Initial antiretroviral therapy regimen and risk of heart failure

OBJECTIVES

Heart failure risk is elevated in people with HIV (PWH). We investigated whether initial antiretroviral therapy (ART) regimens influenced heart failure risk.

DESIGN

Cohort study.

METHODS

PWH who initiated an ART regimen between 2000 and 2016 were identified from three integrated healthcare systems. We evaluated heart failure risk by protease inhibitor, nonnucleoside reverse transcriptase inhibitors (NNRTI), and integrase strand transfer inhibitor (INSTI)-based ART, and comparing two common nucleotide reverse transcriptase inhibitors: tenofovir disoproxil fumarate (tenofovir) and abacavir. Follow-up for each pairwise comparison varied (i.e. 7 years for protease inhibitor vs. NNRTI; 5 years for tenofovir vs. abacavir; 2 years for INSTIs vs. PIs or NNRTIs). Hazard ratios were from working logistic marginal structural models, fitted with inverse probability weighting to adjust for demographics, and traditional cardiovascular risk factors.

RESULTS

Thirteen thousand six hundred and thirty-four PWH were included (88% men, median 40 years of age; 34% non-Hispanic white, 24% non-Hispanic black, and 24% Hispanic). The hazard ratio (95% CI) were: 2.5 (1.5-4.3) for protease inhibitor vs. NNRTI-based ART (reference); 0.5 (0.2-1.8) for protease inhibitor vs. INSTI-based ART (reference); 0.1 (0.1-0.8) for NNRTI vs. INSTI-based ART (reference); and 1.7 (0.5-5.7) for tenofovir vs. abacavir (reference). In more complex models of cumulative incidence that accounted for possible nonproportional hazards over time, the only remaining finding was evidence of a higher risk of heart failure for protease inhibitor compared with NNRTI-based regimens (1.8 vs. 0.8%; P  = 0.002).

CONCLUSION

PWH initiating protease inhibitors may be at higher risk of heart failure compared with those initiating NNRTIs. Future studies with longer follow-up with INSTI-based and other specific ART are warranted.

A comparative effectiveness study of carotid intervention for long-term stroke prevention in patients with severe asymptomatic stenosis from a large integrated health system

OBJECTIVE

The results of current prospective trials comparing the effectiveness of carotid endarterectomy (CEA) vs standard medical therapy for long-term stroke prevention in patients with asymptomatic carotid stenosis (ACS) will not be available for several years. In this study, we compared the observed effectiveness of CEA and standard medical therapy vs standard medical therapy alone to prevent ipsilateral stroke in a contemporary cohort of patients with ACS.

METHODS

This cohort study was conducted in a large integrated health system in adult subjects with 70% to 99% ACS (no neurologic symptom within 6 months) with no prior ipsilateral carotid artery intervention. Causal inference methods were used to emulate a conceptual randomized trial using data from January 1, 2008, through December 31, 2017, for comparing the event-free survival over 96 months between two treatment strategies: (1) CEA within 12 months from cohort entry vs (2) no CEA (standard medical therapy alone). To account for both baseline and time-dependent confounding, inverse probability weighting estimation was used to derive adjusted hazard ratios, and cumulative risk differences were assessed based on two logistic marginal structural models for counterfactual hazards. Propensity scores were data-adaptively estimated using super learning. The primary outcome was ipsilateral anterior ischemic stroke.

RESULTS

The cohort included 3824 eligible patients with ACS (mean age: 73.7 years, 57.9% male, 12.3% active smokers), of whom 1467 underwent CEA in the first year, whereas 2297 never underwent CEA. The median follow-up was 68 months. A total of 1760 participants (46%) died, 445 (12%) were lost to follow-up, and 158 (4%) experienced ipsilateral stroke. The cumulative risk differences for each year of follow-up showed a protective effect of CEA starting in year 2 (risk difference = 1.1%, 95% confidence interval: 0.5%-1.6%) and persisting to year 8 (2.6%, 95% confidence interval: 0.3%-4.8%) compared with patients not receiving CEA.

CONCLUSIONS

In this contemporary cohort study of patients with ACS using rigorous analytic methodology, CEA appears to have a small but statistically significant effect on stroke prevention out to 8 years. Further study is needed to appropriately select the subset of patients most likely to benefit from intervention.

Development of cardiometabolic risk factors following endocrine therapy in women with breast cancer

PURPOSE

Studies comparing the effect of aromatase inhibitor (AI) and tamoxifen use on cardiovascular disease (CVD) risk factors in hormone receptor-positive breast cancer (BC) survivors report conflicting results. We examined associations of endocrine therapy use with incident diabetes, dyslipidemia, and hypertension.

METHODS

The Pathways Heart Study examines cancer treatment exposures with CVD-related outcomes in Kaiser Permanente Northern California members with BC. Electronic health records provided sociodemographic and health characteristics, BC treatment, and CVD risk factor data. Hazard ratios (HR) and 95% confidence intervals (CI) of incident diabetes, dyslipidemia, and hypertension in hormone receptor-positive BC survivors using AIs or tamoxifen compared with survivors not using endocrine therapy were estimated using Cox proportional hazards regression models adjusted for known confounders.

RESULTS

In 8985 BC survivors, mean baseline age and follow-up time was 63.3 and 7.8 years, respectively; 83.6% were postmenopausal. By treatment, 77.0% used AIs, 19.6% used tamoxifen, and 16.0% used neither. Postmenopausal women who used tamoxifen had an increased rate (HR 1.43, 95% CI 1.06-1.92) of developing hypertension relative to those who did not use endocrine therapy. Tamoxifen use was not associated with incident diabetes, dyslipidemia, or hypertension in premenopausal BC survivors. Postmenopausal AI users had higher hazard rates of developing diabetes (HR 1.37, 95% CI 1.05-1.80), dyslipidemia (HR 1.58, 95% CI 1.29-1.92), and hypertension (HR 1.50, 95% CI 1.24-1.82) compared with non-endocrine therapy users.

CONCLUSION

Hormone receptor-positive BC survivors treated with AIs may have higher rates of developing diabetes, dyslipidemia, and hypertension over an average 7.8 years post-diagnosis.

Development of cardiometabolic risk factors following endocrine therapy in women with breast cancer

PURPOSE

Studies comparing the effect of aromatase inhibitor (AI) and tamoxifen use on cardiovascular disease (CVD) risk factors in hormone-receptor positive breast cancer (BC) survivors report conflicting results. We examined associations of endocrine therapy use with incident diabetes, dyslipidemia, and hypertension.

METHODS

The Pathways Heart Study examines cancer treatment exposures with CVD-related outcomes in Kaiser Permanente Northern California members with BC. Electronic health records provided sociodemographic and health characteristics, BC treatment, and CVD risk factor data. Hazard ratios (HR) and 95% confidence intervals (CI) of incident diabetes, dyslipidemia, and hypertension in hormone-receptor positive BC survivors using AIs or tamoxifen compared with survivors not using endocrine therapy were estimated using Cox proportional hazards regression models adjusted for known confounders.

RESULTS

In 8,985 BC survivors, mean baseline age and follow-up time was 63.3 and 7.8 years, respectively; 83.6% were postmenopausal. By treatment, 77.0% used AIs, 19.6% used tamoxifen, and 16.0% used neither. Postmenopausal women who used tamoxifen had an increased rate (HR: 1.43, 95% CI: 1.06-1.92) of developing hypertension relative to those who did not use endocrine therapy. Tamoxifen use was not associated with incident diabetes, dyslipidemia, or hypertension in premenopausal BC survivors. Postmenopausal AI users had higher hazard rates of developing diabetes (HR: 1.37, 95% CI: 1.05-1.80), dyslipidemia (HR: 1.58, 95% CI: 1.29-1.92) and hypertension (HR: 1.50, 95% CI: 1.24-1.82) compared with non-endocrine therapy users.

CONCLUSION

Hormone-receptor positive BC survivors treated with AIs may have higher rates of developing diabetes, dyslipidemia, and hypertension over an average 7.8 years post-diagnosis.

Practice Patterns and Outcomes Associated With Anticoagulation Use Following Sepsis Hospitalizations With New-Onset Atrial Fibrillation

BACKGROUND

Practice patterns and outcomes associated with the use of oral anticoagulation for arterial thromboembolism prevention following a hospitalization with new-onset atrial fibrillation (AF) during sepsis are unclear.

METHODS

Retrospective, observational cohort study of patients ≥40 years of age discharged alive following hospitalization with new-onset AF during sepsis across 21 hospitals in the Kaiser Permanente Northern California health care delivery system, years 2011 to 2018. Primary outcomes were ischemic stroke/transient ischemic attack (TIA), with a safety outcome of major bleeding events, both within 1 year of discharge alive from sepsis hospitalization. Adjusted risk differences for outcomes between patients who did and did not receive oral anticoagulation within 30 days of discharge were estimated using marginal structural models fitted by inverse probability weighting using Super Learning within a target trial emulation framework.

RESULTS

Among 82 748 patients hospitalized with sepsis, 3992 (4.8%) had new-onset AF and survived to hospital discharge; mean age was 78±11 years, 53% were men, and 70% were White. Patients with new-onset AF during sepsis averaged 45±33% of telemetry monitoring entries with AF, and 27% had AF present on the day of hospital discharge. Within 1 year of hospital discharge, 89 (2.2%) patients experienced stroke/TIA, 225 (5.6%) had major bleeding, and 1011 (25%) died. Within 30 days of discharge, 807 (20%) patients filled oral anticoagulation prescriptions, which were associated with higher 1-year adjusted risks of ischemic stroke/TIA (5.69% versus 2.32%; risk difference, 3.37% [95% CI, 0.36-6.38]) and no significant difference in 1-year adjusted risks of major bleeding (6.51% versus 7.10%; risk difference, -0.59% [95% CI, -3.09 to 1.91]). Sensitivity analysis of ischemic stroke-only outcomes showed a risk difference of 0.15% (95% CI, -1.72 to 2.03).

CONCLUSIONS

After hospitalization with new-onset AF during sepsis, oral anticoagulation use was uncommon and associated with potentially higher stroke/TIA risk. Further research to inform mechanisms of stroke and TIA and management of new-onset AF after sepsis is needed.

Risk of Cardiovascular Disease in Women With and Without Breast Cancer: The Pathways Heart Study

PURPOSE

To examine cardiovascular disease (CVD) and mortality risk in women with breast cancer (BC) by cancer therapy received relative to women without BC.

