Content uploaded by Seog Hee Park
Author content
All content in this area was uploaded by Seog Hee Park on Jan 29, 2016
Content may be subject to copyright.
Seong Su Hwang, MD
Hae Hiang Song, PhD
Jun Hyun Baik, MD
So Lyung Jung, MD
Seog Hee Park, MD
Kyu Ho Choi, MD
Young Ha Park, MD
Index terms:
Education
Radiology and radiologists, research
Published online before print
10.1148/radiol.2261011255
Radiology 2003; 226:16–23
Abbreviation:
ICMJE ⫽ International Committee of
Medical Journal Editors
1
From the Department of Radiology,
St Vincent’s Hospital (S.S.H., J.H.B.,
S.L.J., Y.H.P.), Department of Radiol-
ogy, Kangnam St Mary’s Hospital
(S.H.P., K.H.C.), and Department of
Biostatistics (H.H.S.), Catholic Univer-
sity of Korea, 505 Banpo-dong, Seo-
cho-ku, Seoul 137-701, Korea. Received
July 23, 2001; revision requested Sep-
tember 24; final revision received April
15, 2002; accepted May 17. Address
correspondence to H.H.S. (e-mail:
hhsong@catholic.ac.kr).
See also the From the Editor article
in this issue.
Author contributions:
Guarantors of integrity of entire study,
H.H.S., S.H.P.; study concepts, S.S.H.,
S.H.P., K.H.C.; study design, S.S.H.,
H.H.S., Y.H.P.; literature research,
S.S.H.; data acquisition, S.S.H., J.H.B.,
S.L.J.; data analysis/interpretation, S.S.H.,
J.H.B., S.H.P.; statistical analysis, H.H.S.,
J.H.B.; manuscript preparation, S.S.H.,
H.H.S.; manuscript definition of intellec-
tual content, H.H.S., S.H.P.; manuscript
editing, H.H.S., S.S.H.; manuscript revi-
sion/review and final version approval, all
authors.
©
RSNA, 2002
Researcher Contributions and
Fulfillment of ICMJE
Authorship Criteria: Analysis
of Author Contribution Lists
in Research Articles with
Multiple Authors Published
in Radiology
1
PURPOSE: To determine the number of researchers who fulfill the International
Committee of Medical Journal Editors (ICMJE) authorship criteria and to evaluate
individual contributions.
MATERIALS AND METHODS: The author contribution lists of Radiology articles
published between 1998 and 2000 with at least three authors were reviewed. The
fulfillment of ICMJE criteria for authorship and the contribution percentage were
assessed according to each researcher’s position in the byline and nationality (Amer-
ican vs international), number of researchers per article, and year of publication.
RESULTS: Sixty-eight percent of researchers fulfilled the ICMJE authorship criteria.
Position in the byline indicated a significant difference in fulfillment (P ⬍ .001):
98.9% and 85.3% for the first and second authors, respectively, and 52.8% and
66.5% for the middle and last authors, respectively. American researchers had a
higher percentage (78%) of fulfillment than did international researchers (57%)
(P ⬍ .001). Fulfillment decreased as the number of authors per article increased (P ⬍
.001), although there was no significant change throughout 1998 –2000. The mean
contribution percentages decreased greatly from first to second to last to middle
authors. American researchers had a significantly larger mean contribution percent-
age than did international researchers. Of the total 6,686 researchers, 2,316 (35%)
contributed to one or two categories. This rate was higher for middle and interna-
tional authors.
CONCLUSION: The 68% fulfillment of criteria for authorship was closely related to
the large number of researchers contributing to one category or to categories
belonging to the same ICMJE criterion.
©
RSNA, 2002
Publication is an important academic process. As the number of authors per original biomed-
ical article has increased, problems regarding irresponsible authorship have been raised,
including misattribution of authorship credit, such as guest or ghost authorship, lack of
guarantors, and ambiguity in the meaning of the order of author names in the byline (1–6).
In 1985, the International Committee of Medical Journal Editors (ICMJE) established the
criteria for authorship, which were modified during the May 2000 ICMJE meeting. In the
criteria for authorship, the ICMJE states the following (7):
All persons designated as authors should qualify for authorship, and all those who qualify
should be listed. Each author should have participated sufficiently in the work to take
Special Reports
16
R
adiology
public responsibility for appropriate por-
tions of the content. One or more au-
thors should take responsibility for the
integrity of the work as a whole, from
inception to published article. Author-
ship credit should be based only on 1)
substantial contributions to conception
and design, or acquisition of data, or
analysis and interpretation of data; 2)
drafting the article or revising it critically
for important intellectual content; and 3)
final approval of the version to be pub-
lished. Conditions 1, 2, and 3 must all be
met. Acquisition of funding, the collec-
tion of data, or general supervision of the
research group, by themselves, do not
justify authorship.
The order of authors in the byline pro-
vides little information regarding each
author’s contribution in research articles
with multiple authors. To disclose a re-
searcher’s specific contributions for jour-
nal readers, the use of a contribution list
was proposed (1,8) and adopted by The
Lancet in 1997. In this disclosure, the au-
thors were not required to use any pre-
defined categories or checklists of contri-
butions but were free to devise their own
descriptions of the tasks each author per-
formed. Subsequently, a few journals be-
gan to publish author contributions, and
Radiology is the only imaging journal that
does. Radiology has published author con-
tributions in the form of a list of specific
categories since 1998. This information is
published on the first page of each origi-
nal research article (9). Unlike in the con-
tribution list of The Lancet, each author
of an article published in Radiology must
select from among 14 categories those
that best fit the tasks that he or she per-
formed. Radiology also requests that guar-
antors be identified—that is, those who
have responsibility for the integrity of
the work as a whole (10).
