You are on page 1of 6

G Model

BONSOI-3509; No. of Pages 6

ARTICLE IN PRESS
Joint Bone Spine xxx (2011) xxxxxx

Original article

Misleading abstract conclusions in randomized controlled trials in rheumatology: Comparison of the abstract conclusions and the results section
Sylvain Mathieu a,b,c,d , Bruno Giraudeau a,e,f,g , Martin Soubrier d , Philippe Ravaud a,b,c,
a

Inserm U738, Paris, France Centre dpidmiologie clinique, hpital htel-Dieu, Assistance publiqueHpitaux de Paris, Paris, France c Facult de mdecine, universit Paris Descartes, Paris, France d Service de rhumatologie, hpital Gabriel-Montpied, universit Clermont-Ferrand 1, Clermont-Ferrand, France e Inserm, CIC 202, Paris, France f CHRU de Tours, Tours, France g Universit Francois-Rabelais Tours, Tours, France
b

a r t i c l e

i n f o

a b s t r a c t
Introduction: Readers of scientic articles often read only the abstract and its conclusions because of lack of time or of access to the full-length articles. Objectives: To assess the prevalence of misleading conclusions in abstracts of randomized controlled trials (RCTs) in rheumatology, determine whether trials are registered and whether abstract conclusions are based on the primary outcome (PO), and identify the predictors of misleading abstract conclusions. Methods: We searched Medline, Embase and the Cochrane Collaboration for reports of RCTs assessing rheumatoid arthritis, osteoarthritis or spondylarthropathies published between January 2006 and April 2008. Abstract conclusions were misleading if: the PO was not reported in the conclusion; the conclusions were based on only a secondary outcome or subgroup results; the results and conclusions were in disagreement; negative results were suggested as equivalent, or if there was no discussion of benets and risks in cases of serious adverse events. Results: Of the 144 reports selected, we focused on the 105 articles containing a clear PO. Twenty-four reports (23%) contained misleading conclusions. Lack of PO reporting (n = 10) and conclusions disagreeing with article results (n = 7) were the most frequent reasons. Nineteen out of 144 (13.2%) declared study registration with clear and similar registered and published POs and no misleading abstract conclusions. Reports of negative results showed a higher frequency of misleading conclusions as did those assessing osteoarthritis. On multivariable analysis, only negative results predicted misleading abstract conclusions (OR = 9.58 [3.2028.70]). Conclusions: Almost one-quarter of these RCT in rheumatology had misleading conclusions in the abstract, especially those with negative results. 2011 Socit francaise de rhumatologie. Published by Elsevier Masson SAS. All rights reserved.

Article history: Accepted 9 May 2011 Available online xxx Keywords: Misleading conclusions Randomized controlled trials Reporting Transparency

1. Introduction Randomized controlled trials (RCTs) are generally considered as the best guarantee of the efcacy of health care interventions, and the number of published reports of RCTs is on the increase [1]. Biased results from poorly designed or inadequately reported trials can lead to awed decision making in individual and public health care treatment [2]. Providing clear, unambiguous results and decreasing bias in RCTs have become prime objectives for medical authors and journals [35]. The use of reporting guidelines such as the CONsolidated Standards of Reporting Trials (Consort).

Corresponding author. Tel.: +33140256773; fax: +33140256773. E-mail address: philippe.ravaud@htd.aphp.fr (P. Ravaud).

Statements and extensions has improved the quality of reporting [611] in the full text of articles. Since the present study was begun, other recommendations have been drawn up for the writing of abstracts [12]. However, despite these guidelines, the quality of reporting is still uneven, and evidence of bias persists [1317]. Le Henanff et al. assessed the quality of non-inferiority and equivalence randomized trials and found misleading conclusions in four of the 33 articles assessed (12.1%) that fullled guideline requirements specic to these trials, mainly because of discrepancy between the declared aims and the results obtained [18]. Medical practitioners have limited time to read articles and as a result, as Gotzche remarked, Most users of the scientic literature read vastly more conclusions than they read abstracts, and vastly more abstracts than full papers. Conclusions are the worst part of papers,. . . [19]. The problem is compounded by the fact that

