cation on sex- and gender-based research practices. gender on health and disease as scientific imperative. This
In terms of how the policy has been accepted by grant will help to ensure that the reproducibility of scientific reports
applicants, study section members found that the number of increase over time and that new discoveries using sex as a
grants that addressed SABV both in their research strategy biological variable are on the horizon.
and specifically within the experimental design, analyses, and
reporting has significantly increased since the policy was Acknowledgments
introduced. Likewise, there was a significant increase in the
number of grants, which provided consistent justification We thank the individuals who participated in this survey.
of the use of a single sex. These data are encouraging and We also thank Alexandra Rashedi and Megan Runge for their
demonstrate adherence to the policy. Future studies are assistance. This work was supported by the Thomas J. Wat-
needed to examine whether this initial outcome translates kins Memorial Endowed Chair (to TKW) at Northwestern
into sex-based or sex-inclusive research findings with direct University.
impacts on health and disease.
Author Disclosure Statement
Downloaded by 193.56.67.114 from www.liebertpub.com at 02/06/19. For personal use only.
13. National Institutes of Health. Roster Index for Regular 20. Beery AK. Inclusion of females does not increase vari-
Standing Study Sections and Continuing SEPs. Available ability in rodent research studies. Curr Opin Behav Sci 2018;
at: https://public.csr.nih.gov/studysections/standing/pages/ 23:143–149.
default.aspx Accessed October 17, 2018. 21. Clayton JA. Studying both sexes: A guiding principle for
14. Beery AK, Zucker I. Sex bias in neuroscience and biomed- biomedicine. FASEB J 2016;30:519–524.
ical research. Neurosci Biobehav Rev 2011;35:565–572.
15. Mansukhani NA, Yoon DY, Teter KA, et al. Determining if
sex bias exists in human surgical clinical research. JAMA
Surg 2016;151:1022–1030. Address correspondence to:
16. Clayton JA, Collins FS. Policy: NIH to balance sex in cell Teresa K. Woodruff, PhD
and animal studies. Nature 2014;509:282–283. Women’s Health Research Institute
17. Mogil JS, Chanda ML. The case for the inclusion of female Feinberg School of Medicine
subjects in basic science studies of pain. Pain 2005;117:1–5. Northwestern University
18. Prendergast BJ, Onishi KG, Zucker I. Female mice liber- 303 East Superior Street
ated for inclusion in neuroscience and biomedical research. Robert H. Lurie Medical Research Center
Neurosci Biobehav Rev 2014;40:1–5. 10th Floor Suite 10-121
19. Becker JB, Prendergast BJ, Liang JW. Female rats are not Chicago, IL 60611
more variable than male rats: A meta-analysis of neuro-
science studies. Biol Sex Differ 2016;7:34. E-mail: tkw@northwestern.edu
Downloaded by 193.56.67.114 from www.liebertpub.com at 02/06/19. For personal use only.