Yank and Rennie (11) performed a pre-
liminary analysis of the author contribu-
tion lists in The Lancet during a 6-month
period from July to December 1997 and
suggested that their results needed to be
replicated in larger studies of other jour-
nals. In this spirit, we initiated our
study—we analyzed author contribution
lists from articles published during a 3-year
period in Radiology. The purpose of this
study was to determine the number of re-
searchers who fulfill the ICMJE authorship
criteria and to evaluate the extent and na-
ture of individual contributions.
MATERIALS AND METHODS
The study samples were original research
articles with multiple authors published
in Radiology from January 1998 to De-
cember 2000. Three radiologists (S.S.H.,
J.H.B., S.L.J.) collected the data from the
author contribution lists that appear on
the first page of the articles. To evaluate
multiple authorship, we limited our study
to articles with three or more authors,
excluding four articles by a single author
and 42 articles by two authors. Four ad-
ditional articles were excluded because a
name listed in the byline did not appear
in the contribution list. Eleven articles
containing the name of a research group
as an author were also excluded from our
study, since information on each au-
thor’s contribution was not available. A
total of 1,068 original research articles
written by a total of 6,686 researchers
were collected.
To determine authorship trends among
researchers listed in the bylines, research-
ers were grouped as first, second, middle,
and last authors. The last researcher in
the byline was defined only as the “last
author.” Middle authors were those not
listed as first, second, or last. Thus, for
articles with three authors, there was no
middle author. Also, to assess geographic
differences in authorship trends, research
articles were defined as having either
American or international authorship
according to researcher nationality, as
determined by the institution(s) docu-
mented in the article. In articles pub-
lished in collaboration with both Ameri-
can and international authors, the
nationality was that of the first author.
How Many Researchers Fulfill
ICMJE Criteria for Authorship?
There are 14 categories in the author
contribution list for Radiology. To study
the association between this contribu-
tion list and the three conditions of the
ICMJE authorship criteria, we identified
categories in Radiology that applied to
each condition. We consider that contri-
bution to any of the following eight cat-
egories would fulfill the first condition of
the ICMJE criteria for authorship—that
is, any aspect of conception and design,
acquisition of data, or analysis and inter-
pretation of data: “study concepts,”“study
design,”“literature research,”“clinical
studies,”“experimental studies,”“data
acquisition,”“data analysis/interpreta-
tion,” and “statistical analysis.” The sec-
ond condition of the ICMJE criteria—
that is, any part of drafting the article or
revising it critically for important intel-
lectual content, includes four categories:
“manuscript definition of intellectual
content,”“manuscript preparation,”“manu-
script editing,” and “manuscript revi-
sion/review.” The third condition of the
ICMJE criteria, which consists of final ap-
proval of the version to be published,
includes a single category: “manuscript
final version approval.”
In a one-to-one comparison of the Ra-
diology contribution list categories and
ICMJE conditions, the category of “manu-
script final version approval” in Radiology
poses a problem, since this category was
first included in the contribution list in
December 2000. Thus, we presumed that
all authors of Radiology whose manu-
script was published from 1998 to 2000
justly approved their final version, im-
plying that all authors fulfilled the third
condition for authorship criteria. This is
not unreasonable, because the copyright
transfer form signed by all authors stipu-
lates it.
In addition, the category of “guarantor
of integrity of entire study” was omitted
from our comparison, since it was not
included in the ICMJE conditions. In
summary, by excluding “guarantor of in-
tegrity of entire study” and “manuscript
final version approval,” we considered 12
categories in our study of fulfillment of
ICMJE’s authorship criteria. We ascer-
tained the percentage of researchers who
fulfilled the first two conditions of the
ICMJE criteria, and these percentages
were compared with the researchers’ po-
sitions in the byline, researchers’ nation-
ality, number of researchers per article,
and the year of publication.
Assessment of Extent and Nature
of Contributions
In the assessment of the percentages or
types of each researcher’s specific contri-
butions, there is no need to exclude the
category of “guarantor of integrity of en-
tire study,” unlike in the study of the
fulfillment for authorship criteria. We
wanted to find out which author had a
role as “guarantor of integrity of entire
study.” Also, since we examined data
from 1998 to 2000 in Radiology, we ex-
cluded the category of “manuscript final
version approval” for the reason ex-
plained above.
In summary, for the study of the ex-
tent and nature of researcher contribu-
tions, we considered 13 categories by
excluding “manuscript final version ap-
proval.” To assess the extent of the con-
tributions of each researcher to an article
with respect to the 13 categories, we de-
veloped a “contribution percentage.” The
actual contribution of an author is diffi-
cult or even impossible to quantify. How-
Volume 226
䡠
Number 1 Researcher Contributions and Fulfillment of Authorship Criteria
䡠
17
R
adiology
ever, the contribution categories checked
for each author in an article provide some
indication of workload, and thus, a mea-
sure of “contribution percentage” can be
computed. The relative individual contri-
bution of an author is computed on the
basis of the number of categories in which
the author participated, when the total
number of categories checked in that arti-
cle is counted as a 100% contribution. On
the basis of the assumption that each cat-
egory in an article implies equal contribu-
tion, the contribution percentage of an au-
thor was calculated as 100m/n, where m is
the number of categories checked by that
author and n is the number of categories
checked by all authors in that article. For
example, a contribution percentage of 90%
for a researcher means that when the total
number of categories checked by all re-
searchers in that article is 10, the researcher
participated in nine categories.
We compared the mean contribution
percentage of all researchers according to
position in the byline and nationality. In
addition, we evaluated the percentages and
types of contributions of researchers who
contributed to one or two categories and
compared them according to each re-
searcher’s position in the byline or nation-
ality.
Statistical Analysis
Data were analyzed by using the
2
test
or analysis of variance with Scheffe´ mul-
tiple comparison methods (12). Statisti-
cal significance depended on the total
number of samples, and for a large sam-
ple, even a slight difference of percent-
ages or means can be highly significant.