1297-319X/$ see front matter 2011 Socit francaise de rhumatologie. Published by Elsevier Masson SAS. All rights reserved. doi:10.1016/j.jbspin.2011.05.008

Please cite this article in press as: Mathieu S, et al. Misleading abstract conclusions in randomized controlled trials in rheumatology: Comparison of the abstract conclusions and the results section. Joint Bone Spine (2011), doi:10.1016/j.jbspin.2011.05.008

G Model
BONSOI-3509; No. of Pages 6 2

ARTICLE IN PRESS
S. Mathieu et al. / Joint Bone Spine xxx (2011) xxxxxx

conclusions presented in an abstract can be inuenced by various interests that distort or modify the results of a clinical trial [1922]. Our aim was to: 1) assess the prevalence of misleading conclusions in abstracts of reports of RCTs; 2) determine whether abstract conclusions are based on the primary outcome (PO) stated in the article; 3) identify the factors that might predict misleading conclusions. 2. Methods We analyzed RCTs of rheumatology because the pharmacological and non-pharmacological treatments of this disorder vary widely. 2.1. Search strategy One of us (SM) searched Medline, Embase and the Cochrane Central Register of Controlled Trials to identify reports of two parallel-group, superiority RCTs in rheumatoid arthritis (RA), spondylathropathies and osteoarthritis (OA), published between January 1, 2006 and March 31, 2008 in general and specialty medical journals. We used the search terms rheumatoid arthritis or spondylarthropathies or osteoarthritis and randomized controlled trials. Our search concerned only articles published in English. 2.2. Exclusion criteria Potentially relevant articles were selected by one of us (SM) who, after reading the title, keywords and abstract obtained the full text of the article. The exclusion criteria were: 1) commentary or discussion papers; 2) observational studies (cohort studies, case-control studies) or meta-analyses; 3) a particular RCT design (e.g., cluster randomized trials or N-of-1 trials); 4) study design with more than two parallel groups; 5) a non-inferiority or equivalence objective; 6) only the abstract being available. If several papers referred to the same trial, only the main publication was retained. We excluded any subsequent papers as duplicate publications of the initial trial. 2.3. Data extraction One reviewer (SM), who was not blinded to journal and author, extracted all data using a standardized data abstraction form based on preliminary discussion. 2.3.1. Extraction of journal characteristics We found the impact factor and whether the journal involved peer-review from the journals website [23] and whether the journal was a general medical or internal medicine journal from the Institute for Scientic Information (ISI) Web of Knowledge 2007 data. 2.3.2. Information about authors and sponsors We recorded the nature of the funding source (no funding, government, industry, private non-prot, multiple sources, other or not reported) and the nancial ties of each author (no nancial

ties, nancial ties with the sponsor of the study disclosed, nancial ties with any other company disclosed or not reported) [23]. Information about the role of the sponsor was recorded as: 1) 2) 3) 4) the role of sponsor not mentioned; the sponsor not involved in the study design and analyses; the sponsor involved in the study design and statistical analysis; no sponsor.

2.3.3. Trial registration For each published trial report, one of us (SM) systematically searched whether the trial was registered. We checked whether the authors reported registering their trial and whether the registration number was included in the published article. When no registration number was reported in the published article, we searched the following clinical trial registries: ClinicalTrials.gov (NCT) and International Standard Randomized Controlled Trial Number Register (ISRCTN). To identify the record of these trials in the registries, we proceeded as follows. 1. We searched for references to the published article in the registry; 2. If the registry contained no reference to the published article, we checked whether the principal investigator and the funding source matched those reported in the article. When these tallied, the report was included and data were extracted; 3. If multiple registry records were found when the article title was used as the search string, we checked the experimental treatment, the comparator, the design, and the sample size to select the relevant report. If no registration number was found after these searches, the trial for the published article was considered not registered. 2.3.4. Extraction of the primary outcome If the PO was clearly identied in the article, it was recorded as such. If no PO was given, we recorded the outcome used for sample size calculation. If none of this information was available, we recorded the PO registered in a clinical trial registry (obtained as described above). For each selected report, we compared, if possible, the registered PO and that reported in the published text. For articles with differences between the registered and published PO, we used the PO stated in the article. If no PO was found in the article or in the registry, the study was excluded from analysis. The PO was considered unclear if it could not be determined after these three steps and the study was left out. 2.3.5. Denition of a misleading abstract conclusion Each report was assessed by (SM) to determine whether the abstract contained a misleading conclusion. This analysis included the three stages given in Table 1: 1) assessment of the results section of the article [24]; 2) assessment of the abstract conclusions; 3) determining the existence of a misleading conclusion [23,2527] Examples of studies classied as having misleading abstract conclusions are given in Appendix S1 (see the supplementary material associated with this article online). All abstract conclusions identied as being misleading were conrmed by two other authors (PR and BG), with disagreements resolved by consensus.