METHODS

The study population comprised Kaiser Permanente Northern California members. Cases with invasive BC diagnosed from 2005 to 2013 were matched 1:5 to controls without BC on birth year and race/ethnicity. Cancer treatment, CVD outcomes, and covariate data were from electronic health records. Multivariable Cox proportional hazards models estimated hazard ratios (HRs) and 95% CIs of CVD incidence and mortality by receipt of chemotherapy treatment combinations, radiation therapy, and endocrine therapy.

RESULTS

A total of 13,642 women with BC were matched to 68,202 controls without BC. Over a 7-year average follow-up (range < 1-14 years), women who received anthracyclines and/or trastuzumab had high risk of heart failure/cardiomyopathy relative to controls, with the highest risk seen in women who received both anthracyclines and trastuzumab (HR, 3.68; 95% CI, 1.79 to 7.59). High risk of heart failure and/or cardiomyopathy was also observed in women with BC with a history of radiation therapy (HR, 1.38; 95% CI, 1.13 to 1.69) and aromatase inhibitor use (HR, 1.31; 95% CI, 1.07 to 1.60), relative to their controls. Elevated risks for stroke, arrhythmia, cardiac arrest, venous thromboembolic disease, CVD-related death, and death from any cause were also observed in women with BC on the basis of cancer treatment received.

CONCLUSION

Women with BC had increased incidence of CVD events, CVD-related mortality, and all-cause mortality compared with women without BC, and risks varied according to the history of cancer treatment received. Studies are needed to determine how women who received BC treatment should be cared for to improve cardiovascular outcomes.

Association of Glyburide and Subcutaneous Insulin With Perinatal Complications Among Women With Gestational Diabetes

IMPORTANCE

Nearly 30% of individuals with gestational diabetes (GDM) do not achieve glycemic control with lifestyle modification alone and require medication treatment. Oral agents, such as glyburide, have several advantages over insulin for the treatment of GDM, including greater patient acceptance; however, the effectiveness of glyburide for the treatment of GDM remains controversial.

OBJECTIVE

To compare the perinatal and neonatal outcomes associated with glyburide vs insulin using causal inference methods in a clinical setting with information on glycemic control.

DESIGN, SETTING, AND PARTICIPANTS

The population-based cohort study included patients with GDM who required medication treatment from 2007 to 2017 in Kaiser Permanente Northern California. Machine learning and rigorous casual inference methods with time-varying exposures were used to evaluate associations of exposure to glyburide vs insulin with perinatal outcomes. Data analysis was conducted from March 2018 to July 2017.

EXPOSURES

Time-varying exposure to glyburide vs insulin during pregnancy.

MAIN OUTCOMES AND MEASURES

Outcomes evaluated separately included neonatal hypoglycemia, jaundice, shoulder dystocia, respiratory distress syndrome (RDS), neonatal intensive care unit (NICU) admission, size-for-gestational age, and cesarean delivery. Inverse probability weighting (IPW) estimation was used to separately compare perinatal outcomes between those initiating glyburide and insulin. This approach was combined with Super Learning for propensity score estimation to account for both baseline and time-dependent confounding in both per-protocol (primary) and intention-to-treat (secondary) analyses to evaluate sustained exposure to the same therapy.

RESULTS

From 2007 to 2017, 11 321 patients with GDM (mean [SD] age, 32.9 [4.9] years) initiated glyburide or insulin during pregnancy. In multivariate models, the risk of neonatal respiratory distress was 2.03 (95% CI, 0.13-3.92) per 100 births lower and the risk of NICU admission was 3.32 (95% CI, 0.20-6.45) per 100 births lower after continuous exposure to glyburide compared with insulin. There were no statistically significant differences in glyburide vs insulin initiation in risk for neonatal hypoglycemia (0.85 [95% CI, -1.17 to 2.86] per 100 births), jaundice (0.02 [95% CI, -1.46 to 1.51] per 100 births), shoulder dystocia (-1.05 [95% CI, -2.71 to 0.62] per 100 births), or large-for-gestational age categories (-2.75 [95% CI, -6.31 to 0.80] per 100 births).

CONCLUSIONS AND RELEVANCE

Using data from a clinical setting and contemporary causal inference methods, our findings do not provide evidence of a difference in the outcomes examined between patients with GDM initiating glyburide compared with those initiating insulin.

Abstract PD5-03: Development of cardiometabolic risk factors following endocrine therapy: The pathways heart study

Purpose: Endocrine therapy is associated with cardiovascular disease among breast cancer (BC) survivors, with observed opposing effects between aromatase inhibitors (AIs) and tamoxifen. AIs deplete endogenous estrogen levels, while tamoxifen has mixed estrogenic and antiestrogenic activity. Yet, observational studies comparing AI vs. tamoxifen use may be confounded by indication and few have tested their associations with cardiometabolic risk factors. Therefore, we examined the association of AI or tamoxifen use on the incidence of newly diagnosed hypertension, diabetes, and dyslipidemia in a cohort of BC survivors within Kaiser Permanente Northern California (KPNC). Methods: The Pathways Heart Study is an ongoing cohort study within KPNC examining incident CVD outcomes and risk factors in 14,942 women with history of BC. Eligibility was: 1) stage I-IV invasive BC diagnosis between Nov 2005 and Mar 2013; 2) &ge;21 years; and 3) active KPNC membership &ge;12 months at diagnosis. KPNC records were used to collect demographic, socioeconomic, and health characteristics. Endocrine therapy was collected from outpatient pharmacy data. Incident hypertension, diabetes, and dyslipidemia were identified from ICD-9/10 codes, laboratory results, and/or medication use. Hazard ratios (HR) and 95% confidence intervals (CI) from Cox proportional models were used to determine whether AI and tamoxifen use were associated with incident hypertension, diabetes, and dyslipidemia compared to BC survivors not receiving either of these therapies. For each cardiometabolic risk factor, models adjusted for demographic, socioeconomic, and health characteristics and excluded women with the cardiometabolic risk factor at baseline. Results: Among 14,942 women with a history of BC, mean age at baseline was 61.2±12.8 years and mean follow-up time was 7.0±3.5 years (range 1-13.4). The frequency of use was: AI, n=6,070 (40.6%); tamoxifen, n=1,755 (11.8%); and neither, n=7,117 (47.6%). Regression models showed AI use was associated with increased risk of incident hypertension (HR: 1.1, 95% CI: 1.00-1.21) and increased risk of incident dyslipidemia (HR: 1.18, 95% CI: 1.07-1.3) relative to BC survivors who did not use endocrine therapy. (Table). In contrast, tamoxifen use was associated with decreased risk of dyslipidemia (HR: 0.8, 95% CI: 0.68-0.94) relative to BC survivors who did not use endocrine therapy. Neither AI nor tamoxifen use was associated with risk of incident diabetes. Conclusion: Compared to BC survivors who did not use endocrine therapy, women treated with AIs had a higher risk of incident hypertension and dyslipidemia, while women treated with tamoxifen had a lower risk of dyslipidemia. AIs reduce endogenous estrogen levels, which can alter lipid profiles, although prior studies have been inconsistent, possibly due to differences in steroidal and non-steroidal AIs. More work is needed to understand the implications of these associations on long-term cardiovascular health and how to best manage cardiometabolic risk factors in BC survivors with a history of endocrine therapy use.

Table. Adjusted1 hazard ratios (95% confidence intervals) of incident cardiometabolic risk factors among women with a history of breast cancer, by endocrine therapy useNo Endocrine Therapy(n=7,117)Endocrine TherapyAromatase inhibitor(n=6,070)Tamoxifen(n=1,755)Incident HypertensionRef1.10 (1.00, 1.21)0.98 (0.85, 1.14)Incident DiabetesRef0.99 (0.87, 1.13)0.98 (0.80, 1.20)Incident DyslipidemiaRef1.18 (1.07, 1.30)0.80 (0.68, 0.94)1Adjusted for age, race/ethnicity, baseline body mass index, AJCC stage, menopausal status, smoking status, education level, income, chemotherapy, radiation therapy, and prevalent cardiovascular disease.

Citation Format: Heather Greenlee, Eileen Rillamas-Sun, Carlos Iribarren, Richard Cheng, Romain Neugebauer, Jamal S. Rana, Mai Nguyen-Huynh, Zaixing Shi, Cecile A. Laurent, Valerie S. Lee, Janise M. Roh, Hanjie Shen, Dawn L. Hershman, Lawrence H. Kushi, Marilyn L. Kwan. Development of cardiometabolic risk factors following endocrine therapy: The pathways heart study [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr PD5-03.

Abstract PD5-01: Cardiovascular disease risk of breast cancer therapies: The pathways heart study

Purpose: Studies on long-term cardiovascular disease (CVD) risk in breast cancer (BC) survivors are limited. We examined CVD risk associated with exposure to specific BC therapies and explored whether body mass index (BMI) or prevalent CVD risk factors at BC diagnosis modified these associations. Methods: The Pathways Heart Study is a prospective cohort study examining incident CVD outcomes and risk factors in women with BC at Kaiser Permanente Northern California (KPNC). Eligible women were diagnosed with stage I-IV invasive BC from 2005-2013, &ge;21 years old, and KPNC members &ge;12 months at diagnosis. KPNC records provided demographic and BC therapy characteristics. Incident CVD outcomes [ischemic heart disease, heart failure/cardiomyopathy (HF/CM), stroke] were assessed from ICD9/10 codes. Multivariable Cox models estimated hazard ratios (HR) and 95% confidence intervals (CI) of each CVD outcome by cancer therapy received compared to not receiving that therapy, excluding those with prevalent CVD. Separate regression models included interaction terms for cancer therapy by overweight, obesity, diabetes, dyslipidemia, and hypertension to test whether the CVD outcome risk varied by presence of these factors at diagnosis. Results: Among 4,181 BC survivors with mean age of 59.6±12.0 years and mean follow-up of 7.9±3.5 years (range: 0.04-13.3), cancer therapies were not associated with incident CVD. However, CVD risks varied by BMI and prevalence of CVD risk factors at BC diagnosis. Normal weight (NW) women who received anthracyclines had higher risk of ischemic heart disease and HF/CM relative to NW women not receiving these therapies; interaction terms indicated HF/CM risk was statistically different than risks for obese women (Table). NW women who received cyclophosphamide or left-sided radiation had higher risk of HF/CM and stroke relative to NW women not receiving these therapies; these risks were statistically different from obese (for cyclophosphamide) or overweight (for radiation) women. Relative to women not receiving these therapies, higher HRs for HF/CM were observed among non-diabetic women who received cyclophosphamide (2.03, CI: 1.22-3.37), non-dyslipidemic women who received anthracyclines (3.65, CI: 1.69-7.87), and non-hypertensive women who received either anthracyclines (4.04, CI: 1.81-9.03) or cyclophosphamide (2.66, CI: 1.23-5.74) (P for interaction range: 0.04 to 0.06). Conclusion: Certain chemotherapy drugs may increase the risk of CVD in NW BC survivors; overweight and obese BC survivors may experience less risk than NW women. While chemotherapy also appears to increase HF/CM risk for women without diabetes, dyslipidemia, and hypertension, these conditions are more prevalent among overweight/obese women. Analysis within these subgroups is needed and forthcoming.

