PT  - JOURNAL ARTICLE
AU  - Xiaoling Yuan
AU  - Yong Lin
AU  - Yating Wang
AU  - Lanjing Zhang
TI  - Reporting quality of trend analyses in leading medicine and oncology journals
AID  - 10.1101/2020.09.18.303701
DP  - 2020 Jan 01
TA  - bioRxiv
PG  - 2020.09.18.303701
4099  - http://biorxiv.org/content/early/2020/09/21/2020.09.18.303701.short
4100  - http://biorxiv.org/content/early/2020/09/21/2020.09.18.303701.full
AB  - Trend analyses are critical for assessing changes and predicting future of epidemiological parameters. Recent trend-analysis guidelines recommended reporting slopes or beta/coefficient if possible. American Statistical Association and others recommended reporting effect-size. However, reporting of these statistics in trend analyses was largely unclear. We searched the Pubmed for the articles published during Jan. 1, 2008 to Dec. 31, 2018, whose titles included “trend” or “trends” among the following medicine and oncology journals: Ann Intern Med, Ann Oncol, BMJ, J Clin Oncol, J Natl Cancer Inst, JAMA Oncol, JAMA, Lancet, Lancet Oncol and N Engl J Med. We examined proportions of the trend analyses in leading medicine and oncology journals, that reported p-value, effect-size or beta/coefficient/Annual percent change (APC), and identified associated factors. Among the 398 identified reports of trend analysis published during 2008-2018, there were 297 qualified reports, including 38 (12.8%) analyses using non-parametric model, 226 (76.1%) analyses using (piece-wise) linear model, 32 (10.8%) analyses using non-linear parametric model and 1 (0.3%) analyses using semi-parametric model (Cox regression). In these analyses, 193 (66.0%) and 216 (72,7%) analyses reported P-value and effect-size, respectively. Among the 226 trend analyses using linear model, 169 (74.8%) reported p-value, 183 (81.0%) reported effect-size, 94 (41.6%) reported APC and 34 (15.0%) reported Beta/coefficient. Only 13 (5.8%) analyses reported neither p-value/effect size nor beta/coefficient/APC. In multivariable regression models, author affiliation with epidemiology department or statistics department was associated with reporting effect-size, and U.S. senior-authors (versus non-U.S.) more likely reported p-value. No factors were linked to reporting APC. In summary, reporting quality of the included trend analyses was overall good. However, reporting p-value or effect-size did not change with publication years, and reporting APC or Beta/coefficient was uncommon. Additional works and training appear warranted to improve reporting quality of trend analyses.Competing Interest StatementThe authors have declared no competing interest.