Please cite this article in press as: Mathieu S, et al. Misleading abstract conclusions in randomized controlled trials in rheumatology: Comparison of the abstract conclusions and the results section. Joint Bone Spine (2011), doi:10.1016/j.jbspin.2011.05.008

G Model
BONSOI-3509; No. of Pages 6

ARTICLE IN PRESS
S. Mathieu et al. / Joint Bone Spine xxx (2011) xxxxxx 3 Denition of misleading abstract conclusions We considered 1) results associated with the primary outcome stated in the abstract conclusions and 2) a cross-classication of the studies according to the classication of the Results section and abstract conclusions to dene abstract conclusions as misleading by the following characteristics. 1) Results for the primary outcome not reported 2) Conclusions based on only secondary outcome results 3) Conclusions based on only sub-group analysis 4) Conclusions not in agreement with the results of the study (i.e. difference between the classication of the Results section and the abstract conclusions) 5) Conclusions of equivalence in superiority trials with negative results or equivalence suggested by authors with terms such as similar, as effective as, comparable [23,2527] 6) No discussion of the benetrisk balance of the intervention when serious side effects were found in the text

Table 1 Assessment of results section of the full text of articles and abstract conclusions and denition of misleading abstract conclusions. Classication of Results section and abstract conclusions 1) Results statistically in favour of the study intervention 2) Results statistically not in favour of the study intervention (results statistically signicant for the comparator being more efcacious) 3) Results statistically not in favour of or against the study intervention (result did not reach statistical signicance) 4) Unclear results (i.e. results for the primary outcome were insufciently described to determine whether these results were one of the three previous items) 5) Results for the primary outcome not reported

For each study, we used the following classication scheme (assigning the relevant number) to classify: 1) the results of the primary outcome according to the data presented in the Results section of the full text and; 2) the abstract conclusions. This analysis was then used to classify abstract conclusions as misleading. To minimize the bias of a reading by one author, SM rst evaluated all abstract conclusions one by one and then all results sections. Serious adverse events reported in the Results section were dened according to the National Institutes of Health [24].

2.4. Statistical analysis Data are reported as median and interquartile range (IQR) or mean and standard deviation (SD) for quantitative variables and number of articles (%) for categorical variables for the different characteristics of each article. Impact factor and sample size were divided into quartiles, with the upper three quartiles compared with the lowest quartile for impact factor and the inverse comparison for sample size. Differences between groups were tested by Chi2 test for categorical data, or Students unpaired t test for numerical data. A P < 0.05 (two-sided) was considered statistically signicant. Associations between the misleading abstract conclusions (dened as a binary outcome) and manuscript characteristics were studied by logistic regression models. Univariate analysis was performed, followed by multivariate analysis, which included any covariate with P < 0.20 on univariate analysis. Although impact factor and sample size were continuous variables, they were modeled categorically, on the basis of four classes previously dened. SAS v9.1 (SAS Inst., Cary, NC) and R v2.6.0) were used for statistical analyses. 3. Results 3.1. Characteristics of included articles Our nal sample consisted of 144 published reports of RCTs (Fig. 1). The characteristics of the full sample are shown in Table 2. Most studies were published in peer-reviewed (n = 109; 75.7%) and specialty journals (n = 122; 84.7%). Sixty-seven articles (46.5%) concerned patients with OA, 63 (43.8%) RA (two assessing both RA and OA) and 16 (11.1%) spondyloarthropathies. Of the 144 trials selected, 69 (47.9%) disclosed funding from industry sponsors, with 32 of them (46.4%) reporting that the sponsor was involved in the study design (nine), in statistical analysis (one) or in both (22). Ten reports (14.5%) stated that the sponsor was not involved in the design or analysis, and 27 (39.1%) gave no information about the role of the sponsor. In 26% of reports (37), the authors declared nancial ties with the sponsor of the study. However, for many reports, this information was not provided (41.7%). The PO was unclear in 39 articles (27.1%) (Fig. 2) and was more often unclear in articles published in journals that were not peer-reviewed than in those published in peer-reviewed journals (28.2% vs 7.6%; P = 0.002). Non-peer-reviewed journals also had a lower impact factor than peer-reviewed journals (1.35 [IQR: 1.073.00] vs 4.02 (2.254.75]; P < 0.001) (Table 2).