Table. Adjusted HRs (95% CI) of CVD outcomes among breast cancer survivors receiving select cancer therapies* stratified by BMI status at diagnosisBMI Ischemic heart diseaseHeart failure/CardiomyopathyStrokeAnthracycline, n=1283Normal4.22 (1.59, 11.2)5.27 (2.54, 10.9)1.89 (0.79, 4.53)Overweight1.66 (0.73, 3.77)2.17 (1.15, 4.11)0.40 (0.16, 0.99)Obese1.26 (0.56, 2.85)1.1 (0.54, 2.27)a0.33 (0.13, 0.83)aCyclophosphamide, n=1705Normal1.63 (0.61, 4.31)3.28 (1.59, 6.75)2.21 (1.01, 4.84)Overweight1.59 (0.75, 3.39)1.63 (0.9, 2.97)0.73 (0.34, 1.58)Obese0.85 (0.39, 1.86)0.75 (0.38, 1.47)a0.31 (0.13, 0.71)aLeft-Side Radiation, n=1331Normal1.44 (0.56, 3.69)2.04 (1.0, 4.18)2.38 (1.28, 4.42)Overweight1.47 (0.68, 3.16)0.68 (0.34, 1.34)b0.72 (0.37, 1.4)bObese1.32 (0.73, 2.38)1.30 (0.79, 2.16)1.05 (0.61, 1.82)*Cancer therapies with non-significant findings (i.e., Trastuzumab, taxanes, aromatase inhibitors, Tamoxifen, and any-side radiation) are not shown.ap≤0.05 normal weight v. obese; bp≤0.05 normal weight v. overweight

Citation Format: Heather Greenlee, Eileen Rillamas-Sun, Carlos Iribarren, Richard Cheng, Romain Neugebauer, Jamal S. Rana, Mai Nguyen-Huynh, Zaixing Shi, Cecile A. Laurent, Valerie S. Lee, Janise M. Roh, Hanjie Shen, Dawn L. Hershman, Lawrence H. Kushi, Marilyn L. Kwan. Cardiovascular disease risk of breast cancer therapies: The pathways heart study [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr PD5-01.

Risk of Cardiometabolic Risk Factors in Women With and Without a History of Breast Cancer: The Pathways Heart Study

PURPOSE

The incidence of cardiometabolic risk factors in breast cancer (BC) survivors has not been well described. Thus, we compared risk of hypertension, diabetes, and dyslipidemia in women with and without BC.

METHODS

Women with invasive BC diagnosed from 2005 to 2013 at Kaiser Permanente Northern California (KPNC) were identified and matched 1:5 to noncancer controls on birth year, race, and ethnicity. Cumulative incidence rates of hypertension, diabetes, and dyslipidemia were estimated with competing risk of overall death. Subdistribution hazard ratios (sHRs) were estimated by Fine and Gray regression, adjusted for cardiovascular disease-related risk factors, and stratified by treatment and body mass index (BMI).

RESULTS

A total of 14,942 BC cases and 74,702 matched controls were identified with mean age 61.2 years and 65% non-Hispanic White. Compared with controls, BC cases had higher cumulative incidence rates of hypertension (10.9% v 8.9%) and diabetes (2.1% v 1.7%) after 2 years, with higher diabetes incidence persisting after 10 years (9.3% v 8.8%). In multivariable models, cases had higher risk of diabetes (sHR, 1.16; 95% CI, 1.07 to 1.26) versus controls. Cases treated with chemotherapy (sHR, 1.23; 95% CI, 1.11 to 1.38), left-sided radiation (sHR, 1.29; 95% CI, 1.13 to 1.48), or endocrine therapy (sHR, 1.23; 95% CI, 1.12 to 1.34) continued to have higher diabetes risk. Hypertension risk was higher for cases receiving left-sided radiation (sHR, 1.11; 95% CI, 1.02 to 1.21) or endocrine therapy (sHR, 1.10; 95% CI, 1.03 to 1.16). Normal-weight (BMI < 24.9 kg/m²) cases had higher risks overall and within treatment subgroups versus controls.

CONCLUSION

BC survivors at KPNC experienced elevated risks of diabetes and hypertension compared with women without BC depending on treatments received and BMI. Future studies should examine strategies for cardiometabolic risk factor prevention in BC survivors.

Association of Cardiovascular Outcomes and Mortality With Sustained Long-Acting Insulin Only vs Long-Acting Plus Short-Acting Insulin Treatment

IMPORTANCE

Cardiovascular events and mortality are the principal causes of excess mortality and health care costs for people with type 2 diabetes. No large studies have specifically compared long-acting insulin alone with long-acting plus short-acting insulin with regard to cardiovascular outcomes.

OBJECTIVE

To compare cardiovascular events and mortality in adults with type 2 diabetes receiving long-acting insulin who do or do not add short-acting insulin.

DESIGN, SETTING, AND PARTICIPANTS

This retrospective cohort study emulated a randomized experiment in which adults with type 2 diabetes who experienced a qualifying glycated hemoglobin A1c (HbA1c) level of 6.8% to 8.5% with long-acting insulin were randomized to continuing treatment with long-acting insulin (LA group) or adding short-acting insulin within 1 year of the qualifying HbA1c level (LA plus SA group). Retrospective data in 4 integrated health care delivery systems from the Health Care Systems Research Network from January 1, 2005, to December 31, 2013, were used. Analysis used inverse probability weighting estimation with Super Learner for propensity score estimation. Analyses took place from April 1, 2018, to June 30, 2019.

EXPOSURES

Long-acting insulin alone or with added short-acting insulin within 1 year from the qualifying HbA1c level.

MAIN OUTCOMES AND MEASURES

Mortality, cardiovascular mortality, acute myocardial infarction, stroke, and hospitalization for heart failure.

RESULTS

Among 57 278 individuals (39 279 with data on cardiovascular mortality) with a mean (SD) age of 60.6 (11.5) years, 53.6% men, 43.5% non-Hispanic White individuals, and 4 years of follow-up (median follow-up of 11 [interquartile range, 5-20] calendar quarters), the LA plus SA group was associated with increased all-cause mortality compared with the LA group (hazard ratio, 1.27; 95% CI, 1.05-1.49) and a decreased risk of acute myocardial infarction (hazard ratio, 0.89; 95% CI, 0.81-0.97). Treatment with long-acting plus short-acting insulin was not associated with increased risks of congestive heart failure, stroke, or cardiovascular mortality.

CONCLUSIONS AND RELEVANCE

Findings of this retrospective cohort study suggested an increased risk of all-cause mortality and a decreased risk of acute myocardial infarction for the LA plus SA group compared with the LA group. Given the lack of an increase in major cardiovascular events or cardiovascular mortality, the increased all-cause mortality with long-acting plus short-acting insulin may be explained by noncardiovascular events or unmeasured confounding.

Separating Algorithms From Questions and Causal Inference With Unmeasured Exposures: An Application to Birth Cohort Studies of Early Body Mass Index Rebound

Observational studies reporting on adjusted associations between childhood body mass index (BMI; weight (kg)/height (m)2) rebound and subsequent cardiometabolic outcomes have often not paid explicit attention to causal inference, including definition of a target causal effect and assumptions for unbiased estimation of that effect. Using data from 649 children in a Boston, Massachusetts-area cohort recruited in 1999-2002, we considered effects of stochastic interventions on a chosen subset of modifiable yet unmeasured exposures expected to be associated with early (<age 4 years) BMI rebound (a proxy measure) on adolescent cardiometabolic outcomes. We considered assumptions under which these effects might be identified with available data. This leads to an analysis where the proxy, rather than the exposure, acts as the exposure in the algorithm. We applied targeted maximum likelihood estimation, a doubly robust approach that naturally incorporates machine learning for nuisance parameters (e.g., propensity score). We found a protective effect of an intervention that assigns modifiable exposures according to the distribution in the observational study of persons without (vs. with) early BMI rebound for fat mass index (fat mass (kg)/ height (m)2; -1.39 units, 95% confidence interval: -1.63, -0.72) but weaker or no effects for other cardiometabolic outcomes. Our results clarify distinctions between algorithms and causal questions, encouraging explicit thinking in causal inference with complex exposures.

A prognostic information system for real-time personalized care: Lessons for embedded researchers

Embedded researchers could play a central role in developing tools to personalize care using electronic medical records (EMRs). However, few studies have described the steps involved in developing such tools, or evaluated the key factors in success and failure. This case study describes how we used an EMR-derived data warehouse to develop a prototype informatics tool to help oncologists counsel patients with pancreatic cancer about their prognosis. The tool generated real-time prognostic information based on tumor type and stage, age, comorbidity status and lab tests. Our multidisciplinary team included embedded researchers, application developers, user experience experts, and an oncologist leader.This prototype succeeded in establishing proof of principle, but did not reach adoption into actual practice. In pilot testing, oncologists succeeded in generating prognostic information in real time. A few found it helpful in patient encounters, but all identified critical areas for further development before implementation. Generalizable lessons included the need to (1) include a wide range of potential use cases and stakeholders when selecting use cases for such tools; (2) develop talking points for clinicians to explain results from predictive tools to patients; (3) develop ways to reduce lag time between events and data availability; and (4) keep the options presented in the user interface very simple. This case demonstrates that embedded researchers can lead collaborations using EMR-derived data to create systems for real-time personalized patient counseling, and highlights challenges that such teams can anticipate.

Exercise During the First Trimester of Pregnancy and the Risks of Abnormal Screening and Gestational Diabetes Mellitus

OBJECTIVE

To estimate the effects of exercise during the first trimester on the risks of abnormal screening and gestational diabetes mellitus (GDM).