In our study, sample size was determined
on the basis of the number of articles or
the number of researchers. Therefore, the
contribution percentage or mean differ-
ence in the number of researchers, which
was 6,686 in our study, was calculated
more conservatively with a significance
level of .01. However, for a statistical eval-
uation of the number of articles, the usual
significance level of .05 was applied.
RESULTS
A total of 1,068 original research articles
published in Radiology from 1998 through
2000, authored by a total of 6,686 re-
searchers listed in the bylines of these
articles, with each article having at least
three authors, were studied. Twenty-six
(5.3%) of 494 international articles and
four (0.7%) of 574 American articles were
published in collaboration with both
American and international authors and
did not constitute a significant bias in
our statistical analysis. The number of
authors per article varied widely from
three to 14, with a mean of 6.3 authors ⫾
2.3 (SD) per article. The mean number of
authors per article published by Ameri-
can authors was significantly less (mean,
5.9 ⫾ 2.0) than that by international au-
thors (mean, 6.7 ⫾ 2.0) (P ⬍ .001).
There was variation in the number of
contribution categories per article, with a
mean of 11.4 categories ⫾ 1.2. Nine hun-
dred seventy-seven (92%) of the 1,068
articles assessed had between 11 and 13
categories when the category of “manu-
script final version approval” was not in-
cluded. Despite these variations, how-
ever, fulfillment of authorship criteria
results and researcher contribution per-
centages did not correlate with the num-
bers of documented contribution catego-
ries (Table 1) (P ⬎ .01 for both).
Fulfillment of ICMJE Criteria
A total of 2,172 (32.5%) of the 6,686
authors appearing in the bylines did not
fulfill the ICMJE criteria for authorship, ac-
cording to assessment of 12 contribution
categories (excluding the categories of
“guarantor of integrity of entire study” and
“manuscript final version approval”) (Ta-
ble 2). According to researcher position in
the byline, the percentages of researchers
who fulfilled all three conditions of the
ICMJE criteria were significantly different
(P ⬍ .001): 98.9% and 85.3%, respectively,
for the first and second authors, and 52.8%
and 66.5%, respectively, for the middle
and last authors.
For the first condition of ICMJE crite-
ria, which comprises conception and de-
sign, acquisition of data, and analysis
and interpretation of data, the percent-
age of researchers who fulfilled it was
more than 90% for all researcher posi-
tions in the byline, except for the last
author, with 80%. However, the percent-
ages of authors fulfilling the second con-
dition (drafting and revising the article
for important intellectual content) varied
considerably, depending on researcher
position in the byline, with 99% for the
first author and 62% for middle authors.
The geographic difference of the per-
centages of researchers who fulfilled all
three conditions for authorship criteria
was significant, with a larger percentage
(78%) of American researchers compared
with that (57%) of international re-
searchers (P ⬍ .001). This difference was
mainly due to the second criterion, with
a higher percentage (84%) of fulfillment
for the American researchers compared
with that (68%) of the international re-
searchers. The relationship between the
percentages of fulfillment of ICMJE crite-
ria and the number of researchers per
article was negatively associated, with
the percentages of fulfillment decreasing
as the number of authors per article in-
creased (P ⬍ .001).
The percentages of authors who ful-
filled all three conditions for authorship
did not change throughout 1998 –2000
TABLE 1
Contribution Percentage per Number of Author Contribution Categories
No. of Categories
per Article
No. of Articles
(n ⫽ 1,068)
No. of Authors
(n ⫽ 6,686)
No. of Authors Who
Fulfilled Authorship Criteria
Mean Contribution
Percentage of All Authors*
ⱕ8 15 (1.4) 95 (1.4) 51 (53.7) 38.1 ⫾ 23.1
9 27 (2.5) 177 (2.6) 118 (66.7) 39.5 ⫾ 26.8
10 49 (4.6) 288 (4.3) 183 (63.5) 42.1 ⫾ 28.9
11 279 (26.1) 1,669 (25.0) 1,130 (67.7) 41.9 ⫾ 29.1
12 581 (54.4) 3,655 (54.7) 2,506 (68.6) 39.4 ⫾ 27.7
13 117 (11.0) 802 (12.0) 526 (65.6) 38.1 ⫾ 27.5
Total 1,068 (100) 6,686 (100) 4,514 (67.5) 40.0 ⫾ 28.0
Note.—Numbers in parentheses are percentages.
* Data are mean ⫾ SD.
18
䡠
Radiology
䡠
January 2003 Hwang et al
R
adiology
(66.3% in 1998, 68.9% in 1999, and
67.4% in 2000) (P ⫽ .18); this was true
even when the first or second criterion
was examined separately.
Extent and Nature of Contributions
Thirteen of the 14 contribution catego-
ries (excluding “manuscript final version
approval”) were analyzed, and the extent
of contribution was evaluated by using
the contribution percentage of each re-
searcher as noted above. The mean per-
centages shown in Table 3 are classified
according to the number of researchers
per article, researcher position in the by-
line, and researcher nationality. Mean
percentages decreased from 61% for arti-
cles with three researchers to 31% for
articles with more than 10 researchers,
with an overall mean contribution of
40% ⫾ 28. These mean percentages var-
ied greatly depending on researcher posi-
tion in the byline, which decreased from
82% to 51% to 35% to 25% for the first,
second, last, and middle authors, respec-
tively. For each researcher position in the
byline, a pattern of decreasing mean per-
centages persisted as the number of re-
searchers per article increased. Overall,
the mean contribution percentage of
American researchers (45%) was signifi-
cantly higher than that of international
researchers (35%) (P ⬍ .001).
