Limitations and Misinterpretations of E-Values for Sensitivity Analyses of Observational Studies

The E-value was recently introduced on the basis of earlier work as “the minimum strength of association…that an unmeasured confounder would need to have with both the treatment and the outcome to fully explain away a specific treatment–outcome association, conditional on the measured covariates.” E-values have been proposed for wide application in observational studies evaluating causality. However, they have limitations and are prone to misinterpretation. E-values have a monotonic, almost linear relationship with effect estimates and thus offer no additional information beyond what effect estimates can convey. Whereas effect estimates are based on real data, E-values may make unrealistic assumptions. No general rule can exist about what is a “small enough” E-value, and users of the biomedical literature are not familiar with how to interpret a range of E-values. Problems arise for any measure dependent on effect estimates and their CIs—for example, bias due to selective reporting and dependence on choice of exposure contrast and level of confidence. The automation of E-values may give an excuse not to think seriously about confounding. Moreover, biases other than confounding may still undermine results. Instead of misused or misinterpreted E-values, the authors recommend judicious use of existing methods for sensitivity analyses with careful assumptions; systematic assessments of whether and how known confounders have been handled, along with consideration of their prevalence and magnitude; thorough discussion of the potential for unknown confounders considering the study design and field of application; and explicit caution in making causal claims from observational studies.

What Can Medicine Learn From Social Science Studies of Sexual Harassment?

Addressing Medication Costs During Primary Care Visits: A Before–After Study of Team-Based Training

Background: Medications contribute to patients' out-of-pocket costs, yet most clinicians do not routinely screen for patients' cost-of-medication (COM) concerns. Objective: To assess whether a single training session improves COM conversations. Design: Before–after cross-sectional surveys of patients and qualitative interviews with clinicians and staff. Setting: 7 primary care practices in 3 U.S. states. Participants: In total, 700 patients were surveyed from May 2017 to January 2018: 50 patients per practice before the intervention and another 50 patients per practice after the intervention. Eligibility criteria included age 18 years or older and taking 1 or more long-term medications. Qualitative interviews with 45 staff members were conducted. Intervention: A single 60-minute training session for clinicians and staff from each practice on COM importance, team-based screening, and cost-saving strategies. Measurements: Patient data (demographics, number of long-term medications, total monthly out-of-pocket medication costs, and history of cost-related medication nonadherence) were obtained immediately before and 3 months after the intervention. Practice staff were interviewed 3 months after the intervention. Results: A total of 700 patient surveys were completed. Frequency of COM discussion improved in 6 of the 7 practices and remained unchanged in 1 practice. Overall, COM conversations with patients increased from 17% at baseline to 32% postintervention (P = 0.00). There was substantial heterogeneity among sites in before–after differences in patient-reported out-of-pocket COM. Qualitative analyses from key informant interviews showed wide variation in implementation of screening approaches, workflow, adoption of a team-based approach, and strategies for addressing COM. Limitation: It is not known whether improvements in COM conversations were sustained beyond 3 months. Conclusion: A single team training to screen and address patients' medication cost concerns improved COM discussions over the short term. Further research is needed to assess sustained effects and impact on patient costs and medication adherence and to determine whether more intensive, scalable interventions are needed. Primary Funding Source: Robert Wood Johnson Foundation.

Excess Burden of Mental Illness and Hospitalization in Young-Onset Type 2 Diabetes: A Population-Based Cohort Study: Annals of Internal Medicine: Vol 170, No 3