RESEARCH DESIGN AND METHODS

Data come from PETALS, a prospectively followed pregnancy cohort (n = 2,246, 79% minorities) receiving care at Kaiser Permanente Northern California. A Pregnancy Physical Activity Questionnaire was used to assess exercise. Glucose testing results for screening and diagnostic tests were obtained from electronic health records. Inverse probability of treatment weighting and targeted maximum likelihood with data-adaptive estimation (machine learning) of propensity scores and outcome regressions were used to obtain causal risk differences adjusted for potential confounders, including prepregnancy BMI, exercise before pregnancy, and gestational weight gain. Exercise was dichotomized at 1) the cohort's 75th percentile for moderate- to vigorous-intensity exercise (≥13.2 MET-h per week or ≥264 min per week of moderate exercise), 2) current recommendations (≥7.5 MET-h per week or ≥150 min per week of moderate exercise), and 3) any vigorous exercise.

RESULTS

Overall, 24.3% and 6.5% had abnormal screening and GDM, respectively. Exercise meeting or exceeding the 75th percentile decreased the risks of abnormal screening and GDM by 4.8 (95% CI 1.1, 8.5) and 2.1 (0.2, 4.1) fewer cases per 100, respectively, in adjusted analyses.

CONCLUSIONS

Exercise reduces the risks of abnormal screening and GDM, but the amount needed to achieve these risk reductions is likely higher than current recommendations. Future interventions may consider promoting ≥38 min per day of moderate-intensity exercise to prevent GDM.

Gestational weight gain and adverse pregnancy outcomes by pre-pregnancy BMI category in women with chronic hypertension: A cohort study

OBJECTIVES

It is important to understand relationships of gestational weight gain with adverse pregnancy outcomes in women with chronic hypertension, given their high baseline risk of adverse outcomes. We assessed associations of gestational weight gain with adverse pregnancy outcomes in women with chronic hypertension by pre-pregnancy body mass index categories.

STUDY DESIGN

We identified 14,369 women with chronic hypertension using electronic health records from 3 integrated health care delivery systems (2005-2014). Gestational weight gain-for-gestational age charts were used to calculate gestational weight gain z-scores, which account for gestational age. Modified Poisson regression models using generalized estimating equations were used to calculate relative risks and 95% confidence intervals, adjusted for sociodemographic and medical characteristics.

MAIN OUTCOME MEASUREMENTS

Preeclampsia, preterm delivery, cesarean delivery, neonatal intensive care unit admission, birthweight (extracted from the electronic health record).

RESULTS

In women with normal weight or overweight, low gestational weight gain (z-score < -1) was associated with 27-28% greater risk of preterm delivery and 48-82% greater risk of small-for-gestational age birthweight, while high gestational weight gain (z-score > 1) was associated with 40-90% greater risk of preeclampsia and 59-113% greater risk of large-for-gestational age birthweight. In women with obesity, low gestational weight gain was associated with 27-54% lower risk of several adverse pregnancy outcomes, including preeclampsia and cesarean delivery.

CONCLUSIONS

In women with chronic hypertension and normal weight or overweight, moderate gestational weight gain may confer the lowest risk of adverse outcomes. In women with chronic hypertension and obesity, low gestational weight gain may be necessary for the lowest risk of adverse pregnancy outcomes.

Inpatient-outpatient shared electronic health records: telemedicine and laboratory follow-up after hospital discharge

OBJECTIVES

Continuity of patient information across settings can improve transitions after hospital discharge, but outpatient clinicians often have limited access to complete information from recent hospitalizations. We examined whether providers' timely access to clinical information through shared inpatient-outpatient electronic health records (EHRs) was associated with follow-up visits, return emergency department (ED) visits, or readmissions after hospital discharge in patients with diabetes.

STUDY DESIGN

Stepped-wedge observational study.

METHODS

As an integrated delivery system staggered implementation of a shared inpatient-outpatient EHR, we studied 241,510 hospital discharges in patients with diabetes (2005-2011), examining rates of outpatient follow-up office visits, telemedicine (phone visits and asynchronous secure messages), laboratory tests, and return ED visits or readmissions (as adverse events). We used multivariate logistic regression adjusting for time trends, patient characteristics, and medical center and accounting for patient clustering to calculate adjusted follow-up rates.

RESULTS

For patients with diabetes, provider use of a shared inpatient-outpatient EHR was associated with a statistically significant shift toward follow-up delivered through a combination of telemedicine and outpatient laboratory tests, without a traditional in-person visit (from 22.9% with an outpatient-only EHR to 27.0% with a shared inpatient-outpatient EHR; P < .05). We found no statistically significant differences in 30-day return ED visits (odds ratio, 1.02; 95% CI, 0.96-1.09) or readmissions (odds ratio, 0.98; 95% CI, 0.91-1.06) with the shared EHR compared with the outpatient-only EHR.

CONCLUSIONS

Real-time clinical information availability during transitions between health care settings, along with robust telemedicine access, may shift the method of care delivery without adversely affecting patient health outcomes. Efforts to expand interoperability and information exchange may support follow-up care efficiency.

Onset of cardiovascular disease risk factors in women with and without a history of breast cancer: The Pathways Heart Study

12017

Background: Women with a history of breast cancer (BC) are at increased long-term risk of dying from cardiovascular disease (CVD). However, the onset of CVD risk factors in women with BC has not been well-described. We compared risk of incident CVD risk factors in women with and without BC enrolled in the Kaiser Permanente Northern California (KPNC) integrated health system. Methods: Data were extracted from KPNC electronic health records. All invasive BC cases diagnosed between 2005-2013 were identified and matched 1:5 with controls on birth year, race/ethnicity and KPNC membership at the date of BC diagnosis. Cox regression models assessed the hazard of incident hypertension (based on diagnosis codes and filled prescriptions), dyslipidemia (based on diagnosis codes, filled prescriptions, and lab values), and diabetes (KPNC Diabetes Registry). Models were adjusted for baseline BMI, menopausal status, smoking status, neighborhood median household income, education, prevalent CVD conditions, and other baseline CVD risk factors. Subgroups of women who received chemotherapy, radiation therapy, and endocrine therapy were compared with controls. Results: A total of 14,942 women with a new diagnosis of invasive BC were identified and matched to 74,702 controls. On average, women were 62.0 years, 28.3 kg/m2BMI, 64.9% non-Hispanic white. Overall, cases were more likely to develop hypertension (HR: 1.18, 95% CI: 1.13, 1.24) and diabetes (HR: 1.23, 95% CI: 1.16, 1.31). Across the board, receipt of any of the three therapies (chemotherapy, radiation therapy and endocrine therapy) was associated with increased risk of hypertension and diabetes, compared to controls. Risk-factor specific hazard ratios for receipt of chemotherapy were (HR 1.18, 95% CI: 1.10, 1.27) and (HR 1.38, 95% CI: 1.26, 1.51), for hypertension and diabetes, respectively. For receipt of radiation therapy, risk-factor specific hazard ratios were (HR: 1.17, 95% CI: 1.09, 1.26) and (HR: 1.15, 95% CI: 1.04, 1.27), for hypertension and diabetes, respectively. Risk-factor specific hazard ratios for receipt of endocrine therapy were (HR: 1.22, 95% CI: 1.14, 1.30) and (HR: 1.16, 95% CI: 1.06, 1.27), for hypertension and diabetes, respectively. Conclusions: The risk of developing hypertension and diabetes is increased in women with BC who received chemotherapy, radiation therapy, and/or endocrine therapy. Future studies should examine the roles of CVD risk factor diagnosis and management on cardiometabolic risk in women with a BC history.

Risk of cardiovascular disease in women with and without a history of breast cancer: The Pathways Heart Study

12016

Background: Breast cancer (BC) survivors are at increased risk of cardiovascular disease (CVD) following diagnosis, as compared to women without BC. To provide a population-based estimate of CVD risk in BC survivors, we compared risk of CVD events in women with and without BC history enrolled in the Kaiser Permanente Northern California (KPNC) integrated health system. Methods: Data were extracted from KPNC electronic health records. All invasive BC cases diagnosed between 2005-2013 were identified and matched 1:5 with non-BC controls on birth year, race/ethnicity and KPNC membership at date of BC diagnosis. Cox regression models were used to assess differences in the hazard of four major CVD events (ischemic heart disease (IHD), heart failure (HF), cardiomyopathy, and stroke). Models were adjusted for factors known to influence risk of breast cancer or CVD.Other CVD events included arrhythmia, cardiac arrest, carotid disease, myocarditis/pericarditis, transient ischemic attack, valvular disease, and venous thromboembolism (VTE). We additionally examined subgroups of cases who received chemotherapy, radiation, and endocrine therapy, and their controls. Results: A total of 14,942 women with a new diagnosis of invasive BC were identified and matched to 74,702 women without BC history. On average, women were 62.0 years, 28.3 kg/m2BMI, 64.9% non-Hispanic white. Among all cases and controls, there were no significant differences in hazard of developing IHD, cardiomyopathy, and stroke; there was a borderline difference in HF (HR: 1.08, 95% CI: 0.99, 1.19). Cases were more likely to have a cardiac arrest (HR: 1.39, 95% CI: 1.09, 1.78) and develop VTE (HR: 1.97, 95% CI: 1.74, 2.23). Women treated with chemotherapy were more likely than controls to develop HF (HR: 1.44, 95% CI: 1.21, 1.72), cardiomyopathy (HR: 2.01, 95% CI: 1.02, 3.98), and VTE (HR: 3.15, 95% CI: 2.62, 3.79). Women who received radiation therapy were more likely to develop carotid disease (HR: 5.49, 95% CI: 1.22, 24.66) and VTE (HR: 1.65, 95% CI: 1.35, 2.03) than controls. Women who received endocrine therapy were more likely to experience a cardiac arrest (HR: 1.49, 95% CI: 1.07, 2.09) and develop VTE (HR: 1.70, 95% CI: 1.42, 2.03) than controls. Conclusions: Women with BC were at increased risk of heart failure, cardiomyopathy, cardiac arrest, VTE and carotid disease. These risks varied by cancer treatment, with higher risk in those who received chemotherapy. Future studies should explore the effects of chemotherapy class and radiation dose exposure on diverse CVD endpoints in BC survivors.

