Data Withholding in Academic Genetics:  Evidence From a National Survey FREE

Without the free exchange of published scientific information and resources, researchers may unknowingly build on something less than the total accumulation of scientific knowledge or work on problems already solved.¹ However, a number of instances of data withholding (defining data to include the full range of research results, techniques, and materials useful in future investigations and withholding as the failure to share such published data) have been reported.^2- 7 A 1994-1995 survey of academic life scientists found that 34% of respondents were denied research results requested from a fellow university scientist in the previous 3 years, and 8.9% said they had denied a request from another university scientist for access to research results.⁸

Weinberg⁹ asserts that secrecy is more common in genetics and particularly human genetics than in other areas. Reasons may include the increased scientific competitiveness of the field and the opportunities for commercial applications.¹⁰ Research has shown that scientists who reported conducting research on goals similar to that of the Human Genome Project (HGP) were more likely to deny requests for information, data, and materials than were other life scientists.⁸

Understanding the withholding of information, data, and materials may be particularly important in genetics for a number of reasons. First, since academic geneticists publish more articles in peer-reviewed journals, teach more, and serve in more leadership roles in their university and discipline than do their colleagues in other biomedical specialties, the sharing and withholding practices of geneticists may have a disproportionate impact on university policy, the behavior of junior faculty, and the training and socialization of graduate students and postdoctoral fellows.¹¹

Second, understanding the role of genetics in human disease is believed to be important to the future of medicine.¹² Clearly, the progress made in mapping and sequencing the human genome represents a major step toward scientific breakthroughs in genetic-based diagnostics, preventive technologies, and therapeutics. The rate of progress in realizing these medical benefits may depend somewhat on the extent to which the results of genetic investigations flow freely among scientists in the field.

There is scant empirical evidence regarding sharing and withholding in academic genetics. For example, little is known about the extent to which geneticists share and withhold information and how these behaviors have changed over time. Nor do we know much about the reasons researchers withhold information, data, or materials from other academicians and what impact this behavior has on individual researchers or on the field of genetics as a whole. To address these issues, we conducted a national study of data sharing and data withholding in academic genetics, with a comparison group of other life sciences.

As in our previous work, the sample of 3000 life scientists was selected in a multistep process.^13- 14 First, using lists from the National Institutes of Health (NIH), we identified the 100 US educational institutions that received the most funding from the NIH in 1998. Second, at each institution we selected all departments and programs in genetics and human genetics and then randomly selected up to 3 additional life science departments and programs from lists of clinical (medicine, pathology, psychiatry, pediatrics, and surgery) and nonclinical (biochemistry, microbiology, pharmacology, physiology, and anatomy) departments. These specific clinical and nonclinical departments were selected for inclusion because they, among all departments, received the largest number of NIH grants in 1998. Third, using data from the Association of American Medical Colleges faculty roster system, Peterson's Graduate Programs in the Biological Sciences,¹⁵ school and individual Web sites, college bulletins, and direct contact with departments, we identified all full-time faculty at the rank of assistant professor and higher in each selected department and program. In addition, because of our special interest in genetics generally and human genetics specifically, we identified all faculty members who were principal investigators on at least 1 research grant from the HGP administered by the National Human Genome Research Institute (NHGRI) and the Department of Energy in the 5 years preceding the study (excluding those who received grants only from the Ethical, Legal, and Social Implications of Human Genetics Research program).

Finally, a stratified sample of 3000 faculty members was selected. The sample included all 219 grantees of the HGP and all 1547 faculty members in genetics or human genetics departments. The remainder of the sample (n = 1234) was randomly selected so that half came from nonclinical departments (n = 617) and half from clinical departments (n = 617). To avoid including clinical department faculty who were not actively engaged in research, we excluded clinical faculty who had not published at least 1 article in the MEDLINE database in the 3 years preceding the study.

The design of the survey instrument was informed by 2 focus group discussions, 20 semistructured interviews with knowledgeable biomedical researchers, discussions with colleagues, and a review of the literature. The survey instrument was pretested by using 9 cognitive interviews conducted by professional interviewers at the Center for Survey Research of the University of Massachusetts in Boston.