In 40 of the 144 articles, the studies had been registered (27.8%). For 16, the trial registration number appeared in the text (40%), and for 37, the PO was clear. The PO was the same in the text and in the registry for 25 of the 40 articles (25; 62.5%). 3.2. Analysis of articles with a clear primary outcome In all, 105 articles reported a clear PO in the text or in the registry (72.9%) (Fig. 2). The most frequent characteristics of the PO were patient-reported outcomes (n = 56; 53.3%) and clinical measures (n = 40; 38.1%). In 61 (58.1%) of these 105 articles, the results were positive, in 42 (40.0%) negative and in two (1.9%) inaccurate. In 24 (24/105; 22.9%) articles the abstract conclusions were misleading and in seven, there was disagreement between the abstract conclusions and the study results (Table 3). The other reasons for misleading abstract conclusions are in Table 3. In four of 24 articles with misleading conclusions, the efcacy of the study treatment was based on intra-group comparisons, and yet inter-group comparisons revealed no statistically signicant difference between the study treatment and the control. For articles with negative or inaccurate results, the proportion of misleading abstract conclusions was greater than that in articles with positive results (19/42 [45.2%] vs 3/61 [4.9%]; P < 0.001).

1571 articles published in MEDLINE, EMBASE and Cochrane Central Register of Controlled Trials between January 1, 2006 and March 31, 2008

1343 articles excluded on the basis of the title and abstract 428 reports not assessing rheumatoid arthritis (RA), spondylarthropathies (SPA) or osteoarthritis (OA) 538 duplicate reports 277 nonrandomized controlled trials 95 not a two-arm parallel randomized controlled trial (RCT) 5 noninferiority or equivalence trials

228 articles selected

84 articles excluded after reading the full text 23 nonrandomized controlled trials 26 follow-up of RCTs described in other reports 6 reports not assessing RA, SPA or OA 10 articles without a full text 14 articles not in English 5 noninferiority or equivalence trials

144 articles assessed

Fig. 1. Flowchart of the selection of articles in the study.

Please cite this article in press as: Mathieu S, et al. Misleading abstract conclusions in randomized controlled trials in rheumatology: Comparison of the abstract conclusions and the results section. Joint Bone Spine (2011), doi:10.1016/j.jbspin.2011.05.008

G Model
BONSOI-3509; No. of Pages 6 4 Table 2 Characteristics of the selected articles.

ARTICLE IN PRESS
S. Mathieu et al. / Joint Bone Spine xxx (2011) xxxxxx All articles (n = 144) Articles with a clear PO n = 105 (72.9%) Articles with an unclear PO n = 39 (27.1%)

Journal characteristics Year of publication 2006 2007 2008 Type of journal General Speciality Peer-review status Yes No Unknown Impact factor <1 14 510 > 10 Authors characteristics Funding source Single Multiple Not reported No funding Industry University/hospital Government Private non-prot Other (association) Financial ties None With the sponsor of the study With another industrial company Not reported Role of the sponsor None In study design In statistical analysis In study design and statistical analysis Not reported Sample size Quartile 1 Quartile 2 Quartile 3 Quartile 4 Disease assessed Rheumatoid arthritis Osteoarthritis Spondyloarthropathies Registered article Yes Same registered and written primary outcomes Methodological characteristics Description of the sample size calculation Yes Real and calculated sample size similar Nature of primary outcome Patient-reported outcome Clinical outcome Radiological outcome Biological outcome Safety outcome Other Serious adverse events reported No Yes Unknown Data are number (%); PO: primary outcome. a Two studies assessed both rheumatoid arthritis and osteoarthritis.