Exploiting nonsystematic covariate monitoring to broaden the scope of evidence about the causal effects of adaptive treatment strategies

In studies based on electronic health records (EHR), the frequency of covariate monitoring can vary by covariate type, across patients, and over time, which can limit the generalizability of inferences about the effects of adaptive treatment strategies. In addition, monitoring is a health intervention in itself with costs and benefits, and stakeholders may be interested in the effect of monitoring when adopting adaptive treatment strategies. This paper demonstrates how to exploit nonsystematic covariate monitoring in EHR-based studies to both improve the generalizability of causal inferences and to evaluate the health impact of monitoring when evaluating adaptive treatment strategies. Using a real world, EHR-based, comparative effectiveness research (CER) study of patients with type II diabetes mellitus, we illustrate how the evaluation of joint dynamic treatment and static monitoring interventions can improve CER evidence and describe two alternate estimation approaches based on inverse probability weighting (IPW). First, we demonstrate the poor performance of the standard estimator of the effects of joint treatment-monitoring interventions, due to a large decrease in data support and concerns over finite-sample bias from near-violations of the positivity assumption (PA) for the monitoring process. Second, we detail an alternate IPW estimator using a no direct effect assumption. We demonstrate that this estimator can improve efficiency but at the potential cost of increase in bias from violations of the PA for the treatment process.

Determinants of Oral Bisphosphonate Use Beyond 5 Years

BACKGROUND

Few studies have examined factors that determine bisphosphonate (BP) continuation beyond 5 years in clinical practice.

OBJECTIVE

To investigate factors associated with BP continuation among women who completed 5 years of BP therapy.

METHODS

Women who received 5 consecutive years of oral BP treatment entered the cohort during 2002-2014 and were followed up to 5 additional years. Multivariable logistic regression was used to evaluate the association of demographic and clinical factors with adherent treatment continuation.

RESULTS

The cohort included 19,091 women with a median age of 72 years. Baseline and time-varying factors associated with increased odds of BP continuation after 5 years were (a) most recent bone mineral density (BMD) T-score -2 to -2.4 (OR = 1.31, 95% CI = 1.25-1.38), T-score -2.5 to -2.9 (OR = 1.48, 95% CI = 1.39-1.57), and T-score ≤ -3.0 (OR = 1.57, 95% CI = 1.47-1.68) versus T-scores above -2.0; (b) index date before 2008 (OR =1.35, 95% CI = 1.29-1.41); and (c) diabetes mellitus (OR = 1.08, 95% CI = 1.01-1.16). In contrast, factors associated with decreased odds of BP continuation were (a) recent hip (OR = 0.61, 95% CI = 0.52-0.71) or humerus (OR = 0.79, 95% CI = 0.66-0.94) fracture or fracture other than hip, wrist, spine, or humerus (OR = 0.90, 95% CI = 0.84-0.97); (b) Charlson Comorbidity Index score > 2 (OR = 0.91, 95% CI = 0.84-0.98); (c) history of rheumatoid arthritis (OR = 0.89, 95% CI = 0.80-0.99); (d) Hispanic (OR = 0.89, 95% CI=0.85-0.94) or Asian (OR = 0.90, 95% CI = 0.85-0.94) race/ethnicity; and (e) use of proton pump inhibitors (OR = 0.65, 95% CI = 0.59-0.71). Patient age and fracture before BP initiation were not associated with treatment continuation.

CONCLUSIONS

Clinical factors predicting continued BP treatment beyond 5 years include low BMD T-score, absence of recent fracture, and earlier era of treatment. Use of proton pump inhibitors was associated with lower likelihood of BP continuation. Other clinical and demographic factors were also noted to have variable effects on BP treatment continuation.

DISCLOSURES

This study was supported by a grant from the National Institute on Aging and National Institute of Arthritis, Musculoskeletal and Skin Diseases at the National Institutes of Health (NIH; R01AG047230, S1). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH or Kaiser Permanente. Lo has received previous research funding from Amgen and Sanofi, unrelated to the current study. Adams has received previous research funding from Merck, Amgen, Otsuka, and Radius Health, unrelated to the current study. Ettinger has served as an expert witness for Teva Pharmaceuticals, unrelated to the current study. Ott previously attended a scientific advisory meeting for Amgen but declined the honorarium. The other authors have nothing to disclose. These data were presented at the 2018 Annual Meeting of the American Society of Bone and Mineral Research (ASBMR), September 28-October 1, 2018, Montreal, Quebec, Canada.

Comparison of Mortality and Major Cardiovascular Events Among Adults With Type 2 Diabetes Using Human vs Analogue Insulins

Importance

The comparative cardiovascular safety of analogue and human insulins in adults with type 2 diabetes who initiate insulin therapy in usual care settings has not been carefully evaluated using machine learning and other rigorous analytic methods.

Objective

To examine the association of analogue vs human insulin use with mortality and major cardiovascular events.

Design, Setting, and Participants

This retrospective cohort study included 127 600 adults aged 21 to 89 years with type 2 diabetes at 4 health care delivery systems who initiated insulin therapy from January 1, 2000, through December 31, 2013. Machine learning and rigorous inference methods with time-varying exposures were used to evaluate associations of continuous exposure to analogue vs human insulins with mortality and major cardiovascular events. Data were analyzed from September 1, 2017, through June 30, 2018.

Exposures

On the index date (first insulin dispensing), participants were classified as using analogue insulin with or without human insulin or human insulin only.

Main Outcomes and Measures

Overall mortality, mortality due to cardiovascular disease (CVD), myocardial infarction (MI), stroke or cerebrovascular accident (CVA), and hospitalization for congestive heart failure (CHF) were evaluated. Marginal structural modeling (MSM) with inverse probability weighting was used to compare event-free survival in separate per-protocol analyses. Adjusted and unadjusted hazard ratios and cumulative risk differences were based on logistic MSM parameterizations for counterfactual hazards. Propensity scores were estimated using a data-adaptive approach (machine learning) based on 3 nested covariate adjustment sets. Sensitivity analyses were conducted to address potential residual confounding from unmeasured differences in risk factors across delivery systems.

Results

The 127 600 participants (mean [SD] age, 59.4 [12.6] years; 68 588 men [53.8%]; mean [SD] body mass index, 32.3 [7.1]) had a median follow-up of 4 quarters (interquartile range, 3-9 quarters) and experienced 5464 deaths overall (4.3%), 1729 MIs (1.4%), 1301 CVAs (1.0%), and 3082 CHF hospitalizations (2.4%). There were no differences in adjusted hazard ratios for continuous analogue vs human insulin exposure during 10 quarters for overall mortality (1.15; 95% CI, 0.97-1.34), CVD mortality (1.26; 95% CI, 0.86-1.66), MI (1.11; 95% CI, 0.77-1.45), CVA (1.30; 95% CI, 0.81-1.78), or CHF hospitalization (0.93; 95% CI, 0.75-1.11).

Conclusions and Relevance

Insulin-naive adults with type 2 diabetes who initiate and continue treatment with human vs analogue insulins had similar observed rates of major cardiovascular events, CVD mortality, and overall mortality.

Identifying hypertension in pregnancy using electronic medical records: The importance of blood pressure values

OBJECTIVE

To incorporate blood pressure (BP), diagnoses codes, and medication fills from electronic medical records (EMR) to identify pregnant women with hypertension.

STUDY DESIGN

A retrospective cohort study of singleton pregnancies at three US integrated health delivery systems during 2005-2014.

MAIN OUTCOME MEASURES

Women were considered hypertensive if they had any of the following: (1) ≥2 high BPs (≥140/90 mmHg) within 30 days during pregnancy (High BP); (2) an antihypertensive medication fill in the 120 days before pregnancy and a hypertension diagnosis from 1 year prior to pregnancy through 20 weeks gestation (Treated Chronic Hypertension); or (3) a high BP, a hypertension diagnosis, and a prescription fill within 7 days during pregnancy (Rapid Treatment). We described characteristics of these pregnancies and conducted medical record review to understand hypertension presence and severity.

RESULTS

Of 566,624 pregnancies, 27,049 (4.8%) met our hypertension case definition: 24,140 (89.2%) with High BP, 5,409 (20.0%) with Treated Chronic Hypertension, and 5,363 (19.8%) with Rapid Treatment (not mutually exclusive). Of hypertensive pregnancies, 19,298 (71.3%) received a diagnosis, 9,762 (36.1%) received treatment and 11,226 (41.5%) had a BP ≥ 160/110. In a random sample (n = 55) of the 7,559 pregnancies meeting the High BP criterion with no hypertension diagnosis, clinical statements about hypertension were found in medical records for 58% of them.

CONCLUSION

Incorporating EMR BP identified many pregnant women with hypertension who would have been missed by using diagnosis codes alone. Future studies should seek to incorporate BP to study treatment and outcomes of hypertension in pregnancy.

251Revacept, an inhibitor of platelet adhesion in symptomatic carotid stenosis: Results from a phase II study

Acknowledgement/Funding

None

Patients with complex chronic conditions: Health care use and clinical events associated with access to a patient portal

Background

For patients with diabetes, many with multiple complex chronic conditions, using a patient portal can support self-management and coordination of health care services, and may impact the frequency of in-person health care visits.

Objective

To examine the impact of portal access on the number of outpatient visits, emergency visits, and preventable hospitalizations.

Design

Observational study comparing patients’ visit rates with and without portal access, using marginal structural modeling with inverse probability weighting estimates to account for potential bias due to confounding and attrition.

Setting

Large integrated delivery system which implemented a patient portal (2006–2007).

Patients

We examined 165,447 patients with diabetes defined using clinical registries. Our study included both patients with diabetes-only and patients with multiple complex chronic conditions (diabetes plus asthma, congestive artery disease, congestive heart failure, or hypertension).

Measurements

We examined rates of outpatient office visits, emergency room visits, and preventable hospitalizations (for ambulatory care sensitive conditions).

Results

Access to a patient portal was associated with significantly higher rates of outpatient office visits, in both patients with diabetes only and in patients with multiple complex conditions (p<0.05). In patients with multiple complex chronic conditions, portal use was also associated with significantly fewer emergency room visits (3.9 fewer per 1,000 patients per month, p<0.05) and preventable hospital stays (0.8 fewer per 1,000 patients per month, p<0.05). In patients with only diabetes, the results were directionally consistent but not statistically significantly associated with emergency room visits and preventable hospital stays.

Limitations

Observational study in an integrated delivery system.

Conclusion

Access to a patient portal can increase engagement in outpatient visits, potentially addressing unmet clinical needs, and reduce downstream health events that lead to emergency and hospital care, particularly among patients with multiple complex conditions.

Using Pharmacy Data and Adherence to Define Long-Term Bisphosphonate Exposure in Women

BACKGROUND

Assigning drug exposure is a necessary first step in examining bisphosphonate (BP) treatment in observational studies using pharmacy data.

OBJECTIVE

To determine whether the choice of adherence level using the proportion of days covered (PDC) affected BP exposure assignment.