Background: Type 2 diabetes (T2D) increases hospitalization risk. Young-onset T2D (YOD) (defined as onset before age 40 years) is associated with excess morbidity and mortality, but its effect on hospitalizations is unknown. Objective: To determine hospitalization rates among persons with YOD and to examine the effect of age at onset on hospitalization risk. Design: Prospective cohort study. Setting: Hong Kong. Participants: Adults aged 20 to 75 years in population-based (2002 to 2014; n = 422 908) and registry-based (2000 to 2014; n = 20 886) T2D cohorts. Measurements: All-cause and cause-specific hospitalization rates. Negative binomial regression models estimated effect of age at onset on hospitalization rate and cumulative bed-days from onset to age 75 years for YOD. Results: Patients with YOD had the highest hospitalization rates by attained age. In the registry cohort, 36.8% of YOD bed-days before age 40 years were due to mental illness. The adjusted rate ratios showed increased hospitalization in YOD versus usual-onset T2D (onset at age ≥40 years) (all-cause, 1.8 [95% CI, 1.7 to 2.0]; renal, 6.7 [CI, 4.2 to 10.6]; diabetes, 3.7 [CI, 3.0 to 4.6]; cardiovascular, 2.1 [CI, 1.8 to 2.5]; infection, 1.7 [CI, 1.4 to 2.1]; P < 0.001 for all). Models estimated that intensified risk factor control in YOD (hemoglobin A1c level <6.2%, systolic blood pressure <120 mm Hg, low-density lipoprotein cholesterol level <2.0 mmol/L [<77.3 mg/dL], triglyceride level <1.3 mmol/L [<115.1 mg/dL], waist circumference of 85 cm [men] or 80 cm [women], and smoking cessation) was associated with a one-third reduction in cumulative bed-days from onset to age 75 years (97 to 65 bed-days). Limitation: Possible residual confounding. Conclusion: Adults with YOD have excess hospitalizations across their lifespan compared with persons with usual-onset T2D, including an unexpectedly large burden of mental illness in young adulthood. Efforts to prevent YOD and intensify cardiometabolic risk factor control while focusing on mental health are urgently needed. Primary Funding Source: Asia Diabetes Foundation.

Fitness and Body Mass Index During Adolescence and Disability Later in Life: A Cohort Study: Annals of Internal Medicine: Vol 170, No 4

Background: Low physical fitness, obesity, and the combination of the two in adolescence may be related to risk for disability in adulthood, but this has rarely been studied. Objective: To examine individual and combined associations of cardiorespiratory fitness and obesity in male adolescents with later receipt of a disability pension due to all and specific causes. Design: Population-based cohort study. Setting: Sweden. Participants: 1 079 128 Swedish adolescents aged 16 to 19 years who were conscripted into the military between 1972 and 1994. Measurements: Cardiorespiratory fitness and body mass index (BMI) were measured at conscription and were related to information on later receipt of a disability pension obtained from the Social Insurance Agency. Results: Over a median follow-up of 28.3 years, 54 304 men were granted a disability pension. Low cardiorespiratory fitness was strongly associated with later receipt of a disability pension due to all causes (hazard ratio, 3.74 [95% CI, 3.55 to 3.95] for lowest vs. highest fitness decile) and specific causes (psychiatric, musculoskeletal, injuries, nervous system, circulatory, and tumors). Obesity was associated with greater risk for receipt of a disability pension due to all and specific causes, with the greatest risks observed for class II and III obesity. Compared with being unfit, being moderately or highly fit was associated with attenuated risk for receipt of a disability pension across BMI categories. Limitation: The cohort did not include women, had data on smoking and alcohol intake only in a subsample, and lacked repeated measures of exposures and covariates. Conclusion: Low cardiorespiratory fitness, obesity, and the combination of the two were strongly associated with later chronic disability due to a wide range of diseases and causes. Although additional well-designed studies are required, these findings support the importance of high cardiorespiratory fitness and healthy body weight during adolescence to prevent later chronic disease. Primary Funding Source: Karolinska Institutet.

Intensive Models of Hepatitis C Care for People Who Inject Drugs Receiving Opioid Agonist Therapy: A Randomized Controlled Trial: Annals of Internal Medicine: Vol 170, No 9