62 70 12 22 122 109 19 16 17 82 41 4

48 (77.4) 48 (68.6) 9 (75.0) 16 (72.7) 89 (73.0) 87 (79.8) 8 (42.1) 10 (62.5) 8 (47.1) 57 (69.5) 36 (87.8) 4 (100.0)

14 (22.6) 22 (31.4) 3 (25.0) 6 (27.3) 33 (37.0) 22 (20.2) 11 (57.9) 6 (37.5) 9 (52.9) 25 (30.5) 5 (12.2) 0 (0.0)

84 23 33 4 69 22 22 5 14 45 37 2 60 41 9 1 22 71 43.8 88.5 221.3 7111.0 63 67a 16 40 25

67 (79.8) 19 (82.6) 16 (48.5) 3 (75.0) 59 (85.5) 16 (72.7) 19 (86.4) 2 (40.0) 10 (71.4) 35 (77.8) 34 (91.9) 2 (100.0) 34 (56.7) 33 (80.5) 7 (77.8) 1 (100.0) 21 (95.5) 43 (60.6) 58.0 121.0 247.0 7111.0 47 (74.6) 47 (70.1)a 13 (81.3) 37 (92.5) 24 (96.0)

17 (20.2) 4 (17.4) 17 (51.5) 1 (25.0) 10 (14.5) 6 (27.3) 3 (13.6) 3 (60.0) 4 (28.6) 10 (22.2) 3 (8.1) 0 (0.0) 26 (43.3) 8 (19.5) 2 (22.2) 0 (0.0) 1 (4.5) 28 (39.4) 31.0 55.0 83.5 652.0 16 (25.4) 20 (29.9) 3 (18.7) 3 (7.5) 1 (4.0)

75 54 73 52 12 6 3 2 82 35 27

69 (92.0) 51 (94.4) 56 (76.7) 41 (78.8) 10 (83.3) 5 (83.3) 2 (66.4) 1 (50.0) 58 (70.7) 34 (97.1) 13 (48.1)

6 (8.0) 3 (5.6) 17 (23.3) 11 (21.2) 2 (16.7) 1 (16.7) 1 (33.3) 1 (50.0) 24 (29.3) 1 (2.9) 14 (51.0)

Please cite this article in press as: Mathieu S, et al. Misleading abstract conclusions in randomized controlled trials in rheumatology: Comparison of the abstract conclusions and the results section. Joint Bone Spine (2011), doi:10.1016/j.jbspin.2011.05.008

G Model
BONSOI-3509; No. of Pages 6

ARTICLE IN PRESS
S. Mathieu et al. / Joint Bone Spine xxx (2011) xxxxxx 5 All (n = 105) Articles with positive results (n = 61) 3 (4.9) Articles with negative results (n = 42) 19 (45.2) 7 (36.8) 3 (15.8) 1 (50.0) 5 (26.3) 6 (31.6) 1 (33.3) 1 (50.0) Articles with inaccurate results (n = 2) 2 (100.0)

Table 3 Frequency of misleading conclusions in articles with a clear primary outcome.

Misleading conclusion Yes Reasons for misleading conclusions Primary outcome not reported Conclusions based on only secondary outcome results Conclusions based on only sub-group analysis Negative results suggested as equivalent Conclusions not in agreement with results Serious side effects without discussion of benetrisk balance of the intervention

24 (22.9) * 10 (41.7) 3 (12.5) 1 (4.2) 5 (20.8) 7 (29.2) 1 (4.2)

3 (100.0)

Data are number (%). *: three conclusions had two reasons for misleading conclusions; : two conclusions had two reasons for misleading conclusions; : one conclusion had two reasons for misleading conclusions; : compared with positive results: P < 0.001.