In Table 4, a researcher’s specific con-
tribution to each of 13 categories is
shown in terms of the number of contrib-
uting authors in relation to the total
number of authors of articles in which
that category was checked in the contri-
bution list. Overall, researchers contrib-
uted most frequently to “manuscript
revision/review” (59%) and “clinical
studies” (54%) and contributed least fre-
quently to “statistical analysis” (25%)
and “guarantor of integrity of entire
study” (27%). For the American and in-
ternational articles, the most frequently
indicated categories differ—that is, inter-
national researchers indicated that they
participated most frequently in the cate-
gories of “clinical studies” (50%) and
“manuscript revision/review” (47%), but
the categories with more than 50% par-
ticipation for the American researchers
were “manuscript revision/review” (72%),
“manuscript editing” (58%), “clinical
studies” (59%), “experimental studies”
(52%), “data acquisition” (55%), and
“data analysis/interpretation” (55%).
The categories in which authors partici-
pated varied greatly according to author
order in the bylines. The first authors’ prin-
cipal contributions were noteworthy in all
13 categories, with the percentages exceed-
ing 60% for each category. Also, the cate-
gories of “guarantor of integrity of entire
study” and “literature research” are listed
as the major contributions of the first au-
thors, compared with the contributions of
non–first authors. The second authors’
contributions ranged from 28% to 67% in
the 13 categories. On the other hand, the
last authors’ principal contribution was
“manuscript revision/review” (76%); the
last authors were also least involved in
“statistical analysis” (13%). Overall, last au-
thors’ contributions were less frequent in
all 13 categories than those of the second
authors. The middle authors’ major contri-
bution categories were “clinical studies”
(51%), “manuscript revision/review” (50%),
“data acquisition” (41%), and a minimal
contribution to “guarantor of integrity of
entire study” (6%).
Researchers Contributing to One or
Two Categories
In Table 5, 1,133 (17%) of 6,686 re-
searchers contributed to a single category,
and 1,183 (18%) researchers contributed to
TABLE 2
Fulfillment of ICMJE Authorship Conditions
Article Characteristics
ICMJE Conditions*
P Value
‡
First Second Third (%)
†
All
Author position in byline ⬍.001
First (n ⫽ 1,068) 1,065 (99.7) 1,058 (99.1) 100 1,056 (98.9)
Second (n ⫽ 1,068) 1,043 (97.7) 934 (87.5) 100 911 (85.3)
Last (n ⫽ 1,068) 858 (80.3) 907 (84.9) 100 710 (66.5)
Middle (n ⫽ 3,482) 3,149 (90.4) 2,160 (62.0) 100 1,837 (52.8)
Author nationality ⬍.001
American (n ⫽ 3,364) 3,185 (94.7) 2,809 (83.5) 100 2,631 (78.2)
International (n ⫽ 3,322) 2,930 (88.2) 2,250 (67.7) 100 1,883 (56.7)
Year of publication .18
1998 (n ⫽ 2,152) 1,985 (92.2) 1,586 (73.7) 100 1,426 (66.3)
1999 (n ⫽ 2,176) 2,002 (92.0) 1,667 (76.6) 100 1,499 (68.9)
2000 (n ⫽ 2,358) 2,128 (90.2) 1,806 (76.6) 100 1,589 (67.4)
No. of authors per article ⬍.001
3(n ⫽ 312) 299 (95.8) 289 (92.6) 100 276 (88.5)
4(n ⫽ 532) 509 (95.7) 464 (87.2) 100 441 (82.9)
5(n ⫽ 925) 852 (92.1) 787 (85.1) 100 717 (77.5)
6(n ⫽ 1,266) 1,156 (91.3) 1,004 (79.3) 100 898 (70.9)
7(n ⫽ 1,029) 939 (91.3) 748 (72.7) 100 661 (64.2)
8(n ⫽ 1,024) 919 (89.7) 697 (68.1) 100 595 (58.1)
9(n ⫽ 684) 611 (89.3) 456 (66.7) 100 390 (57.0)
10 (n ⫽ 400) 368 (92.0) 256 (64.0) 100 225 (56.3)
⬎10 (n ⫽ 514) 462 (89.9) 358 (69.6) 100 311 (60.5)
Total (n ⫽ 6,686) 6,115 (91.5) 5,059 (75.7) 100 4,514 (67.5)
Note.—Numbers in parentheses are percentages.
* First condition includes contributions to conception and design, acquisition of data, and analysis and interpretation of data; second condition
includes contributions to drafting the article and revising it critically for important intellectual content; third condition includes contributions to final
approval of the version to be published.
†
It was assumed that the third condition was fulfilled on the basis of its inclusion on the copyright transfer form all authors must sign.
‡
P values obtained by performing the
2
test.
Volume 226
䡠
Number 1 Researcher Contributions and Fulfillment of Authorship Criteria
䡠
19
R
adiology
two categories. The percentages of re-
searchers with contributions to one or
two categories were significantly higher
for middle and international authors
(P ⬍ .001 for both). The authors contrib-
uting to one category most commonly
participated in “clinical studies” (348 of
1,133, 30.7%), “data acquisition” (212 of
1,133, 18.7%), and “manuscript revision/
review” (170 of 1,133, 15.0%), with par-
ticipation in the rest of the categories at
1.3% to 6.4%. Twenty-six authors (1.4%)
contributed to the single category of
“guarantor of integrity of entire study,”
and 25 of the 26 were international au-
thors.
Contributions to two categories were
made by 1,183 (18%) of 6,686 research-
TABLE 3
Contribution Percentage according to Number of Authors per Article
No. of
Authors
per
Article
All Authors
Mean Contribution Percentage per Author Position
in Byline (%)*
American Authors International Authors
No. of
Articles
Mean
Contribution
Percentage
of All
Authors*
First
(n ⫽ 1,068)
Second
(n ⫽ 1,068)
Last
(n ⫽ 1,068)
Middle
(n ⫽ 3,482)
No. of
Articles
Mean
Contribution
Percentage
of All
Authors*
No. of
Articles
Mean
Contribution
Percentage
of All
Authors*
3 103 (9.6) 61.2 ⫾ 26.9 83.4 ⫾ 19.0 54.9 ⫾ 23.0 47.2 ⫾ 21.9
...