The Center for Survey Research administered the survey by mail between March and July 2000. Subjects were sent a letter, a fact sheet describing the study, a survey instrument, and a postage-paid postcard. They were asked to complete the survey and mail the postcard separately from the completed survey to the center. This process enabled us to track nonrespondents via the postcard while ensuring respondents' complete anonymity, since the survey instrument had no unique identifying information. Nonrespondents were mailed a letter encouraging their participation, mailed additional surveys, and then contacted by telephone and encouraged to participate. Of the potential 3000 respondents, 4 were in the sample twice, 7 had died, and 96 were ineligible because they were retired, out of the country, not located at the sampled institution, or lacking faculty appointments. Of the remaining 2893 subjects, 1849 responded, yielding an overall response rate of 64%.

In addition, 256 nonrespondents were interviewed briefly by telephone to determine how they differed from the respondents. Nonrespondents were significantly more likely than respondents to be full professors and less likely to be geneticists. They were also significantly more likely to receive a high number of requests for information, data, and materials related to their published research than the respondents.

Respondents identified themselves as geneticists by responding yes to the following question: "Do you consider yourself a genetics researcher? By genetics researcher we mean someone whose research involves any of the following: (1) identification of genomes, genes, or gene products in any organism; (2) study of the structure, function, or regulation of genes or genomes; (3) comparison of genes and genomes between species or populations." Respondents who answered no were considered other life scientists.

Geneticists were asked what their primary fields of genetic research were. The response categories were behavioral; biochemical; bioinformatics; cancer; common complex disorders (other than cancer); cytogenetics; developmental; mapping and sequencing; mutagenesis; pharmacogenetics; population genetics, evolution, and epidemiology; prenatal and perinatal; single gene (mendelian) disorders; structure and function (including genotype-phenotype relations, transgenesis); gene therapy; and other (respondents were asked to specify).

We asked geneticists what organism they worked with. The categories were amphibians, bacteria, Drosophila, fungi, humans (including materials of human origin), mammals (other than human), nematodes, plants, viruses, yeast, zebra fish, and other.

The survey used multiple measures of data withholding. First we asked geneticists how many times in the last 3 years they had asked other scientists to provide information, data, or materials concerning published research. We then asked those who had made such requests to estimate the number of times their requests were denied. Respondents who indicated that at least 1 of their requests was denied were considered to have had data withheld from them.

We also asked how many times in the last 3 years geneticists had received requests from other academic scientists for information, data, or materials concerning their published research. We then asked those who received such requests to estimate the number that they denied. Respondents who reported that they had denied another's request were considered to have engaged in postpublication data withholding.

We asked geneticists whether, in the last 3 years, they had requested any of the following from another academic scientist after a finding was published. The follow-up questions asked about additional information about laboratory techniques not included in the publication, pertinent findings that were not included in the publication, phenotypic information not included in the publication, genetic sequences not included in the publication, and biomaterials (probes, cell lines, tissues, reagents, and organisms) mentioned in the publication. For each of the follow-up questions, the response categories were "I made no request," "I received all of what I requested on every request," and "One or more of my requests was denied."

Respondents were asked the following: "On those occasions when you have intentionally withheld information, data, or materials about your published results from other academic scientists, how important was each of the following as a motivating factor?" The follow-up questions concerned their need to protect their ability to publish or that of a graduate student, postdoctoral fellow, or junior faculty member; their need to honor the requirements of an industrial research sponsor; their need to protect the commercial value of the results; their need to preserve patient confidentiality; the effort required to actually produce the materials or information; the financial cost of actually providing the materials or information transfer; and the likelihood that the other person would never reciprocate. The response categories were "very important," "moderately important," "not very important," "not at all important," and "does not apply." Responses were coded into dichotomous variables, with "very important" and "moderately important" coded as 1 and "not very important" and "not at all important" coded as 0.

The survey had 2 batteries of questions regarding the potential effects of withholding. The first asked whether, as a result of another academic scientist's failure to share information, data, or materials, the researcher had ever had a publication significantly delayed, been unable to confirm others' published research, abandoned a promising line of research, stopped collaborating with another academic scientist, complained to a funding agency, journal, or professional association, refused to share with that person or group, or delayed sharing with that person or group. The response categories were yes and no.

The second battery asked researchers how data withholding among academic scientists affected the progress of science in their field, the level of communication in their field, the education of students and postdoctoral fellows, the progress of their research, the quality of their relationships with other academic scientists, and their satisfaction with their professional career. The response categories were "no effect," "detracts somewhat," and "detracts greatly."