Misleading abstract conclusions in articles with negative results were mainly due to lack of PO reporting (n = 7/19) (Table 3). On univariable analysis, reports with negative results were signicantly more likely to contain misleading abstract conclusions. Articles with misleading abstract conclusions more often assessed OA than RA and spondyloarthropathies (35.6% vs 17.0 and 0%, respectively; P < 0.012) (Appendix S2). On multivariate analysis, only negative trial results were associated with misleading abstract conclusions (OR = 9.58 [3.20 28.70]).

4. Discussion A clear description of the PO was given in 105 of 144 abstracts of reports of superiority RCTs in rheumatology. In 24 (23%) of the 105, the conclusions in the abstract were misleading. Only negative trial results were associated with misleading abstract conclusions. Of the 144 articles, only 19 (13.2%) declared study registration in the text, had the same PO in both the published text and the registry, and did not contain misleading abstract conclusions. Articles reporting negative results are more likely to contain misleading abstract conclusions. For instance, we found ve articles in which negative results were suggested as being equivalent. This is because the chance of an article being published can depend on whether the results are positive or negative [28,29]. Chan et al. reported that articles with statistically signicant efcacy outcomes had more than 2-fold greater odds of being fully reported than those with non-signicant efcacy outcomes [13]. Consequently, authors may be less likely to submit manuscripts reporting negative results [30]. Only 34 articles reported serious adverse events, and of these three did not discuss the benetrisk balance of the intervention in their abstract conclusions. However, authors should weigh the advantages and disadvantages of an intervention in abstract conclusions, as recommended in the Consort statement for abstracts [19]. The existence of adverse effects can have a major impact on whether a particular intervention is deemed acceptable and useful, and readers need both efcacy and safety results to make rational and balanced decisions. It is now mandatory for a trial to be registered before publication [31], but only 40 articles (27.8%) in our sample did so. Of these, 16 (40.0%) declared the information in the text and 15 (37.5%) reported a PO different from the registered one, which agrees with other recent ndings [32]. Although trial registration aims to improve the reporting of RCTs results [3,33,34], our study shows that registration is still inadequate and does not prevent authors from changing the PO when reporting the trial results. Reviewers should, therefore, systematically check whether the features of the manuscript they are reviewing agree with those of the registered study. Our study has several limitations. First, the Consort statements extension for abstracts refers to overgeneralization in misleading conclusions, an element we did not deal with in our denition because we considered it too subjective to be assessed. Thus, our ndings probably underestimate the proportion of misleading abstract conclusions. Second, safety issues were considered only as discussed or not, with no attention given to the quality of the reporting of safety data, which is often insufcient or inadequate [21,35,36]. Third, we focused on RCTs in one medical eld (rheumatology). We restricted our selection of articles because

144 articles selected

39 articles with an unclear or no primary outcome in the text or in the registered protocol

105 articles with a clear primary outcome 68 unregistered trials

12 misleading abstract conclusions* 6 conclusions do not match results 4 primary outcome not reported 1 conclusions based on sub-group analysis 2 negative results suggested as equivalent 1 based on secondary outcome results *2 articles had multip le reasons for misleading conclusions

37 registered studies with a clear primary outcome in article

12 articles with different registered and published primary outcome or unable to be defined

6 misleading abstract conclusions 3 primary outcome not reported 2 negative results suggested as equivalent 1 based on secondary outcome results 1 serious adverse events not reported 1 article had multi ple reasons for misleading conclusions

25 registered studies with a clear primary outcome similar in the protocol and the article

6 misleading abstract conclusions 1 conclusions do not match results 3 primary outcome not reported 1 based on secondary outcome results 1 negative results suggested as equivalent

19 registered studies with a clear primary outcome similar in the protocol and in the article and no misleading abstract conclusions

Fig. 2. Summary of articles selected, statement of primary outcome in the text and in the trial registration and misleading abstract conclusions.