METHODS

10,381 female health plan members who initiated oral BP therapy between 2002 and 2010 and had received 5 consecutive years of treatment were identified and subsequently followed up to 5 additional years. In each 90-day interval of follow-up, a woman was considered "on treatment" if she received the drug for more than a predetermined PDC based on pharmacy days supply and "off treatment" if she received the drug for less than that PDC. Women who continued on therapy above the PDC threshold during follow-up were considered continuously on therapy. Women who were off treatment during the first 90-days of follow-up were classified as off therapy and were followed to determine if they remained continuously off treatment. This study evaluated the extent to which varying the PDC threshold (≥ 0.5, ≥ 0.6, and ≥ 0.7) affected the proportion of women classified as "continuously on" or "continuously off" BP during follow-up.

RESULTS

Under PDC thresholds of 0.5, 0.6, and 0.7, 48%, 43%, and 36% of women who remained on follow-up were categorized as continuously on treatment at year 2 of follow-up, and 18%, 14%, and 12% were categorized as continuously on treatment by the end of follow-up. Using these same PDC thresholds, 9%, 12%, and 15% of women were categorized as off therapy during the first quarter of follow-up and were highly likely to remain off therapy: 4%, 5%, and 5% were classified as continuously off therapy at year 2, and 4% of women were classified as such by the end of follow-up for all 3 thresholds.

CONCLUSIONS

A PDC of 0.6 was chosen as a practical threshold for drug adherence. Varying the PDC to 0.5 or 0.7 resulted in modest changes in the proportions of women considered continuously on BP therapy.

DISCLOSURES

This study was supported by a grant from the National Institute of Aging and National Institute of Arthritis, Musculoskeletal and Skin Diseases at the National Institutes of Health (R01AG047230, S1). Lo has received previous research funding from Amgen and Sanofi, outside of the current study. Chandra has received previous research funding from Amgen outside of the current study. Adams has received previous research funding from Merck, Amgen, Otsuka, and Radius Health, outside of the current study. Ott previously attended a scientific advisory meeting for Amgen but declined the honorarium. Ettinger previously served as an expert witness for Teva Pharmaceuticals.

Targeted learning with daily EHR data

Electronic health records (EHR) data provide a cost- and time-effective opportunity to conduct cohort studies of the effects of multiple time-point interventions in the diverse patient population found in real-world clinical settings. Because the computational cost of analyzing EHR data at daily (or more granular) scale can be quite high, a pragmatic approach has been to partition the follow-up into coarser intervals of pre-specified length (eg, quarterly or monthly intervals). The feasibility and practical impact of analyzing EHR data at a granular scale has not been previously evaluated. We start filling these gaps by leveraging large-scale EHR data from a diabetes study to develop a scalable targeted learning approach that allows analyses with small intervals. We then study the practical effects of selecting different coarsening intervals on inferences by reanalyzing data from the same large-scale pool of patients. Specifically, we map daily EHR data into four analytic datasets using 90-, 30-, 15-, and 5-day intervals. We apply a semiparametric and doubly robust estimation approach, the longitudinal Targeted Minimum Loss-Based Estimation (TMLE), to estimate the causal effects of four dynamic treatment rules with each dataset, and compare the resulting inferences. To overcome the computational challenges presented by the size of these data, we propose a novel TMLE implementation, the "long-format TMLE," and rely on the latest advances in scalable data-adaptive machine-learning software, xgboost and h2o, for estimation of the TMLE nuisance parameters.

Automated symptom and treatment side effect monitoring for improved quality of life among adults with diabetic peripheral neuropathy in primary care: a pragmatic, cluster, randomized, controlled trial

AIMS

To evaluate the effectiveness of automated symptom and side effect monitoring on quality of life among individuals with symptomatic diabetic peripheral neuropathy.

METHODS

We conducted a pragmatic, cluster randomized controlled trial (July 2014 to July 2016) within a large healthcare system. We randomized 1834 primary care physicians and prospectively recruited from their lists 1270 individuals with neuropathy who were newly prescribed medications for their symptoms. Intervention participants received automated telephone-based symptom and side effect monitoring with physician feedback over 6 months. The control group received usual care plus three non-interactive diabetes educational calls. Our primary outcomes were quality of life (EQ-5D) and select symptoms (e.g. pain) measured 4-8 weeks after starting medication and again 8 months after baseline. Process outcomes included receiving a clinically effective dose and communication between individuals with neuropathy and their primary care provider over 12 months. Interviewers collecting outcome data were blinded to intervention assignment.

RESULTS

Some 1252 participants completed the baseline measures [mean age (sd): 67 (11.7), 53% female, 57% white, 8% Asian, 13% black, 20% Hispanic]. In total, 1179 participants (93%) completed follow-up (619 control, 560 intervention). Quality of life scores (intervention: 0.658 ± 0.094; control: 0.653 ± 0.092) and symptom severity were similar at baseline. The intervention had no effect on primary [EQ-5D: -0.002 (95% CI -0.01, 0.01), P = 0.623; pain: 0.295 (-0.75, 1.34), P = 0.579; sleep disruption: 0.342 (-0.18, 0.86), P = 0.196; lower extremity functioning: -0.079 (-1.27, 1.11), P = 0.896; depression: -0.462 (-1.24, 0.32); P = 0.247] or process outcomes.

CONCLUSIONS

Automated telephone monitoring and feedback alone were not effective at improving quality of life or symptoms for people with symptomatic diabetic peripheral neuropathy.

TRIAL REGISTRATION

ClinicalTrials.gov (NCT02056431).

Comparative effectiveness of treatment of actinic keratosis with topical fluorouracil and imiquimod in the prevention of keratinocyte carcinoma: A cohort study

BACKGROUND

The effectiveness of 5-fluorouracil compared with that of imiquimod for preventing keratinocyte carcinoma is unknown.

OBJECTIVE

To compare the effectiveness of 5-fluorouracil and that of imiquimod in preventing keratinocyte carcinoma in a real-world practice setting.

METHODS

We identified 5700 subjects who filled prescriptions for 5-fluorouracil or imiquimod for treatment of actinic keratosis in 2007. An intention-to-treat analysis controlling for potential confounding variables was used to calculate 2- and 5-year cumulative risk differences for subsequent keratinocyte carcinoma overall and in field-treated areas.

RESULTS

5-Fluorouracil was associated with a statistically significant decreased risk of any keratinocyte carcinoma compared with imiquimod (adjusted hazard ratio [aHR], 0.86; 95% confidence interval [CI], 0.76-0.97), but there were no significant differences in risk by tumor subtype (for squamous cell carcinoma: aHR, 0.89; 95% CI, 0.74-1.07; for basal cell carcinoma: aHR, 0.87; 95% CI, 0.74-1.03) or site-specific keratinocyte carcinoma (aHR, 0.96; 95% CI, 0.81-1.14). There were no significant differences in 2- or 5-year cumulative risk of keratinocyte carcinoma among those treated with 5-fluorouracil versus with imiquimod.

LIMITATIONS

Generalizability to other practice settings may be limited.

CONCLUSIONS

Whereas 5-fluorouracil was more effective in reducing keratinocyte carcinoma risk overall, we found no differences in the short- or long-term risk of subsequent site-specific keratinocyte carcinoma in a real-world practice setting.

Comparative effectiveness of treatment of actinic keratosis with topical fluorouracil and imiquimod in the prevention of keratinocyte carcinoma: A cohort study

BACKGROUND

The effectiveness of 5-fluorouracil compared with that of imiquimod for preventing keratinocyte carcinoma is unknown.

OBJECTIVE

To compare the effectiveness of 5-fluorouracil and that of imiquimod in preventing keratinocyte carcinoma in a real-world practice setting.

METHODS

We identified 5700 subjects who filled prescriptions for 5-fluorouracil or imiquimod for treatment of actinic keratosis in 2007. An intention-to-treat analysis controlling for potential confounding variables was used to calculate 2- and 5-year cumulative risk differences for subsequent keratinocyte carcinoma overall and in field-treated areas.

RESULTS

5-Fluorouracil was associated with a statistically significant decreased risk of any keratinocyte carcinoma compared with imiquimod (adjusted hazard ratio [aHR], 0.86; 95% confidence interval [CI], 0.76-0.97), but there were no significant differences in risk by tumor subtype (for squamous cell carcinoma: aHR, 0.89; 95% CI, 0.74-1.07; for basal cell carcinoma: aHR, 0.87; 95% CI, 0.74-1.03) or site-specific keratinocyte carcinoma (aHR, 0.96; 95% CI, 0.81-1.14). There were no significant differences in 2- or 5-year cumulative risk of keratinocyte carcinoma among those treated with 5-fluorouracil versus with imiquimod.

LIMITATIONS

Generalizability to other practice settings may be limited.

CONCLUSIONS

Whereas 5-fluorouracil was more effective in reducing keratinocyte carcinoma risk overall, we found no differences in the short- or long-term risk of subsequent site-specific keratinocyte carcinoma in a real-world practice setting.

Abstract 119: Using Blood Pressure Values from Electronic Medical Records to Identify Hypertension During Pregnancy

Objective: Hypertension is a major risk factor for poor pregnancy outcomes. Many observational studies have relied on diagnosis codes, particularly from the delivery hospitalization, to identify hypertension in pregnancy. We augmented diagnosis codes with electronic blood pressure (BP) data to improve the identification of pregnant women with hypertension.

Methods: We studied pregnant women aged 15-49 years enrolled in three Kaiser Permanente health plans who delivered during 2005-2014. Using diagnosis codes, BP values, and antihypertensive medication dispensings, we defined hypertension as: (1) ≥ 2 high BPs (≥ 140/90 mmHg) within 30 days of each other (2highBPs); or (2) ≥ 1 antihypertensive medication fill with ≥1 hypertension diagnosis code from 120 days prior to pregnancy through 20 weeks gestation (chronicHTN); or (3) ≥ 1 high BP, a hypertension diagnosis code, and an antihypertensive fill within 7 days (RapidTx). Among women meeting our study definition, we examined receipt of hypertension diagnosis codes and prevalence of severe hypertension (1+ BP ≥160/110 mmHg).

Results: Among 553,477 eligible women, 29,933 (5%) met our definition of hypertension, including 26,855 identified via 2highBPs, 5,774 via chronicHTN and 6,198 via RapidTx (not mutually exclusive). Among women meeting our hypertension definition overall, only 64% had 1+ hypertension diagnosis code assigned during pregnancy, and 49% had one at delivery. Among hypertensive women identified via 2highBPs, only 60% (16,057/26,855) had a hypertension diagnosis code in pregnancy and 45% (12,131/ 26,855) at delivery. However, 53% of our hypertensive women (14,972/ 29,933) overall and 56% (14,972/ 26,855) of the 2highBPs women had severe hypertension at some time during pregnancy.