75 (13.1) 62.8 ⫾ 26.7 28 (5.7) 57.1 ⫾ 26.9
4 134 (12.5) 53.0 ⫾ 28.8 85.2 ⫾ 14.1 53.6 ⫾ 22.8 37.4 ⫾ 22.7 34.3 ⫾ 21.3 86 (15.0) 53.7 ⫾ 29.3 48 (9.7) 51.6 ⫾ 27.8
5 184 (17.2) 47.6 ⫾ 28.4 84.7 ⫾ 16.6 50.5 ⫾ 23.7 38.2 ⫾ 23.4 30.8 ⫾ 17.7 116 (20.2) 49.7 ⫾ 28.4 68 (13.8) 44.2 ⫾ 28.1
6 210 (19.7) 41.8 ⫾ 27.7 82.0 ⫾ 17.4 50.5 ⫾ 23.7 34.8 ⫾ 22.7 27.8 ⫾ 17.3 114 (19.9) 46.3 ⫾ 27.7 96 (19.4) 36.5 ⫾ 26.8
7 149 (14.0) 36.7 ⫾ 26.6 79.3 ⫾ 20.3 48.8 ⫾ 23.3 31.9 ⫾ 21.3 24.3 ⫾ 15.5 71 (12.4) 40.0 ⫾ 27.1 78 (15.8) 33.7 ⫾ 25.8
8 128 (12.0) 33.2 ⫾ 25.4 76.7 ⫾ 18.0 49.4 ⫾ 23.0 30.3 ⫾ 22.5 21.8 ⫾ 14.3 44 (7.7) 36.8 ⫾ 25.7 84 (17.0) 31.3 ⫾ 25.1
9 74 (6.9) 33.4 ⫾ 26.9 80.5 ⫾ 20.6 53.6 ⫾ 23.0 32.8 ⫾ 25.0 22.3 ⫾ 16.4 31 (5.4) 39.5 ⫾ 27.1 43 (8.7) 29.0 ⫾ 26.0
10 41 (3.8) 30.9 ⫾ 24.6 78.8 ⫾ 19.4 47.3 ⫾ 23.9 30.8 ⫾ 18.6 21.7 ⫾ 15.1 18 (3.1) 33.2 ⫾ 24.8 23 (4.7) 29.2 ⫾ 24.3
⬎10 45 (4.2) 30.9 ⫾ 23.7 82.7 ⫾ 20.6 43.2 ⫾ 23.9 28.4 ⫾ 17.1 23.8 ⫾ 15.1 19 (3.3) 38.4 ⫾ 23.7 26 (5.3) 25.1 ⫾ 22.0
Total 1,068 (100) 40.0 ⫾ 28.0 81.8 ⫾ 18.1 51.3 ⫾ 23.0 35.4 ⫾ 22.8 25.0 ⫾ 16.5 574 (100) 45.1 ⫾ 28.2
†
494 (100) 34.8 ⫾ 26.9
†
Note.—Data in parentheses are percentages.
* Data are mean ⫾ SD.
†
The mean difference was determined by using the two-tailed Wilcoxon rank sum test (P ⬍ .001).
TABLE 4
Individual Author Contributions
Contribution
Categories* No. of Authors
Author Position in Byline
American Authors
International
AuthorsFirst Second Last Middle
Guarantor of
integrity of
entire study 1,798/6,682 (26.9) 915/1,067 (85.8) 361/1,067 (33.8) 314/1,067 (29.4) 208/3,481 (6.0) 942/3,360 (28.0) 856/3,322 (25.8)
Study concepts 2,611/6,603 (39.5) 894/1,055 (84.7) 621/1,055 (58.9) 452/1,055 (42.8) 644/3,438 (18.7) 1,474/3,323 (44.4) 1,137/3,280 (34.7)
Study design 2,554/6,553 (39.0) 925/1,046 (88.4) 623/1,046 (59.6) 382/1,046 (36.5) 624/3,415 (18.3) 1,422/3,279 (43.4) 1,132/3,274 (34.6)
Literature
research 1,873/6,490 (28.9) 945/1,041 (90.8) 353/1,041 (33.9) 147/1,041 (14.1) 428/3,367 (12.7) 1,025/3,302 (31.0) 848/3,188 (26.6)
Clinical studies 2,974/5,472 (54.3) 642/858 (74.8) 513/858 (59.8) 347/858 (40.4) 1,472/2,898 (50.8) 1,521/2,561 (59.4) 1,453/2,911 (49.9)
Experimental
studies 828/1,758 (47.1) 220/272 (80.9) 150/272 (55.1) 91/272 (33.5) 367/942 (39.0) 478/922 (51.8) 350/836 (41.9)
Data
acquisition 3,156/6,499 (48.6) 869/1,039 (83.6) 613/1,039 (59.0) 306/1,039 (29.5) 1,368/3,382 (40.4) 1,795/4,272 (42.0) 1,361/3,227 (42.2)
Data analysis/
interpretation 2,615/6,440 (40.6) 923/1,030 (89.6) 575/1,030 (55.8) 269/1,030 (26.1) 848/3,350 (25.3) 1,404/3,227 (43.5) 1,211/3,213 (37.7)
Statistical
analysis 1,177/4,759 (24.7) 519/756 (68.7) 213/756 (28.2) 101/756 (13.4) 344/2,491 (13.8) 638/2,347 (27.2) 539/2,412 (22.3)
Manuscript
definition of
intellectual
content 2,603/6,447 (40.4) 875/1,032 (84.8) 597/1,032 (57.8) 463/1,032 (44.9) 668/3,351 (19.9) 1,419/3,228 (44.0) 1,184/3,219 (36.8)
Manuscript
preparation 2,115/6,653 (31.8) 1,026/1,063 (96.5) 427/1,063 (40.2) 229/1,063 (21.5) 433/3,464 (12.5) 1,203/3,341 (36.0) 912/3,312 (27.5)
Manuscript
editing 2,960/6,556 (45.1) 765/1,050 (72.9) 615/1,050 (58.6) 523/1,050 (49.8) 1,057/3,406 (31.0) 1,919/3,318 (57.8) 1,041/3,238 (32.1)
Manuscript
revision/
review 3,760/6,398 (58.8) 635/1,022 (62.1) 686/1,022 (67.1) 774/1,022 (75.7) 1,665/3,332 (50.0) 2,270/3,191 (71.1) 1,490/3,207 (46.5)
Note.—Data are number of authors who contributed to a specific category/number of authors in all articles who contributed to that specific category.
