Sweetened and unsweetened coffee consumption associated with lower death risk

1. Sweetened and unsweetened coffee consumption associated with lower death risk

Abstract: https://www.acpjournals.org/doi/10.7326/M21-2977      

Editorial: https://www.acpjournals.org/doi/10.7326/M22-1465 

Note : HD soundbites of the author explaining study findings are available for download at https://www.dssimon.com/MM/ACP-coffee-sugar

URLs go live when the embargo lifts

A cohort study has found that compared to non-coffee drinkers, adults who drank moderate amounts (1.5 to 3.5 cups per day) of unsweetened coffee or coffee sweetened with sugar were less likely to die during a 7-year follow up period. The results for those who used artificial sweeteners were less clear. The findings are published in Annals of Internal Medicine.

Previous studies observing the health effects of coffee have found that coffee consumption is associated with a lower risk of death but did not distinguish between unsweetened coffee and coffee consumed with sugar or artificial sweeteners.

Researchers from Southern Medical University in Guangzhou, China used data from the U.K. Biobank study health behavior questionnaire to evaluate the associations of consumption of sugar-sweetened, artificially sweetened, and unsweetened coffee with all-cause and cause-specific mortality. More than 171,000 participants from the U.K. without known heart disease or cancer were asked several dietary and health behavior questions to determine coffee consumption habits. The authors found that during the 7-year follow up period, participants who drank any amount of unsweetened coffee were 16 to 21 percent less likely to die than participants who did not drink coffee. They also found that participants who drank 1.5 to 3.5 daily cups of coffee sweetened with sugar were 29 to 31 percent less likely to die than participants who did not drink coffee. The authors noted that adults who drank sugar-sweetened coffee added only about 1 teaspoon of sugar per cup of coffee on average. Results were inconclusive for participants who used artificial sweeteners in their coffee.

Any accompanying editorial by the editors of Annals of Internal Medicine notes that while coffee has qualities that could make health benefits possible, confounding variables including more difficult to measure differences in socioeconomic status, diet, and other lifestyle factors may impact findings. The authors add that the participant data is at least 10 years old and collected from a country where tea is a similarly popular beverage. They caution that the average amount of daily sugar per cup of coffee recorded in this analysis is much lower than specialty drinks at popular coffee chain restaurants, and many coffee consumers may drink it in place of other beverages that make comparisons to non-drinkers more difficult. Based on this data, clinicians can tell their patients that there is no need for most coffee drinkers to eliminate the beverage from their diet but to be cautious about higher calorie specialty coffees.

Media contacts: For an embargoed PDF or to speak with an author, please contact Angela Collom at acollom@acponline.org.

-------------------------------------------------

2.  Machine learning assistance does not negatively impact risk of bias assessments

Abstract: https://www.acpjournals.org/doi/10.7326/M22-0092 

Editorial: https://www.acpjournals.org/doi/10.7326/M22-1439 

URL goes live when the embargo lifts

A randomized control trial (RCT) has found that using RobotReviewer, an open-access platform

that partially automates risk of bias (RoB) assessments, to assist researchers in RoB assessments is noninferior to assessments conducted without using RobotReviewer. The findings are published in Annals of Internal Medicine.

As both the quantity of and methodological standards in research increase, conducting systemic reviews efficiently and accurately becomes more difficult. RoB assessment is an important but resource-intensive stage of the systemic review creation process, with a single assessment of one trial requiring an hour or more of the reviewer’s time. The use of automation and machine learning has been proposed to assist in this process, but uptake has been limited due to concerns about reliability. Previous studies have examined the reliability of RobotReviewer-only RoB assessments compared to human assessments, but none have examined the use of RobotReviewer-assisted RoB assessments.

Researchers from University College London and Monash University analyzed data collected from seven reviews, 145 studies, 290 individual assessment forms, and 1160 RoB judgments across four of seven Cochrane RoB domains provided by RobotReviewer. The authors found that the proportion of accurate RobotReview-assisted assessments was 0.89, while the proportion of accurate human-only assessments was 0.90. For each individual Cochrane RoB domain, the respective proportions of accurate assessments were 0.90 and 0.92 for sequence generation, 0.84 and 0.90 for allocation concealment, 0.94 and 0.90 for blinding of participants and personnel, and 0.87 and 0.89 for blinding of outcome assessors. The authors also measured person-time for RobotReviewer-assisted assessments and human-only assessments, but the results were inconclusive. The authors highlight that if systematic reviewers consider integrating the use of RobotReviewer into their work, the evidence from the study suggests that the overall quality of their resulting review will not be adversely affected.

An accompanying editorial by authors from the University of Alberta argue that despite limitations, this study provides an important contribution to the growing body of evidence that support automated or semiautomated review processes. They also write that the study demonstrates an important model for necessary future research examining the feasibility of automated and semiautomated processes. They urge the systemic review community to embrace the real-world use of semiautomated processes into their workflows and embedding evaluations to evaluate effectiveness.

Media contacts: For an embargoed PDF or to connect with the corresponding author, Anneliese Arno, PhD, please contact Angela Collom at acollom@acponline.org.  

-------------------------------------------------

Also new in this issue:

Rethinking Professionalism Assessments in Medical Education

Rachel Mintz, BS; Leah Pierson, BA; and David Gibbes Miller, MSc

Ideas and Opinions

Abstract: https://www.acpjournals.org/doi/10.7326/M22-0488