Crossing the Death Valley of publication process in cornea and ocular surface diseases: from abstracts

s submitted to national and international meetings represent an important opportunity to present the current research activities and to gain useful feedback from the scientific community. The subsequent submission to a scientific journal through a peer-review process allows a more rigorous evaluation of the data presented, as well as their dissemination to a global audience through international search databases. In general, the selection process of abstracts submitted to a meeting depends on a Scientific Committee that reviews and selects studies according to the potential scientific impact of the contribution. However, not all meeting abstracts end up being published in a peer-reviewed journal later. Previous studies have documented the publication rate of abstracts presented at national and international conferences of various specialties and have been used as an indicator of conference quality(1-5). A systematic review indicated that the publication rate of accepted abstracts presented at overall biomedical congresses ranged from 22% to 63%, depending on the methodology, specialty, author experience, and results of the single research(6). However, till date, very poor information is available regarding the publication out come of abstracts presented at international Ophthalmo logy Congresses(7). The aims of the present study were to investigate the (a) demographics of abstracts presented at the Association for Research in Vision and Ophthalmology (ARVO) 2013 Annual Meeting, (b) publication rate by peerreviewed journals, and (c) potential predictive factors for better outcomes of publication.


INTRODUCTION
Abstracts submitted to national and international meetings represent an important opportunity to present the current research activities and to gain useful feedback from the scientific community. The subsequent submission to a scientific journal through a peer-review process allows a more rigorous evaluation of the data presented, as well as their dissemination to a global audience through international search databases.
In general, the selection process of abstracts submitted to a meeting depends on a Scientific Committee that reviews and selects studies according to the potential scientific impact of the contribution. However, not all meeting abstracts end up being published in a peer-reviewed journal later. Previous studies have documented the publication rate of abstracts presented at national and international conferences of various specialties and have been used as an indicator of conference quality (1)(2)(3)(4)(5) . A systematic review indicated that the publication rate of accepted abstracts presented at overall biomedical congresses ranged from 22% to 63%, depending on the methodology, specialty, author experience, and results of the single research (6) . However, till date, very poor information is available regarding the publication out come of abstracts presented at international Ophthalmo logy Congresses (7) .
The aims of the present study were to investigate the (a) demographics of abstracts presented at the Association for Research in Vision and Ophthalmology (ARVO) 2013 Annual Meeting, (b) publication rate by peerreviewed journals, and (c) potential predictive factors for better outcomes of publication.

METHODS
In this retrospective cohort study, abstracts presented at the ARVO 2013 Annual Meeting (Seattle, WA, USA) in the Cornea scientific section were identified and retrieved from the archive website. For each abstract, the following data were recorded into a database: abstract number, authors' names, number of authors, affiliation of the first author, number of centers, sample size, main topic, research scope (basic science or clinical research), methodology (prospective or retrospective, randomized or nonrandomized, presence of a control group), and financial support.
A comprehensive literature search was then conducted using the PubMed-NCBI and Scopus databases. The search was restricted to the period between December 2012 (abstract submission deadline for ARVO 2013) and February 2018. Two independent investigators initially searched for the abstracts by combining the names of the first and last author. If no corresponding paper was found, the first author name was combined with various keywords from the title and the text of the abstract. The search was repeated for all possible keywords until either a match was found or all combinations were exhausted. Only publications with consistent hypotheses, study designs, and results were accepted as true matches. For each matching publication, the first author's name, time of publication, journal's name, and impact factor (IF) at the time of the publication were collected.
The SPSS statistical software (SPSS Inc, Chicago, IL) was used for data analysis. Univariate linear regression was performed to evaluate the association between variables and the likelihood of publication and between variables and the IF of the journal. A multivariate logistic regression and a stepwise linear regression were respectively performed to predict the publication rate to predict the IF of the journal, including independent variables that reached a significance level of <0.05 in the univariate analysis. A Kaplan-Meier analysis was performed to analyze the time course of the publication of abstracts. P<0.05 was considered to be statistically significant.

