Back

Exploratory Data Analysis Recognizing Geometrical Patterns in Meta-analysis

Kimoto, K.; Yamakuchi, M.; Takenouchi, K.; Hashiguchi, T.

2022-06-07 epidemiology
10.1101/2022.06.06.22275944 medRxiv
Show abstract

BackgroundsStrange controversies have remained in thrombosis-related fields (travelers and COVID-19-related thrombosis), although travelers thrombosis is well-known even among non-professionals as "economy class syndrome." We hypothesized there might be something overlooked behind those strange situations. MethodsSince ordinary review methods (e.g., systematic review or meta-analysis) had already been conducted, we focused on reviewing a "previously published" "chart." Also, we developed a novel "review method" for the meta-regression analysis result. We applied those to some previously published and well-known data. ResultsWe newly found an approximately 28 days cycle of thrombosis onset over several weeks after travel in a figure. Also, we found an eighteen-day cycle of thrombosis onset in another chart. In COVID-19 cardiovascular biomarker studies, we newly extracted subgroup patterns in a scatterplot (Troponin T and NT-proBNP) that applied simple linear regression analysis. Also, these subgroups had already appeared in the cardiomyopathy study. ConclusionsTravelers thrombosis sometimes occurs over two months after leaving the risky in-flight environment. This phenomenon has been explained that the thrombus is formed in a cabin but dislodged after. However, from the cyclic patterns, explaining that the "high-risk period of thrombosis with Oral Contraceptive (OC) use initiation" coincided with "travel" is more reasonable (e.g., honeymoon and OC initiation). Regarding risk-benefit balance, it is conceivable that "spreading the risk" by starting the dosing away from the travel period is essential to ensure safer use because the in-flight environment may have a non-zero effect, and optimal care by a primary care physician (prescriber) is not available during the travel. In COVID-19, there seems to be a complex scatterplot structure that is unsuitable for usually used simple linear fitting. In a literature review, a pattern on a chart should be given more paying attention.

Matching journals

The top 9 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.