We asked respondents to estimate how the overall willingness of academic scientists in their area of research to share information, data, and materials had changed in the last decade. The response categories were "They are much more willing to share now," "They are somewhat more willing to share now," "Remained the same," "They are somewhat less willing to share now," and "They are much less willing to share now."

Several measures were used as control variables when geneticists were compared with nongeneticists on measures of withholding. The control variables included sex; whether the respondent trained in the United States; whether in the last 3 years the respondent had received research grants or contracts from companies whose work was related to their area of scientific expertise; whether their university research had resulted in any commercial activities, including a patent application, a patent granted, a patent licensed, a product under regulatory review, a product on the market, or a start-up company; the number of peer-reviewed articles they had published in the last 3 years (low, 0-5; medium, 6-15; high, ≥16); and if their research involved living humans as research subjects.

In addition, we created variables representing the volume of requests that respondents received (low, 1-6; high, ≥7) and made to others (low, 1-6; high, ≥7) in the last 3 years. These variables were used only when the likelihood of engaging in data withholding was examined.

Because we were primarily interested in withholding in academic genetics, the primary analytic group was the subsample of survey respondents who considered themselves geneticists. The responses of these self-identified geneticists were analyzed with standard statistical procedures to generate means for continuous variables and percentages for categorical or nominal variables. All analyses were weighted to adjust for differences caused by the likelihood of being sampled and for differences in nonresponse rates within survey strata. Differences in proportions were tested with logistic regression analyses. All analyses were conducted by using SUDAAN (Research Triangle Institute, Research Triangle Park, NC), a statistical package that correctly computes the SEs when determining statistical significance for survey data derived from complex sampling methods.

When comparing geneticists with other life scientists, we used multivariate logistic regression controlling for the effects of sex, whether the scientists were trained in the United States, whether they used humans as research subjects, the number of publications in the last 3 years, whether they had research funding from industry, and whether they had engaged in commercial activities. These variables were selected according to the results of our previous research into the causes of data withholding in the life sciences⁸ and the unpublished results of the personal interviews and focus groups.

In addition, for multivariate analyses that examined the likelihood of denying a request, we included variables representing the volume of requests received in the last 3 years. For analyses that examined the likelihood of having a request denied, we included a variable representing the volume of requests made of others in the last 3 years.

Of the 1849 life scientists who responded to the survey, 1240 considered themselves geneticists. Table 1 shows the characteristics of geneticists and nongeneticists among our respondents. Of the geneticists, three quarters were male, about half were full professors (49%), 27% were associate professors, and 23% were assistant professors. In terms of professional activities, 35% had engaged in commercial activities, 31% published 5 or fewer articles in the last 3 years, and 20% published 16 or more. Seventy-seven percent trained in the United States, and 30% reported that their research subjects included living humans.

Table 1 also shows that geneticists and other life scientists were similar in terms of their sex and academic rank. However, other life scientists were significantly more likely than geneticists to have trained in the United States, less likely to have research support from industry, and less likely to have engaged in commercial activities. In addition, nongeneticists were significantly more likely to conduct research involving living humans and also more likely to have a low number of publications in the last 3 years.

Among geneticists, the most frequently mentioned research field was structural and functional genetics (52%), followed by biochemical genetics (45%), mapping and sequencing (28%), cancer genetics (28%), developmental genetics (27%), mutagenesis (27%), single-gene disorders (16%), common complex disorders (15%), population studies (11%), bioinformatics (10%), and gene therapy (10%). The most frequently reported research organisms were nonhuman mammals (59%), humans (58%), bacteria (46%), yeast (19%), viruses (17%), and Drosophila (8%). Data regarding the research field and primary research organism of nongeneticists were not collected.

Eighty-four percent of geneticists had made at least 1 request in the previous 3 years of another academic researcher for additional information, data, or materials concerning published research. Ninety-two percent reported receiving such a request. Among those who made a request of another academician, 47% reported that at least 1 request was denied. However, only 12% denied a request they received from another academic researcher. Respondents estimated that they had made an average of 8.8 requests for information, data, or materials regarding published research in the previous 3 years, of which 10% of all requests were denied.