RESULTS
A total of 939 abstracts submitted to the ARVO 2013 Annual Meeting in the Cornea scientific section were included in this analysis. Table 1 shows the demographic characteristics of these abstracts. Regarding the geographic distribution of these abstracts, the affiliation of the first author spanned 41 different countries. In the majority of them, the affiliation was with an institution located in North America (531 abstracts; 56.6% of the total), followed by Europe (184 abstracts; 19.6%) and Asia (179 abstracts; 19.1%). South America, Oceania, and Africa were represented with 23 (2.5%), 22 (2.3%), and no abstracts, respectively ( Figure 1).
Of the analyzed abstracts, 360 (38.3% of the total) were published until February 2018. The number of authors during the publication process increased significantly from 5.0 ± 2.1 (mean ± standard deviation) in the abstracts to 6.2 ± 2.6 in the published articles (p<0.001). In 234 of the published articles (65%), the first author remained the same as that of the abstract, whereas in 126 studies (35%), there was a change in the first author. In particular, in 67 cases, the gender of the first author remained the same, whereas in 59 cases, there was a sex mismatch (36 cases from female to male, 23 cases from male to female). Journals with >1% of the total publications are listed in table 2. These 19 journals together accounted for 71.4% of all published articles. The median IF of the journals in which the abstracts were published was 3.4 (range 0-10.7). In total, 71.9% of the published abstracts appeared in journals related to the field of Ophthalmology. Among these, the official journals of the ARVO Society had the highest number of published articles (76; 21.1% of the total). Table 3 shows the number of abstracts grouped according to each main subsection along with the corresponding publication rate and the journal IF. There was no significant difference in the publication rate among A univariate binary logistic regression analysis was conducted to ascertain the effect of the variables on the likelihood that the abstracts were published (Table  4). Results showed that basic science studies had a significantly higher publication rate than clinical studies. Similarly, prospective, controlled, and financially supported studies had a higher publication rate. In contrast, multicenter studies had a lower publication rate than single-center studies. A multivariate logistic regression revealed that studies were more likely to get published if they were financially supported (OR=1.482, p=0.005), had a control group (OR=1.511, p=0.016), and had a basic science research scope (OR=1.388, p=0.020). The entire model was significant (p<0.001) and explained 33% of the variance in the publication rate.
A Venn diagram analysis was performed to explore the effect of the financial support, the presence of a control group, and the basic research scope on the publication rate (Figure 2). The publication rate was 41.5% when at least one of the three parameters was present (i.e., the abstract was financially supported and had a basic research scope or a control group) and increased to 47.3% when all the three parameters were present.
The median time interval between abstract submission and article publication was 22 months (95% confidence interval: 20.0-24.1). Figure 3 shows the results of Kaplan-Meier analyses for the time course of the publication of abstracts based on the research scope (Part A), the presence of a control group (Part B), and the funding support (Part C). A log-rank test disclosed a statistically significant difference in the publication time distribution of basic science vs clinical abstracts (χ 2 =7.636, p=0.006), controlled vs uncontrolled abstracts (χ 2 =6.921, p=0.009), and funded vs not funded abstracts (χ 2 =13.892, p=0.001).  A univariate linear regression analysis performed to explore the association between abstract variables and journal IF (Table 5) revealed that studies with a higher number of authors and financial support were published in a journal with a higher IF, whereas multicenter studies were published in a lower IF journal. A multivariate stepwise linear regression demonstrated that the journal IF was higher in case of financially supported studies (β=0.163, p=0.002) and lower for multicenter studies (β=-0.170, p=0.001). The entire model was significant (p<0.001) and explained 6% of the variance in the IF.