Background: Many people who inject drugs (PWID) are denied treatment for hepatitis C virus (HCV) infection, even if they are receiving opioid agonist therapy (OAT). Research suggests that HCV in PWID may be treated effectively, but optimal models of care for promoting adherence and sustained virologic response (SVR) have not been evaluated in the direct-acting antiviral (DAA) era. Objective: To determine whether directly observed therapy (DOT) and group treatment (GT) are more effective than self-administered individual treatment (SIT) in promoting adherence and achieving SVR among PWID receiving OAT. Design: Three-group, randomized controlled trial conducted from October 2013 to April 2017. (ClinicalTrials.gov: NCT01857245) Setting: Three OAT programs in Bronx, New York. Participants: Persons aged 18 years and older with genotype 1 HCV infection who were willing to receive HCV therapy on site in the OAT program. Of 190 persons screened, 158 were randomly assigned to a study group and 150 initiated treatment: DOT (n = 51), GT (n = 48), and SIT (n = 51). Intervention: 2 intensive interventions (DOT and GT) and 1 control condition (SIT). Measurements: Primary: adherence, measured by using electronic blister packs. Secondary: HCV treatment completion and SVR 12 weeks after treatment completion. Results: Mean age was 51 years; 65% of participants had positive results on urine drug testing during the 6 months before treatment, and 75% reported ever injecting drugs. Overall adherence, estimated from mixed-effects models using the daily timeframe, was 78% (95% CI, 75% to 81%) and was greater among participants randomly assigned to DOT (86% [CI, 80% to 92%]) than those assigned to SIT (75% [CI, 70% to 81%]; difference, 11% [CI, 5% to 18%]; Bonferroni-corrected P = 0.001). No significant difference in adherence was observed between participants randomly assigned to GT (80% [CI, 74% to 86%]) and those assigned to SIT (difference, 4.7% [CI, −2% to 11%]; Bonferroni-corrected P = 0.29). The HCV treatment completion rate was 97%, with no differences among groups (P = 0.53). Overall SVR was 94% (CI, 89% to 97%); the SVR rate was 98% in the DOT group, 94% in the GT group, and 90% in the SIT group (P = 0.152). Limitation: These findings may not be generalizable to PWID not enrolled in OAT programs. Conclusion: All models of onsite HCV care delivered to PWID in OAT programs resulted in high SVR, despite ongoing drug use. Directly observed therapy was associated with greater adherence than SIT. Primary Funding Source: National Institute on Drug Abuse and Gilead Sciences.

Smoking, Vaping, and Use of Edibles and Other Forms of Marijuana Among U.S. Adults

Self-reported Physical Health of Women Who Did and Did Not Terminate Pregnancy After Seeking Abortion Services: A Cohort Study: Annals of Internal Medicine: Vol 171, No 4

Background: Research demonstrates worse short-term morbidity and mortality associated with childbirth than with abortion, but little research has examined long-term physical health in women with unwanted pregnancies after abortion versus childbirth. Objective: To examine the physical health of women who seek and receive or are denied abortion. Design: Prospective cohort study. Setting: 30 U.S. abortion facilities from 2008 to 2010. Participants: Of 1132 women seeking abortion who consented to participate, 874 were included in this analysis (328 who had first-trimester abortion, 383 who had second-trimester abortion, and 163 who gave birth). Measurements: Self-reported overall health; chronic abdominal, pelvic, back, and joint pain; chronic headaches or migraines; obesity; asthma; gestational and nongestational hypertension and diabetes; and hyperlipidemia were assessed semiannually for 5 years. Mortality was assessed by using verbal autopsy and public records. Results: No significant differences were observed in self-rated health or chronic pain after first-trimester versus second-trimester abortion. At 5 years, 27% (95% CI, 21% to 34%) of women who gave birth reported fair or poor health compared with 20% (CI, 16% to 24%) of women who had first-trimester abortion and 21% (CI, 18% to 25%) who had second-trimester abortion. Women who gave birth also reported more chronic headaches or migraines and joint pain, but experienced similar levels of other types of chronic pain and obesity. Gestational hypertension was reported by 9.4% of participants who gave birth. Eight of 1132 participants died during follow-up, 2 in the postpartum period. Maternal mortality did not differ statistically by group. Limitation: Self-reported outcome measures, uncertain generalizability, and 41% loss to follow-up at 5 years. Conclusion: Although some argue that abortion is detrimental to women's health, these study data indicate that physical health is no worse in women who sought and underwent abortion than in women who were denied abortion. Indeed, differences emerged suggesting worse health among those who gave birth. Primary Funding Source: An anonymous foundation.

Brief Commentary: Glycemic Targets for Older Adults With Type 2 Diabetes

Ten Principles for More Conservative, Care-Full Diagnosis