Numbers in parentheses are percentages.
* The category of “manuscript final version approval” was not included in the evaluation of individual contributions.
20
䡠
Radiology
䡠
January 2003 Hwang et al
R
adiology
ers. Similar to researchers making a single
contribution, researchers contributing to
two categories most commonly partici-
pated in “manuscript revision/review”
(506 of 1,183, 42.8%), “clinical studies”
(426 of 1,183, 36.0%), and “data acquisi-
tion” (396 of 1,183, 33.5%). Results of
detailed statistical analysis showed that
researchers contributing to one or two
categories were more frequently involved
in articles with a larger number of au-
thors in the byline (P ⬍ .001).
DISCUSSION
Fulfillment of Authorship Criteria
Authorship must bring with it respon-
sibility as well as reward, considering the
original Latin word auctor and the etymo-
logic term authority (13). Thus, author-
ship in biomedical journals establishes
accountability, responsibility, and credit
(4). In the medical research community,
however, misconduct relating to the as-
signment of authorship has been re-
ported to be increasing and most com-
monly appears as “honorary or guest”
authors who do not meet authorship cri-
teria (14,15). Concerning medical articles
with multiple authors published today,
several factors seem to be responsible for
author inflation, such as the increased
number of multicenter trials or the com-
plexity of research.
Other reasons that motivated authors
to add as many co-authors as possible
have been suggested: for example, hon-
orary or guest authorship due to the need
to please someone such as the head of the
group or the person who acquired fund-
ing for the study by including him or her,
even if that person did not substantially
contribute to the research. Another ex-
ample is swap authorship, defined as the
listing of another person’s name in one’s
article and vice versa (1,4,5).
Since the ICMJE authorship criteria
were established in 1985 and implemen-
tation has been encouraged in many
medical journals, numerous investigators
have studied the extent to which these
authorship criteria have been fulfilled
among published articles. Goodman (14)
found that about one third of 84 authors
did not meet authorship criteria in an
analysis of 12 articles. After surveying
184 research articles with multiple au-
thors from 10 medical journals, Shapiro
et al (15) reported that 268 (26%) of
1,014 authors had insufficient contribu-
tions to the research to merit authorship.
Hoen et al (16) studied 1 year of issues of
the Dutch Journal of Medicine and found
that 128 (36%) of 352 authors failed to
fulfill ICMJE authorship criteria. Yank
and Rennie (11) found that 346 (44%) of
785 authors in 6 months of issues of The
Lancet did not fulfill ICMJE authorship
criteria. In our analysis of research arti-
cles published in Radiology between 1998
and 2000, 2,172 (32.5%) of 6,686 re-
searchers in the bylines did not meet
ICMJE criteria. According to the policies
of both Radiology and the ICMJE, re-
searchers who do not have substantial
contribution to all three conditions for
authorship criteria should only be listed
in the acknowledgments (7,10). How-
ever, there were researchers who did not
meet the authorship criteria who ap-
peared in the bylines.
Similar to the findings of the previous
study of Yank and Rennie (11), our find-
ings showed that researcher participation
in the first condition of authorship crite-
ria—that is, contributions to conception
and design, acquisition of data, and anal-
ysis and interpretation of data—was
higher than that in the second condi-
tion—that is, contributions to drafting
the article or revising it critically for im-
portant intellectual content.
The “substantial” contribution stated
in the ICMJE criteria supports exclusion
of researchers with a limited role in all
three conditions, but simply checking
each of the three ICMJE conditions does
not reveal the type or degree of each au-
thor’s contribution. On the other hand,
if the 13 Radiology contribution catego-
ries are used in a meaningful and repro-
ducible way, these categories can reveal
more detailed information than can the
three ICMJE conditions.
The results of our analysis of these two
methods of disclosure of author contri-
butions demonstrated a misunderstand-
ing or unawareness of the authorship
criteria, or, if the criteria were well under-
stood, the presence of honorary or guest
authors in Radiology. A few indications of
a limited role of such authors are as fol-
lows: 1,133 authors checked a single cat-
egory, a clear indication of not fulfilling
the authorship criteria—that is, 348 au-
thors checked only “clinical studies,” 212
checked only “data acquisition,” and 170
checked only “manuscript revision/re-
view.” Such results were more common
among middle and international au-
thors.