Please cite this article in press as: Mathieu S, et al. Misleading abstract conclusions in randomized controlled trials in rheumatology: Comparison of the abstract conclusions and the results section. Joint Bone Spine (2011), doi:10.1016/j.jbspin.2011.05.008

G Model
BONSOI-3509; No. of Pages 6 6

ARTICLE IN PRESS
S. Mathieu et al. / Joint Bone Spine xxx (2011) xxxxxx [11] Plint AC, Moher D, Morrison A, et al. Does the CONSORT checklist improve the quality of reports of randomised controlled trials? A systematic review. Med J Aust 2006;185:2637. [12] Hopewell S, Clarke M, Moher D, et al. CONSORT for reporting randomized controlled trials in journal and conference abstracts: explanation and elaboration. PLoS Med 2008;5:e20. [13] Chan AW, Krleza-Jeric K, Schmid I, et al. Outcome reporting bias in randomized trials funded by the Canadian Institutes of Health Research. CMAJ 2004;171:73540. [14] Chan AW, Hrobjartsson A, Haahr MT, et al. Empirical evidence for selective reporting of outcomes in randomized trials: comparison of protocols to published articles. JAMA 2004;291:245765. [15] Al-Marzouki S, Roberts I, Evans S, et al. Selective reporting in clinical trials: analysis of trial protocols accepted by The Lancet. Lancet 2008;372:201. [16] Dwan K, Altman DG, Arnaiz JA, et al. Systematic review of the empirical evidence of study publication bias and outcome reporting bias. PLoS One 2008;3:e3081. [17] Higgins JPT, Altman DG. Assessing risk of bias in included studies. In: Higgins JPT, Green S, editors. Cochrane handbook for systematic reviews of interventions. Hoboken: NJ: Wiley; 2008. p. 187241. [18] Le Henanff A, Giraudeau B, Baron G, et al. Quality of reporting of non inferiority and equivalence randomized trials. JAMA 2006;295:114751. [19] Gtzsche PC. Readers as research detectives. Trials 2009;10:2. [20] Al-Marzouki S, Roberts I, Marshall T, et al. The effect of scientic misconduct on the results of clinical trials: a Delphi survey. Contemp Clin Trials 2005;26:3317. [21] Jonville-Bra AP, Giraudeau B, Autret-Leca E. Reporting of drug tolerance in randomized clinical trials: when data conict with authors conclusions. Ann Int Med 2006;144:306. [22] Dimick JB, Diener-West M, Lipsett PA. Negative results of randomized clinical trials published in the surgical literature: equivalency or error? Arch Surg 2001;136:796800. [23] Bero L, Oostvogel F, Baccheti P, et al. Factors associated with ndings of published trials of drug-drug comparisons: why some statins appear more efcacious than others. Plos Med 2007;4:e184. [24] http://www.nia.nih.gov/NR/rdonlyres/D95A1337-6340-4C95-96F9ABD3FD3C0F03/0/NIAGlossaryofClinicalResearchTermsFINAL.doc. [25] Freiman JA, Chalmers TC, Smith Jr H, et al. The importance of beta, the type II error and sample size in the design and interpretation of the randomized control trial. Survey of 71 negative trials. N Engl J Med 1978;299:6904. [26] Moher D, Dulberg CS, Wells GA. Statistical power, sample size, and their reporting in randomized controlled trials. JAMA 1994;272:1224. [27] Altman DG, Bland JM. Absence of evidence is not evidence of absence. BMJ 1995;311:485. [28] Dickersin K. How important is publication bias? A synthesis of available data. AIDS Educ Prev 1997;9:1521. [29] Johnson RT, Dickersin K. Publication bias against negative results from clinical trials: three of the seven deadly sins. Nat Clin Pract Neurol 2007;3:5901. [30] Olson CM, Rennie D, Cook D, et al. Publication bias in editorial decision making. JAMA 2002;287:28258. [31] De Angelis CD, Drazen JM, Frizelle FA, et al. Is this clinical trial fully registered? A statement from the International Committee of Medical Journal editors. Lancet 2005;365:18279. [32] Mathieu S, Boutron I, Moher D, et al. Comparison of registered and published primary outcomes in randomized controlled trials. JAMA 2009;302:977. [33] Goldbeck-Wood S. Changes between protocol and manuscript should be declared at submission. BMJ 2001;322:14601. [34] Siegel JP. Editorial review of protocols for clinical trials. N Engl J Med 1990;323:1355. [35] Ioannidis JP, Evans SJ, Gotzsche PC, et al. Better reporting of harms in randomized trials: an extension of the Consort statement. Ann Intern Med 2004;141:7818. [36] Ioannidis JP, Contopoulos-Ioannidis DG. Reporting of safety data from randomised trials. Lancet 1998;352:17523. [37] Lee BB. A misleading conclusion and abstract. Anesth Analg 2003;96:62633. [38] Mills E, Montori VM, Thabane L. Angiotensine-converting enzyme inhibitor meta-analysis provides misleading conclusion. Am J Cardiol 2004;93: 2403. [39] Postma MJ. Cost-effectiveness analysis of Human Papillomavirus (HPV) Vaccination in the Netherlands: recent publication reinforces favourable cost-effectiveness despite misleading conclusion. Vaccine 2010;28: 8734.