For perspective on the prevalence and consequences of data withholding in genetics, we compared the responses of geneticists with those of 600 other life scientists who responded to the survey (9 respondents did not classify themselves as either a geneticist or other life scientist). Statistically controlling for the independent effects of sex, academic rank, having trained in the United States, having industry research support, using humans as research subjects, engaging in commercial activities, and the number of publications in the last 3 years, we found that the odds of geneticists making a request for information, data, and materials were significantly higher than for other life scientists (odds ratio [OR], 4.28; 95% confidence interval [CI], 3.11-5.89) (Table 2). However, geneticists were no more likely than other life scientists to report that their requests were denied (OR, 0.97; 95% CI, 0.69-1.40) after the volume of requests made in addition to the control variables mentioned above were controlled for (Table 2).

The odds of geneticists having received a request for information, data, or materials were significantly higher than for other life scientists (OR, 5.41; 95% CI, 3.73-7.87). However, after the volume of requests received in addition to the control variables mentioned above were controlled for, geneticists were no more likely than other life scientists to report denying such requests (OR, 1.39; 95% CI, 0.81-2.40) (Table 2). The 2 factors that were significantly associated with an increased likelihood of denying others' requests were having received a high number of requests in the last 3 years (OR, 1.78; 95% CI, 1.10-2.90) and having engaged in commercial activities (OR, 1.72; 95% CI, 1.06-2.81).

Geneticists requesting biomaterials from another academician after a publication were most likely to report having a request denied (35%) compared with requesting sequence information (28%), pertinent findings (25%), phenotypic information (22%), and additional information regarding laboratory techniques not included in the publication (16%).

Figure 1 shows the reasons that geneticists, who constituted 67% of the respondents, gave for intentionally withholding information, data, or materials concerning their own published research from other scientists. Investigators who reported having 1 of their requests denied were not significantly more likely than those who had not had a request denied to say they had themselves denied a request (13% vs 8%; P = .09).

Twenty-eight percent of all geneticists reported that they had been unable to replicate published research as a direct result of another academic scientist's unwillingness to share information, data, or materials. Twenty-eight percent also reported that they had ended a collaboration as a result of withholding. Other consequences included having a publication significantly delayed (24%), abandoning a promising line of research (21%), delaying sharing with that person or group (18%), and refusing to share with that person or group (13%).

Seventy-seven percent of geneticists felt that data withholding detracted somewhat or greatly from the level of communication in science; 73%, that data withholding slowed the rate of progress in their field of science; and 63%, that data withholding harmed the quality of their relationships with peers. Geneticists also reported adverse effects of data withholding on their own research (58%), the education of students and postdoctoral fellows (56%), and their satisfaction with their careers (45%).

Other life scientists were significantly less likely than geneticists to report that data withholding detracted from the progress of their research (38% vs 58%, respectively; P<.001) and the overall level of progress in their scientific field (56% vs 73%, respectively; P<.001).

Thirty-five percent of geneticists thought other academic scientists were somewhat or much less willing to share information, data, and materials compared with a decade ago; 51%, that willingness had remained unchanged; 14%, that it had increased.

We found that 95% of HGP-funded geneticists had received a postpublication request for information, data, and materials compared with 92% of non–HGP-funded geneticists (P = .20). Among those who received a request, HGP-funded geneticists were no less likely to deny a request than were non–HGP-funded geneticists (15% vs 12%; P = .46).

This study provides the first detailed, systematic, quantitative portrait of the phenomenon of data withholding in genetics or any other field of academic investigation. Our findings suggest that data withholding in genetics is not widespread, given than 12% of geneticists reported denying requests from other academicians for information, data, and materials. At the same time, the impact of withholding appears to be much more widespread, given that almost half of all geneticists who had made a request of another academic for information, data, and materials related to published research had had that request denied. Further, it is possible that withholding has increased in recent years because more than one third (35%) of geneticists also believe that data withholding is becoming more common in their field.

A critical question is whether withholding affects the daily work of genetics investigators or the health of this field of investigation. A number of respondents reported adverse effects on their ability to reproduce the work of other investigators, the timeliness of their own publications, and their ability to pursue chosen research directions. Further, large numbers of geneticists see adverse effects on communication within their field, the education of young scientists, and the rate of scientific progress. The adverse effects on research progress at the individual and field level were more likely to be reported by geneticists than by investigators who had experienced data withholding in other life science fields.