Although the researcher(s) listed as
“guarantor of integrity of entire study”
should make a substantial contribution
to all three parts of the authorship crite-
ria, in our opinion, as well as be respon-
sible for the integrity of the entire study,
there were 26 authors who contributed
solely to the category of “guarantor of
integrity of entire study,” and 25 of them
were international authors. For those au-
thors who checked two categories, the
results were similar to those of authors
TABLE 5
Authors Contributing to One or Two Categories
Article Characteristics
Contribution to
One Category
Contribution to
Two Categories Total P Value*
Author position in byline ⬍.001
First (n ⫽ 1,068) 1 (0.1) 5 (0.5) 6 (0.6)
Second (n ⫽ 1,068) 30 (2.8) 94 (8.8) 124 (11.6)
Last (n ⫽ 1,068) 180 (16.9) 201 (18.8) 381 (35.7)
Middle (n ⫽ 3,482) 922 (26.5) 883 (25.4) 1,805 (51.8)
Author nationality ⬍.001
American (n ⫽ 3,364) 357 (10.6) 485 (14.4) 842 (25.0)
International (n ⫽ 3,322) 776 (23.4) 698 (21.0) 1,474 (44.4)
Year of publication .02
1998 (n ⫽ 2,152) 391 (18.2) 405 (18.8) 796 (37.0)
1999 (n ⫽ 2,176) 374 (17.2) 350 (16.1) 724 (33.3)
2000 (n ⫽ 2,358) 368 (15.6) 428 (18.2) 796 (33.8)
No. of authors per article ⬍.001
3(n ⫽ 312) 4 (1.3) 19 (6.1) 23 (7.4)
4(n ⫽ 532) 26 (4.9) 74 (13.9) 100 (18.8)
5(n ⫽ 925) 79 (8.5) 123 (13.3) 202 (21.8)
6(n ⫽ 1,266) 166 (13.1) 235 (18.6) 401 (31.7)
7(n ⫽ 1,029) 180 (17.5) 213 (20.7) 393 (38.2)
8(n ⫽ 1,024) 238 (23.2) 224 (21.9) 462 (45.1)
9(n ⫽ 684) 201 (29.4) 116 (17.0) 317 (46.3)
10 (n ⫽ 400) 115 (28.8) 83 (20.8) 198 (49.5)
⬎10 (n ⫽ 514) 124 (24.1) 96 (18.7) 220 (42.8)
Total (n ⫽ 6,686) 1,133 (16.9) 1,183 (17.7) 2,316 (34.6)
Note.—Numbers in parentheses are percentages.
* P values obtained by performing the
2
test.
Volume 226
䡠
Number 1 Researcher Contributions and Fulfillment of Authorship Criteria
䡠
21
R
adiology
who checked a single category. Under
our presumption of the third condition
of authorship criteria being fulfilled, each
author has to participate in more than
two categories to meet both the first and
second conditions of the ICMJE. Authors
contributing to one category do not meet
both criteria. Also, authors contributing
to two categories may not fulfill both cri-
teria, unless each category comes from
the first and second conditions of the
ICMJE. All of the authors contributing to
one category (1,133 authors) and 656
(56%) of 1,183 authors contributing to
two categories did not meet the author-
ship criteria.
Including 26 authors with single con-
tributions to the category of “guarantor
of integrity of entire study,” the authors
who contributed to one or two categories
represented 82% (1,789 of 2,172) of au-
thors who did not fulfill the ICMJE crite-
ria. Consistent with the results of Shapiro
et al (15), these authors were principally
responsible for a decreased fulfillment of
authorship. These results show that there
is an evident need for clarification and
continual education of the terms of the
establishment of the authorship criteria,
especially for non-American researchers.
Although there may not be wide agree-
ment on or recognition of ICMJE author-
ship criteria, we believe that fulfilling the
ICMJE criteria as the standard for author-
ship is essential for justified authorship.
In our analysis, we examined the asso-
ciation between the percentage of re-
searchers who fulfilled the authorship
criteria and the number of researchers
per article, as well as the researchers’ po-
sitions in the byline. These percentages
decreased significantly as the number of
researchers per article increased, from
89% in an article with three authors to
61% in an article with more than 10 au-
thors.
The mean contribution percentages
showed the same trend, which demon-
strates that the principal problem for au-
thorship seems to be closely related to
the number of researchers per article.
This increased number of researchers per
article also shows a decreased individual
contribution to the research. In our study,
there were several differences in author-
ship trends between American and inter-
national researchers. Articles with interna-
tional researchers had a lower fulfillment
of ICMJE authorship criteria, a higher
mean number of researchers per article, a
lower contribution percentage per re-
searcher, and a higher number of research-
ers contributing to a single category. It is
possible that factors such as sociocultural
differences, different reward, and tenure
systems might determine the geographic
differences of authorship trends.
Although Radiology has made an effort
to reduce inappropriate authorship, the
results of our study showed no substan-
tial improvement of the fulfillment of
authorship criteria from 1998 to 2000,
thus demonstrating that researcher aware-
ness or interest in authorship criteria has
not increased during this 3-year period.
Individual Author Contributions
Overall, the percentages of researcher
contributions varied in each category.
The categories most frequently partici-
pated in, including “manuscript revision/
review,”“clinical studies,” and “data ac-
quisition,” had a 50%– 60% contribution
rate. According to researcher position in
the byline, as in previous studies (15,17),
the contribution of the first author was
noteworthy, with percentages exceeding
60% contribution in all categories. Con-
tributions of nonfirst authors varied
greatly, however, with a relatively larger
contribution of the second authors com-
pared with that of the last or middle au-
thors, and there were markedly variable
types and numbers of categories fre-
quently participated in. With regard to
the geographic differences among re-
searcher contributions, there was a ten-
dency toward lower participation in arti-
cles by international researchers than
that in articles by American researchers in
all the contribution categories. This ten-
dency toward lower participation was
correlated closely with a greater number
of researchers per article, a lower contri-
bution percentage per researcher, and a
larger proportion of researchers contrib-
uting to a single category.
To quantify individual author contri-
butions, in an era of no available infor-
mation about each author’s contribu-
tions, such as self-documented or checked
categories of author contributions, Mou-
lopoulos et al (18) considered a simple
contribution percentage based on the to-
tal number of authors, computed by as-
suming that the contributions of differ-
ent authors in an article were all equal.
To compare the curricula vitae of four
candidates for a professional appoint-
ment, they further devised a different
measure of contribution by giving more
weight to the first or higher-positioned
authors in the byline. However, when
the contribution of the last authors is
greater than that of the middle authors,
as was shown in our analysis, this mea-
sure is not valid. In our study, author
contributions could be quantified and
compared simply by calculating the con-
tribution percentage. Although contribu-
tion percentage as an index for author-
ship is simple and easy to apply, it may
be difficult to generalize because it re-
quires some premises, such as the pres-
ence of specific categorization of author
contributions and a presumption of giv-
ing equal value to all contribution cate-
gories.