Conclusion: Incorporating BP values identifies additional pregnant women with hypertension who would have been missed by approaches using diagnosis codes alone. Women identified by our method frequently had severely elevated BP, showing the importance of including these women in future studies of hypertension during pregnancy.

simcausal R Package: Conducting Transparent and Reproducible Simulation Studies of Causal Effect Estimation with Complex Longitudinal Data

The simcausal R package is a tool for specification and simulation of complex longitudinal data structures that are based on non-parametric structural equation models. The package aims to provide a flexible tool for simplifying the conduct of transparent and reproducible simulation studies, with a particular emphasis on the types of data and interventions frequently encountered in real-world causal inference problems, such as, observational data with time-dependent confounding, selection bias, and random monitoring processes. The package interface allows for concise expression of complex functional dependencies between a large number of nodes, where each node may represent a measurement at a specific time point. The package allows for specification and simulation of counterfactual data under various user-specified interventions (e.g., static, dynamic, deterministic, or stochastic). In particular, the interventions may represent exposures to treatment regimens, the occurrence or non-occurrence of right-censoring events, or of clinical monitoring events. Finally, the package enables the computation of a selected set of user-specified features of the distribution of the counterfactual data that represent common causal quantities of interest, such as, treatment-specific means, the average treatment effects and coefficients from working marginal structural models. The applicability of simcausal is demonstrated by replicating the results of two published simulation studies.

Adjuvant gemcitabine-docetaxel chemotherapy for stage I uterine leiomyosarcoma: Trends and survival outcomes

OBJECTIVE

To assess recent trends of administering adjuvant gemcitabine-docetaxel (GD) chemotherapy for Stage I uterine leiomyosarcoma, and to compare disease-free and overall survival between women who received and did not receive adjuvant GD chemotherapy.

METHODS

All patients diagnosed with Stage I uterine leiomyosarcoma in a California-Colorado population-based health plan inclusive of 2006-2013 were included in a retrospective cohort. Adjuvant GD chemotherapy rates, clinico-pathologic characteristics and survival estimates were assessed.

RESULTS

Of 111 women with Stage I uterine leiomyosarcoma, 33 received adjuvant GD (median 4cycles), 77 received no chemotherapy, and 1 patient excluded for non-GD chemotherapy. GD-chemotherapy and no-chemotherapy groups were similar with respect to age, stage (IA/IB), uterine weight, mitotic index, body mass index, and Charlson comorbidity score. Non-Hispanic white women were twice as likely to receive adjuvant chemotherapy as non-white or Hispanic women (37.7 vs. 17.1%, P=0.02). The proportion of women receiving adjuvant GD chemotherapy increased from 6.5% in 2006-2008 to 46.9% in 2009-2013 (P<0.001). There was no significance difference in unadjusted Kaplan-Meyer estimated disease-free (P=0.95) or overall survival (P=0.43) between GD-chemotherapy and no-chemotherapy cohorts. Corresponding adjusted Cox proportional hazard ratios for adjuvant GD chemotherapy compared to no chemotherapy were 1.01 (95% confidence interval [CI] 0.57-1.80, P=0.97) for recurrence and 1.28 (95% CI 0.69-2.36, P-0.48) for mortality.

CONCLUSIONS

Use of adjuvant GD chemotherapy for Stage I uterine leiomyosarcoma has increased significantly in the last decade, despite unclear benefit. Compared to no chemotherapy, 4-6cycles of adjuvant GD chemotherapy does not appear to alter survival outcomes.

Identification of the joint effect of a dynamic treatment intervention and a stochastic monitoring intervention under the no direct effect assumption

The management of chronic conditions is characterized by frequent re-assessment of therapy decisions in response to the patient's changing condition over the course of the illness. Evidence most suitable to inform care thus often concerns the contrast of adaptive treatment strategies that repeatedly personalize treatment decisions over time using the latest accumulated data available from the patient's previous clinic visits such as laboratory exams (e.g., hemoglobin A1c measurements in diabetes care). The frequency at which such information is monitored implicitly defines the causal estimand that is typically evaluated in an observational or randomized study of such adaptive treatment strategies. Analytic control of monitoring with standard estimation approaches for time-varying interventions can therefore not only improve study generalizibility but also inform the optimal timing of clinical surveillance. Valid inference with these estimators requires the upholding of a positivity assumption that can hinder their applicability. To potentially weaken this requirement for monitoring control, we introduce identifiability results that will facilitate the derivation of alternate estimators of effects defined by general joint treatment and monitoring interventions in the context of time-to-event outcomes. These results are developed based on the nonparametric structural equation modeling framework using a no direct effect assumption originally introduced in a prior paper that inspired this work. The relevance and scope of the results presented here are illustrated with examples in diabetes comparative effectiveness research.

Analysis of Body Mass Index and Mortality in Patients With Colorectal Cancer Using Causal Diagrams

IMPORTANCE

Physicians and investigators have sought to determine the relationship between body mass index (BMI [calculated as weight in kilograms divided by height in meters squared]) and colorectal cancer (CRC) outcomes, but methodologic limitations including sampling selection bias, reverse causality, and collider bias have prevented the ability to draw definitive conclusions.

OBJECTIVE

To evaluate the association of BMI at the time of, and following, colorectal cancer (CRC) diagnosis with mortality in a complete population using causal diagrams.

DESIGN, SETTING, AND PARTICIPANTS

This retrospective observational study with prospectively collected data included a cohort of 3408 men and women, ages 18 to 80 years, from the Kaiser Permanente Northern California population, who were diagnosed with stage I to III CRC between 2006 and 2011 and who also had surgery.

EXPOSURES

Body mass index at diagnosis and 15 months following diagnosis.

MAIN OUTCOMES AND MEASURES

Hazard ratios (HRs) for all-cause mortality and CRC-specific mortality compared with normal-weight patients, adjusted for sociodemographics, disease severity, treatment, and prediagnosis BMI.

RESULTS

This study investigated a cohort of 3408 men and women ages 18 to 80 years diagnosed with stage I to III CRC between 2006 and 2011 who also had surgery. At-diagnosis BMI was associated with all-cause mortality in a nonlinear fashion, with patients who were underweight (BMI <18.5; HR, 2.65; 95% CI, 1.63-4.31) and patients who were class II or III obese (BMI ≥35; HR, 1.33; 95% CI, 0.89-1.98) exhibiting elevated mortality risks, compared with patients who were low-normal weight (BMI 18.5 to <23). In contrast, patients who were high-normal weight (BMI 23 to <25; HR, 0.77; 95% CI, 0.56-1.06), low-overweight (BMI 25 to <28; HR, 0.75; 95% CI, 0.55-1.04), and high-overweight (BMI 28 to <30; HR, 0.52; 95% CI, 0.35-0.77) had lower mortality risks, and patients who were class I obese (BMI 30 to <35) showed no difference in risk. Spline analysis confirmed a U-shaped relationship in participants with lowest mortality at a BMI of 28. Associations with CRC-specific mortality were similar. Associations of postdiagnosis BMI and mortality were also similar, but patients who were class I obese had significantly lower all-cause and cancer-specific mortality risks.

CONCLUSIONS AND RELEVANCE

In this study, body mass index at the time of diagnosis and following diagnosis of CRC was associated with mortality risk. Though evidence shows that exercise in patients with cancer should be encouraged, findings suggest that recommendations for weight loss in the immediate postdiagnosis period among patients with CRC who are overweight may be unwarranted.

The Diabetes Telephone Study: Design and challenges of a pragmatic cluster randomized trial to improve diabetic peripheral neuropathy treatment

BACKGROUND

Challenges to effective pharmacologic management of symptomatic diabetic peripheral neuropathy include the limited effectiveness of available medicines, frequent side effects, and the need for ongoing symptom assessment and treatment titration for maximal effectiveness. We present here the rationale and implementation challenges of the Diabetes Telephone Study, a randomized trial designed to improve medication treatment, titration, and quality of life among patients with symptomatic diabetic peripheral neuropathy.

METHODS

We implemented a pragmatic cluster randomized controlled trial to test the effectiveness of an automated interactive voice response tool designed to provide physicians with real-time patient-reported data about responses to newly prescribed diabetic peripheral neuropathy medicines. A total of 1834 primary care physicians treating patients in the diabetes registry at Kaiser Permanente Northern California were randomized into the intervention or control arm. In September 2014, we began identification and recruitment of patients assigned to physicians in the intervention group who receive three brief interactive calls every 2 months after a medication is prescribed to alleviate diabetic peripheral neuropathy symptoms. These calls provide patients with the opportunity to report on symptoms, side effects, self-titration of medication dose and overall satisfaction with treatment. We plan to compare changes in self-reported quality of life between the intervention group and patients in the control group who receive three non-interactive automated educational phone calls.

RESULTS

Successful implementation of this clinical trial required robust stakeholder engagement to help tailor the intervention and to address pragmatic concerns such as provider time constraints. As of 27 October 2015, we had screened 2078 patients, 1447 of whom were eligible for participation. We consented and enrolled 1206 or 83% of those eligible. Among those enrolled, 53% are women and the mean age is 67 (standard deviation = 12) years. The racial ethnic make-up is 56% White, 8% Asian, 13% Black or African American, and 19% Hispanic or Latino.

CONCLUSION

Innovative strategies are needed to guide improvements in healthcare delivery for patients with symptomatic diabetic peripheral neuropathy. This trial aims to assess whether real-time collection and clinical feedback of patient treatment experiences can reduce patient symptom burden. Implementation of a clinical trial closely involving clinical care required researchers to partner with clinicians. If successful, this intervention provides a critical information feedback loop that would optimize diabetic peripheral neuropathy medication titration through widely available interactive voice response technology.

A Case Study of the Impact of Data-Adaptive Versus Model-Based Estimation of the Propensity Scores on Causal Inferences from Three Inverse Probability Weighting Estimators

OBJECTIVE

Consistent estimation of causal effects with inverse probability weighting estimators is known to rely on consistent estimation of propensity scores. To alleviate the bias expected from incorrect model specification for these nuisance parameters in observational studies, data-adaptive estimation and in particular an ensemble learning approach known as Super Learning has been proposed as an alternative to the common practice of estimation based on arbitrary model specification. While the theoretical arguments against the use of the latter haphazard estimation strategy are evident, the extent to which data-adaptive estimation can improve inferences in practice is not. Some practitioners may view bias concerns over arbitrary parametric assumptions as academic considerations that are inconsequential in practice. They may also be wary of data-adaptive estimation of the propensity scores for fear of greatly increasing estimation variability due to extreme weight values. With this report, we aim to contribute to the understanding of the potential practical consequences of the choice of estimation strategy for the propensity scores in real-world comparative effectiveness research.