Although genetics investigators are clearly concerned about the effects of data withholding on their field, its significance for the well-being of genetics in particular and the scientific enterprise generally remains to be fully explored. The progress of the HGP and daily reports of advances in genetics suggest the field remains vibrant. This and previous work indicating that geneticists publish more than other investigators and are more productive in other academic domains lends credence to this appearance of health and dynamism.¹⁴ Data withholding may paradoxically occur most commonly during extremely rapid progress, since scientists are generating large numbers of new findings that stimulate much jockeying for scientific priority. The commercial applications of genetics research, along with increasing dependence on industry funding and the rise of commercial norms in the academy, may be partially responsible as well for data withholding.

From a policy perspective, the question is not whether progress in genetics continues, but whether it is as rapid as it could be if data sharing were maximized. The NHGRI has taken a leadership position in encouraging openness among its investigators by encouraging the rapid release and dissemination of new sequence data by its funded investigators. A number of journals also require as a condition of publication that data and materials be placed in public depositories or otherwise made available to other scientists. Unfortunately, our research suggests that these inducements not have been sufficient to prevent the adverse effects of data withholding on genetics as a field. Our findings also do not suggest that data withholding is less common among HGP-funded investigators, despite the NHGRI's laudable efforts to encourage sharing.

This result makes particularly relevant our findings on the types of data that are most commonly withheld and the reasons scientists give for doing so. First, scientists are most likely to encounter refusals when they approach other academic investigators for access to biomaterials. Some of these denials likely stem from the scarcity of precious materials or from human subjects concerns. However, it may be that material transfer agreements have become so complex and demanding that they inhibit sharing.¹⁶ If so, our findings suggest a pressing need to clarify and expedite the process for sharing biomaterials.

Second, some of the reasons respondents gave for refusing requests may be remediable through several potential interventions. These reasons include the time and effort required to comply with requests and the actual costs of doing so. In principle, these obstacles to sharing could be reduced by providing investigators additional resources that are explicitly devoted to disseminating the results of their research after publication. Traditionally, federal sponsors of biomedical research seem to have assumed that investigators' commitment to the norms of science would ensure that they would absorb the costs of honoring those values, including costs of data-sharing activities. Under the circumstances of modern research, this assumption may no longer be reasonable. Additional funds may not sufficiently address all of the reasons for data withholding, such as protecting one's commercial and academic priority.

We found that geneticists were no more likely than nongeneticists to deny requests for access to published information, data, and materials from other faculty.⁸ This finding was unlike that in our previous research, probably because our previous analyses did not control for the number of requests received, which significantly affects the likelihood of denials for those receiving a high number of requests. However, as in our previous research, having engaged in commercialization of university-based research was significantly associated with increased likelihood of data withholding.⁸

This study has several limitations inherent in survey research. Because we relied on self-reporting, our estimate of the percentage of faculty who withheld results from others likely constitutes a lower bound estimate of the proportion who actually participate in this behavior, since respondents are often reticent to admit engaging in behavior that may be perceived as less than desirable.^17- 18 Further, since the 256 nonrespondents were significantly more likely than respondents to have received a high number of postpublication requests (results not shown) and since receiving a high number of requests is a significant predictor of denials (Table 2), nonrespondents may have been more likely to deny a request than respondents, which would have resulted in an underestimate of the percentage of geneticists who denied others' requests. Another limitation is that our respondents may not represent the universe of academic geneticists, since half reported that they were full professors, suggesting that our sample may be tilted toward more senior faculty. Similarly, since our sample included only research-intensive universities, our results may not be applicable to faculty in institutions receiving less extramural research support. Finally, this report examines only 1 form of data withholding: refusal to share data associated with published research results. Other forms of secrecy, such as significant delays in honoring requests, delays in publication, refusals to publicly present research findings, and not discussing research with others, may also affect the progress of science and need to be factored into future research and policy formulation in genetics and other fields of investigation.

Many questions about the prevalence and consequences of data withholding in genetics and other disciplines remain unanswered. However, our findings suggest that data withholding occurs in one of the most salient scientific disciplines of our times, that current efforts to reduce it have not been fully effective, and that additional measures to improve openness of communication in genetics and the sharing of published information data and materials seem justified.