The order of authorship in the byline is
a joint decision of the authors, who
should be prepared to explain the order
in which they are listed (7). If the mean-
ing of the author order can be explained,
the author order in the byline may be a
way to reduce inappropriate authorship,
in addition to listing each author’s spe-
cific contribution. However, findings in
previous studies of researcher contribu-
tions, along with our study results, dem-
onstrated the variable extent and nature
of researcher contributions, despite their
positions in the byline (1,11,15,17).
There have been some preconceptions
about author order in the byline. In a
typical example, many authors reserve
the position of last author for an honor-
ary person, such as a senior staff member
or department director, but this may en-
courage honorary authorship (3). If a
brief sentence is attached at the begin-
ning or end of an article or within the
author contribution list regarding how
author order in the byline was deter-
mined, we believe it will provide useful
information to journal readers. For in-
stance, the sentence “Except for the first
author, author order in the byline was
arranged alphabetically by last name” in-
dicates that author order in the byline
was not meaningful. To the contrary, the
sentence “The order of author names was
based on the number of contributions
made by each” indicates that the author
order in the byline was meaningful.
In summary, our analysis of the infor-
mation in the author contribution list
demonstrated some important findings.
First, in our study, only 68% of the re-
searchers listed in the byline fulfilled the
ICMJE authorship criteria, and this low
percentage of fulfillment was closely re-
lated to the number of researchers per
article, especially to the number of re-
searchers who contributed to a single cat-
egory. Second, the order of researcher
names in the byline had limited refer-
ence to the extent and nature of contri-
butions made to the research. Third, geo-
graphic differences exist in the nature
and extent of researcher contributions
22
䡠
Radiology
䡠
January 2003 Hwang et al
R
adiology
and in the fulfillment of authorship. The
fulfillment of authorship of American re-
searchers seems to be greater than that of
authors from other countries and is
partly related to the smaller number of
researchers contributing to a single cate-
gory in American articles. As suggested
by Athanasoulis (5) and Probyn et al (19),
our results are one piece of evidence sug-
gesting the necessity for continual edu-
cation on valid authorship by various
means, especially for international re-
searchers. In addition, efforts to make the
ICMJE criteria more widely known may
reduce confusion regarding authorship
for those researchers who are unfamiliar
with the ICMJE criteria (16). Our findings
should be interpreted with the under-
standing that our results were derived
only from the journal Radiology, and
thus, one should be cautious not to gen-
eralize them to other journals.
Acknowledgments: We thank Bonnie Hami,
MA, Department of Radiology, University Hos-
pitals of Cleveland, for her editorial assistance
in the preparation of this manuscript. We also
thank Jung Soon Lee, MA, and Hyo Jin Son,
MA, Department of Radiology, St Vincent’s
Hospital, for their assistance in the data entry.
References
1. Rennie D, Yank V, Emanuel L. When au-
thorship fails: a proposal to make contrib-
utors accountable. JAMA 1997; 278:579–
585.
2. Rennie D, Flanagin A. Authorship! author-
ship! ghost, guest, grafters, and the two-
sided coin. JAMA 1994; 271:469 –471.
3. Riesenberg D, Lundberg G. The order of
authorship: who’sonfirst? (editorial).
JAMA 1990; 264:1857.
4. Flanagin A, Carey LA, Fontanarosa PB, et
al. Prevalence of articles with honorary
authors and ghost authors in peer-re-
viewed medical journals. JAMA 1998;
280:222–224.
5. Athanasoulis CA. Authors need to be ed-
ucated on authorship principles. Radiol-
ogy 2000; 217:598–599.
6. Davidoff F. News from the International
Committee of Medical Journal Editors.
Ann Intern Med 2000; 133:229–231.
7. International Committee of Medical Jour-
nal Editors. Uniform requirements for
manuscripts submitted to biomedical
journals. Updated May 2000. Available at:
www.icmje.org. Accessed June 6, 2000.
8. Godlee F. Definition of authorship may
be changed. BMJ 1996; 312:1501–1502.
9. Proto AV. Radiology—1998 and the fu-
ture. Radiology 1998; 206:1–2.
10. Publication information for authors. Ra-
diology 2000; 217:41A–50A.
11. Yank V, Rennie D. Disclosure of re-
searcher contributions: a study of original
research articles in The Lancet. Ann In-
tern Med 1999; 130:661–670.
12. Altman D. Practical statistics for medical
research. London, England: Chapman &
Hall, 1991; 205–212.
13. Sheikh A. Publication ethics and the re-
search assessment exercise: reflections on
the troubled question of authorship.
J Med Ethics 2000; 26:422–426.
14. Goodman NW. Survey of fulfillment of
criteria for authorship in published med-
ical research. BMJ 1994; 309:1482.
15. Shapiro DW, Wenger NS, Shapiro MF.
The contributions of authors to multiau-
thored biomedical research papers. JAMA
1994; 271:438–442.
16. Hoen WP, Walvoort HC, Overbeke AJ.
What are the factors determining author-
ship and the order of the authors’ names?
a study among authors of the Nederlands
Tijdschrift voor Geneeskunde (Dutch
Journal of Medicine). JAMA 1998; 280:
217–218.
17. Slone RM. Coauthors’ contributions to
major papers published in the AJR: fre-
quency of undeserved coauthorship. AJR
Am J Roentgenol 1996; 167:571–579.
18. Moulopoulos SD, Sideris DA, Georgilis
KA. For debate . . . individual contribu-
tions to multiauthor papers. BMJ 1983;
287:1608–1610.
19. Probyn LJ, Asch MR, Proto AV. The effect
of changes in guidelines for authorship
on current radiology publications. Radi-
ology 2000; 215:615–616.
Volume 226
䡠
Number 1 Researcher Contributions and Fulfillment of Authorship Criteria
䡠
23
R
adiology