DISCUSSION
Dissemination of the data obtained from a research project generally begins with the presentation of an abs-tract at a meeting and often culminates with the subsequent complete publication in a peer-reviewed journal. Previous studies analyzing the publication outcomes of meeting abstracts had identified different characteristics of the abstracts that were associated with higher publication rates (1)(2)(3)(4)(5) . A recent investigation on the abstracts presented at the American Academy of Ophthalmology (AAO) 2008 Annual Meeting reported a publication rate of 39.1% and demonstrated that the factors that correlated with a higher publication rate were oral presentations focused on certain types of subspecialty (7) . Furthermore, a focus on rare diseases, affiliation with a US-based institute, and funding correlated with publication in journals with a higher IF.
In the present study, we analyzed the publication outcomes of approximately 1000 abstracts presented    (8) , we considered only full-text manuscripts with consistent hypotheses, designs, and results as true matches. We determined an overall publication rate of 38.3%, which was similar to that reported for the AAO 2008 Annual Meeting (7) but slightly lower than that reported from a Cochrane systematic review incorporating almost 30,000 abstracts from overall biomedical meetings (6) . More than two-thirds of the published articles were disseminated in journals related to the ophthalmic field, with the highest number being published in the official journals of the ARVO Society. The median IF of the overall journals in which the abstracts were published was 3.4. This value is slightly higher than the median IF re ported for papers originated from AAO abstracts (2.9) (7) . This result is even more significant considering that the number of ARVO abstracts presented in the Cornea section is similar to the total number of the overall AAO abstracts, irrespective of the topic. Although the IF is an indicator that suffers from some limitations, it is valued quite strongly with regard to the perceived quality of the journal (9) . A relatively high IF of the published ARVO abstracts may depend on the fact that some articles were published in basic science journals with a very high IF, even when the majority of them were published in ophthalmological journals.
Previous studies investigating the relationship between the geographic distribution of abstracts and their likelihood of publication have reported contradictory conclusions (7,10,11) . In our analysis, the first author in more than half of the abstracts was affiliated with an institu-tion located in North America. However, there was no association between the nationality of the first author and the publication rate or the journal IF.
An evaluation of the potential predictors of publication may provide valuable information for scientists approaching toward the preparation and submission of abstracts. In the present study, the parameters of financial support, presence of a control group, and basic research scope were associated with a higher likelihood of the abstracts being published. In particular, the publication rate increased to 47.3% for the abstracts that were financially supported and had a control group and a basic research scope.
Financially supported abstracts were also published in journals with a higher IF than non-supported studies, which is consistent with the results reported by Mimouni et al. (7) . This finding may be the result of the abundant resources provided to the researchers in case of funded studies. Moreover, the funding party may operate a filtering process, thereby granting financial support only to the studies considered as worth investigating. The financial sponsorship of clinical trials provided by pharmaceutical industries has increased significantly over the years (12) . Despite an increasing focus on the transparency surrounding financial conflict of interest, the role of pharmaceutical industries in the production of scientific evidence still represents a relevant concern for several researchers. Previous studies have demonstrated that trials funded by pharmaceutical industries are more likely to be associated with statistically significant proindustry findings (13)(14)(15) . The possible explanations include publication bias and the selection of an inappropriate comparator to the product being investigated.
Abstracts that had a basic (as opposed to clinical) research scope were more likely to get published. This finding is consistent with the results of a previous systematic review that analyzed almost 15,000 abstracts submitted to dozens of meetings (16) . The authors of that review hypothesized that basic and clinical research studies may differ in terms of both quality of conduction of the study and presentation of the results. However, the existence of a bias in favor of basic research cannot be excluded, as previous studies have emphasized the tendency of chairpersons and senior research advisors to perceive basic research activities of higher quality compared with clinical research activities (17) .
In general, a rigorous method for scoring the quality of a research is to assign an appropriate level of evidence according to the study design. In the present study, abstracts that had a control group exhibited a higher publication rate than uncontrolled studies. This result was not surprising, because in the hierarchical system of classifying evidence, controlled studies are deemed to have higher quality than uncontrolled studies (18) .
The data derived from this study might help young researchers when approaching toward the conception of a research project and later to the preparation and submission of an abstract. However, they should not forget that the ultimate aim of a biomedical research is not to generate publications anyhow but to add novel data that could be useful to improve the current understanding, diagnosis, and treatment of diseases.
The major limitation of this study is the lack of information concerning the reasons for unpublished abstracts. In fact, it would be interesting to analyze the number of abstracts that were not submitted for publication at all or those that did not survive the peer-review process. A previous study surveyed those authors whose abstracts were presented at an orthopedic meeting and never published as full publications. Only few authors had confirmed that their manuscripts were submitted and rejected, whereas more than one-third of them stated that they never submitted the abstract for full publication (19) . The common barriers to publication were the lack of time and interest for full publication and the difficulties in collaboration among co-authors. Another limitation of this study is associated with the fact that the academic title of the first author of the abstracts was not available, which consequently hampered the inclusion of this additional parameter in the predictive analysis.
Almost 40% of the abstracts accepted in the scientific section of Cornea of the ARVO 2013 annual meeting were published in peer-reviewed journals within 5 years from submission at the congress. The parameters associated with a higher publication rate were funding support, basic research scope, and the presence of a control group.