METHOD

We implement secondary analyses of Electronic Health Record data from a large cohort of type 2 diabetes patients to evaluate the effects of four adaptive treatment intensification strategies for glucose control (dynamic treatment regimens) on subsequent development or progression of urinary albumin excretion. Three Inverse Probability Weighting estimators are implemented using both model-based and data-adaptive estimation strategies for the propensity scores. Their practical performances for proper confounding and selection bias adjustment are compared and evaluated against results from previous randomized experiments.

CONCLUSION

Results suggest both potential reduction in bias and increase in efficiency at the cost of an increase in computing time when using Super Learning to implement Inverse Probability Weighting estimators to draw causal inferences.

Adherence to Metformin, Statins, and ACE/ARBs Within the Diabetes Health Plan (DHP)

BACKGROUND

Reducing patient cost-sharing and engaging patients in disease management activities have been shown to increase uptake of evidence-based care.

OBJECTIVE

To evaluate the effect of employer purchase of a disease-specific plan with reduced cost-sharing and disease management (the Diabetes Health Plan/DHP) on medication adherence among eligible employees and dependents.

DESIGN

Employer-level "intent to treat" cohort study, including data from eligible employees and their dependents with diabetes, regardless of whether they were enrolled in the DHP.

SETTING

Employers that contracted with a large national health plan administrator in 2009, 2010, and/or 2011.

PARTICIPANTS

Ten employers that purchased the DHP and 191 employers that did not (controls). Inverse probability weighting (IPW) estimation was used to adjust for inter-group differences.

INTERVENTION

The DHP includes free or low-cost medications and physician visits. Enrollment strategies and specific benefit designs are determined by the employer and vary in practice. DHP participants are notified up front that they must engage in their own health care (e.g., receiving diabetes-related screening) in order to remain enrolled.

MAIN OUTCOME MEASURE

Mean employee adherence to metformin, statins, and ACE/ARBs at the employer level at one year post-DHP implementation, as measured by the proportion of days covered (PDC).

RESULTS

Baseline adherence to the three medications was similar across DHP and control employers, ranging from 64 to 69 %. In the first year after DHP implementation, predicted employer-level adherence for metformin (+4.9 percentage points, p = 0.017), statins (+4.8, p = 0.019), and ACE/ARBs (+4.4, p = 0.02) was higher with DHP purchase.

LIMITATIONS

Non-randomized, observational study.

CONCLUSIONS

The Diabetes Health Plan, an innovative health plan that combines reduced cost-sharing and disease management with an up-front requirement of enrollee participation in his or her own health care, is associated with a modest improvement in medication adherence at 12 months.

High-dimensional propensity score algorithm in comparative effectiveness research with time-varying interventions

The high-dimensional propensity score (hdPS) algorithm was proposed for automation of confounding adjustment in problems involving large healthcare databases. It has been evaluated in comparative effectiveness research (CER) with point treatments to handle baseline confounding through matching or covariance adjustment on the hdPS. In observational studies with time-varying interventions, such hdPS approaches are often inadequate to handle time-dependent confounding and selection bias. Inverse probability weighting (IPW) estimation to fit marginal structural models can adequately handle these biases under the fundamental assumption of no unmeasured confounders. Upholding of this assumption relies on the selection of an adequate set of covariates for bias adjustment. We describe the application and performance of the hdPS algorithm to improve covariate selection in CER with time-varying interventions based on IPW estimation and explore stabilization of the resulting estimates using Super Learning. The evaluation is based on both the analysis of electronic health records data in a real-world CER study of adults with type 2 diabetes and a simulation study. This report (i) establishes the feasibility of IPW estimation with the hdPS algorithm based on large electronic health records databases, (ii) demonstrates little impact on inferences when supplementing the set of expert-selected covariates using the hdPS algorithm in a setting with extensive background knowledge, (iii) supports the application of the hdPS algorithm in discovery settings with little background knowledge or limited data availability, and (iv) motivates the application of Super Learning to stabilize effect estimates based on the hdPS algorithm.

Racial/ethnic differences in hip and diaphyseal femur fractures

Contemporary femur fracture rates were examined in northern California women and compared by race/ethnicity. During 2006-2012, hip fracture rates declined, but diaphyseal fracture rates increased, especially in Asians. Women with diaphyseal fracture were younger and more likely to be bisphosphonate-treated. These disparities in femur fracture should be further examined.

INTRODUCTION

The epidemiology of diaphyseal femur fracture differs from proximal femur (hip) fracture, although few studies have examined demographic variations in the current era. This study examines contemporary differences in low-energy femur fracture by race/ethnicity in a large, diverse integrated health-care delivery system.

METHODS

The incidence of hip and diaphyseal fracture in northern California women aged ≥50 years old during 2006-2012 was examined. Hip (femoral neck and pertrochanteric) fractures were classified by hospital diagnosis codes, while diaphyseal (subtrochanteric and femoral shaft) fractures were further adjudicated based on radiologic findings. Demographic and clinical data were obtained from health plan databases. Fracture incidence was examined over time and by race/ethnicity.

RESULTS

There were 10,648 (97.3 %) hip and 300 (2.7 %) diaphyseal fractures among 10,493 women. The age-adjusted incidence of hip fracture fell from 281 to 240 per 100,000 women and was highest for white women. However, diaphyseal fracture rates increased over time, with a significant upward trend in Asians (9 to 27 per 100,000) who also had the highest rate of diaphyseal fracture. Women with diaphyseal fracture were younger than women with hip fracture, more likely to be of Asian race and to have received bisphosphonate drugs. Women with longer bisphosphonate treatment duration were also more likely to have a diaphyseal fracture, especially younger Asian women.

CONCLUSION

During 2006 to 2012, hip fracture rates declined, but diaphyseal fracture rates increased, particularly among Asian women. The association of diaphyseal fracture and bisphosphonate therapy should be further investigated with examination of fracture pattern.

Targeted learning in real-world comparative effectiveness research with time-varying interventions

In comparative effectiveness research (CER), often the aim is to contrast survival outcomes between exposure groups defined by time-varying interventions. With observational data, standard regression analyses (e.g., Cox modeling) cannot account for time-dependent confounders on causal pathways between exposures and outcome nor for time-dependent selection bias that may arise from informative right censoring. Inverse probability weighting (IPW) estimation to fit marginal structural models (MSMs) has commonly been applied to properly adjust for these expected sources of bias in real-world observational studies. We describe the application and performance of an alternate estimation approach in such a study. The approach is based on the recently proposed targeted learning methodology and consists in targeted minimum loss-based estimation (TMLE) with super learning (SL) within a nonparametric MSM. The evaluation is based on the analysis of electronic health record data with both IPW estimation and TMLE to contrast cumulative risks under four more or less aggressive strategies for treatment intensification in adults with type 2 diabetes already on 2+ oral agents or basal insulin. Results from randomized experiments provide a surrogate gold standard to validate confounding and selection bias adjustment. Bootstrapping is used to validate analytic estimation of standard errors. This application does the following: (1) establishes the feasibility of TMLE in real-world CER based on large healthcare databases; (2) provides evidence of proper confounding and selection bias adjustment with TMLE and SL; and (3) motivates their application for improving estimation efficiency. Claims are reinforced with a simulation study that also illustrates the double-robustness property of TMLE.

Impact of specific glucose-control strategies on microvascular and macrovascular outcomes in 58,000 adults with type 2 diabetes

OBJECTIVE

Comparative effectiveness research methods are used to compare the effect of four distinct glucose-control strategies on subsequent myocardial infarction and nephropathy in type 2 diabetes.

RESEARCH DESIGN AND METHODS

A total of 58,000 adults with type 2 diabetes and A1C <7% (53 mmol/mol) while taking two or more oral agents or basal insulin had subsequent A1C ≥7% (53 mmol/mol) to 8.5% (69 mmol/mol). Follow-up started on date of first A1C ≥7% and ended on date of a specific clinical event, death, disenrollment, or study end. Glucose-control strategies were defined as first intensification of glucose-lowering therapy at A1C ≥7, ≥7.5, ≥8, or ≥8.5% with subsequent control for treatment adherence. Logistic marginal structural models were fitted to assess the discrete-time hazards for each dynamic glucose-control strategy, adjusting for baseline and time-dependent confounding and selection bias through inverse probability weighting.

RESULTS

After adjustment for age, sex, race/ethnicity, comorbidities, blood pressure, lipids, BMI, and other covariates, progressively more aggressive glucose-control strategies were associated with reduced onset or progression of albuminuria but not associated with significant reduction in occurrence of myocardial infarction or preserved renal function based on estimated glomerular filtration rate over 4 years of follow-up.

CONCLUSIONS

In a large representative cohort of adults with type 2 diabetes, more aggressive glucose-control strategies have mixed short-term effects on microvascular complications and do not reduce the myocardial infarction rate over 4 years of follow-up. These findings are consistent with the results of recent clinical trials, but confirmation over longer periods of observation is needed.

Super learning to hedge against incorrect inference from arbitrary parametric assumptions in marginal structural modeling

OBJECTIVE

Clinical trials are unlikely to ever be launched for many comparative effectiveness research (CER) questions. Inferences from hypothetical randomized trials may however be emulated with marginal structural modeling (MSM) using observational data, but success in adjusting for time-dependent confounding and selection bias typically relies on parametric modeling assumptions. If these assumptions are violated, inferences from MSM may be inaccurate. In this article, we motivate the application of a data-adaptive estimation approach called super learning (SL) to avoid reliance on arbitrary parametric assumptions in CER.

STUDY DESIGN AND SETTING

Using the electronic health records data from adults with new-onset type 2 diabetes, we implemented MSM with inverse probability weighting (IPW) estimation to evaluate the effect of three oral antidiabetic therapies on the worsening of glomerular filtration rate.

RESULTS

Inferences from IPW estimation were noticeably sensitive to the parametric assumptions about the associations between both the exposure and censoring processes and the main suspected source of confounding, that is, time-dependent measurements of hemoglobin A1c. SL was successfully implemented to harness flexible confounding and selection bias adjustment from existing machine learning algorithms.

CONCLUSION

Erroneous IPW inference about clinical effectiveness because of arbitrary and incorrect modeling decisions may be avoided with SL.