of our particular experience in this specialty and because being familiar with a topic makes it easier to assess the results of a trial. However, other studies are needed to see whether our results are conrmed in other medical areas. We found discussion of misleading or wrong conclusions in articles on anaesthesia, cardiology or infectious diseases in letters to the editor or editorials [3739] but to our knowledge, the present study is the rst assessment of the frequency of misleading abstract conclusions in a sample of RCTs. Fourth, we focused on RCTs with an objective of superiority. We decided to exclude non-inferiority or equivalence trials because their methodology is unrelated to superiority trials and their interpretation is completely different. Finally, only the abstract conclusions identied as being misleading by one of us (SM) were conrmed by two of the other authors (PR and BG). At worst, this underestimates the prevalence of misleading conclusions. In conclusion, this study shows that almost one-quarter of the abstracts in a sample of 144 reports of rheumatology RCTs, especially those with negative results, contained misleading conclusions. Authors should report their results with greater transparency, and reviewers and editors should ensure that the salient features agree with information provided in the trial registration. Disclosure of interest The authors declare that they have no conicts of interest concerning this article. Appendix A. Supplementary data Supplementary data (Fig. S1, S2) associated with this article can be found at http://www.sciencedirect.com, at doi:10.1016/j.jbspin.2011.05.008. References
[1] Mc Donald S. Number and size of randomized trials reported in general health care journals from 1948 to 1997. Int J Epidemiol 2002;31:1257. [2] Altman DG, Schulz KF, Moher D, et al. The revised CONSORT statement for reporting randomized trials: explanation and elaboration. Ann Intern Med 2001;134:66394. [3] DerSimonian R, Charette LJ, McPeek B, et al. Reporting on methods in clinical trials. N Engl J Med 1982;306:13327. [4] Godlee F. Publishing study protocols: making them more visible will improve registration, reporting and recruitment. BMC News Views 2001;2:4. [5] Chan AW. Bias, spin, and misreporting: time for full access to trial protocols and results. PLoS Med 2008;5:e230. [6] Poynard T. valuation de la qualit mthodologique des essais thrapeutiques randomiss. Presse Med 1988;17:3158. [7] Moher D, Schulz KF, Altman D, et al. The CONSORT statement: revised recommendations for improving the quality of reports of parallel-group randomized trials. JAMA 2001;285:198791. [8] Kane RL, Wang J, Garrard J. Reporting in randomized clinical trials improved after adoption of the CONSORT statement. J Clin Epidemiol 2007;60: 2419. [9] Moher D, Jones A, Lepage L, et al. Use of the CONSORT statement and quality of reports of randomized trials: a comparative before-and-after evaluation. JAMA 2001;285:19925. [10] Mills EJ, Wu P, Gagnier J, et al. The quality of randomized trial reporting in leading medical journals since the revised CONSORT statement. Contemp Clin Trials 2005;26:4807.

Please cite this article in press as: Mathieu S, et al. Misleading abstract conclusions in randomized controlled trials in rheumatology: Comparison of the abstract conclusions and the results section. Joint Bone Spine (2011), doi:10.1016/j.jbspin.2011.05.008

You might also like