Annals of Internal Medicine — New This Month, June 23

The practice-relevant headline is a negative trial: adding CBT or a self-management program to patient-centered opioid tapering did not improve taper success — the taper's structure is what works. The month's best reference read is the In the Clinic on obstructive sleep apnea, which explicitly moves past the 2014 ACP in-lab recommendation toward home sleep apnea testing. Two independent cohorts revisited the GLP-1 / ischemic optic neuropathy question and both landed in the same place — a small absolute excess that largely dissolves under tighter confounding control. No Beyond the Guidelines case ran this month.

General IMRCTHigh yield · IM

Patient-Centered Prescription Opioid Tapering Methods

In adults on long-term opioids, does adding pain-CBT or a self-management program to a patient-centered taper vs taper alone improve taper success?

Population: 562 adults at 11 U.S. sites with pain ≥6 months on a morphine equivalent daily dose (MEDD) ≥10 for ≥3 months, without moderate/severe opioid use disorder (191 / 203 / 168). July 2018 – November 2023.
Intervention: Patient-centered taper with close monitoring and electronic supports, plus pain-CBT or a chronic pain self-management program (CPSMP).
Comparison: Patient-centered taper only.
Outcome: Taper success at 12 mo: 50.9% (95% CI 42.9–58.9) taper only, 48.6% (41.0–56.2) +pain-CBT, 44.5% (36.0–53.3) +CPSMP. Neither adjunct helped (−2.4 pp [CI −11.9 to 7.2]; −5.2 pp [CI −15.3 to 4.8]). Study-related adverse events were highest in the taper-only arm: 66% vs 54% (+CBT) and 64% (+CPSMP).

Clinical takeaway. About half of appropriately selected patients succeed with a patient-centered taper alone — the taper's structure (voluntary, slow, closely monitored, pausable) is what works, not a bolted-on behavioral program. CBT's value here looks like tolerability (fewer withdrawal-related adverse events), not success. Sullivan's editorial stresses that "patient-centered" means patient control and the ability to stop or slow — this trial does not license involuntary dose reduction. COVID-19 shrank and unbalanced the arms, behavioral attendance was low, and there was no no-taper arm. What this changes: don't require CBT or a class before offering a taper — invest in the taper's structure and monitoring instead.

EndocrinologyPooled randomized crossoverNotable · small effect

Prolonged Short Sleep and Its Effect on Body Weight and Composition

Does 6 weeks of mild sleep restriction (−1.5 h/night) vs adequate sleep actually change body weight?

Population: Pooled analysis of 2 randomized crossover trials; n = 95 adults ≥20 y at elevated cardiometabolic risk with habitual sleep ≥7 h/night.
Intervention: 6 weeks of sleep restriction (−1.5 h/night).
Comparison: 6 weeks of sustained adequate sleep, with a multiweek washout.
Outcome: Sleep fell 78.4 min/night. Body weight +0.45 kg (CI 0.33–0.57), waist +0.52 cm (CI 0.25–0.79), whole-body volume +0.56 L, leptin +2.03 ng/mL, and sedentary time +17.2 min/day (CI 11.7–22.7).

Clinical takeaway. The first reasonably real-world causal evidence that mild, chronic short sleep (~6 h) — not just severe experimental restriction — nudges weight upward, plausibly by increasing sedentary time. The magnitude is small and the authors concede 6 weeks may be too short to move body composition. What this changes: incremental — supports asking about sleep in weight counseling; no discrete practice change.

EndocrinologyPhase 2b RCTNotable · surrogate endpoint

Weekly and Biweekly Bofanglutide Versus Semaglutide in Chinese Patients With Type 2 Diabetes

Can a biweekly GLP-1 receptor agonist match or beat weekly semaglutide on HbA1c?

Population: 272 adults, 37 sites in China, drug-naive or on stable oral agents; HbA1c 7.0–11.0% (mean 8.35%), mean BMI 27.9 kg/m². Open-label, 1:1:1:1:1.
Intervention: Bofanglutide (GZR18) 12, 18, or 24 mg every 2 weeks, or 24 mg weekly.
Comparison: Semaglutide 1 mg weekly (not the 2 mg maximum dose).
Outcome: HbA1c change at wk 24: −1.87 / −2.28 / −1.94% (12/18/24 mg Q2W), −2.32% (24 mg weekly) vs −1.60% for semaglutide. GI adverse events 81.8–87.3% vs 51.9% (mostly grade 1–2); no severe hypoglycemia.

Clinical takeaway. The interesting idea is biweekly dosing — less injection burden for a lifelong therapy. But this is phase 2b, open-label, HbA1c-only, Chinese-only, and benchmarked against submaximal semaglutide, so the apparent superiority is soft; GI adverse events were markedly higher. What this changes: nothing yet — awareness of an extended-interval GLP-1 RA in development.

RheumatologySTOP-HZ · exploratory RCTNotable · surrogate, n=59

Timing of Recombinant Zoster Vaccination and JAK Inhibitor Initiation in Rheumatoid Arthritis

When starting a JAK inhibitor, should you finish the zoster vaccine series first?

Population: 59 analyzed patients (29 vs 30) aged ≥50 y with RA initiating tofacitinib, 12 sites in Japan. All received RZV at day 1 and week 8. Baseline characteristics not fully balanced.
Intervention: Delay tofacitinib to week 8 — i.e. after the second RZV dose (group B).
Comparison: Start tofacitinib at day 1, with the first RZV dose (group A).
Outcome: Primary endpoint was the geometric mean fold rise in VZV-specific IgG at week 12. Delaying the JAK inhibitor gave higher early antibody responses; starting immediately gave faster arthritis improvement. By week 12 antibody responses were similar. Herpes-zoster incidence was not measured.

Clinical takeaway. A real, if small, tradeoff: JAK inhibition appears to delay but not prevent RZV-induced immunity. When disease activity allows waiting, completing RZV before starting a JAK inhibitor may improve early immunogenicity — but titers are a surrogate, zoster incidence was never measured, and n=59 with baseline imbalance is exploratory. What this changes: unresolved — supports "vaccinate first when you can afford to wait," without proving clinical benefit.

Reviews & In the Clinic

PulmonaryHigh yield · IM

In the Clinic: Obstructive Sleep Apnea

Which sleep study should you order for suspected uncomplicated OSA?

Key learning point. Home sleep apnea testing (HSAT) is the preferred initial test in uncomplicated OSA — RCTs show a home testing-and-treatment strategy gives similar outcomes to in-lab polysomnography, at lower cost, with faster treatment initiation and patient preference. The review states plainly that the 2014 ACP guideline endorsing in-lab studies predates this evidence. HSAT records respiratory channels only (oximetry, airflow, chest movement) and has no EEG, so it does not measure sleep directly — which is exactly why it is reserved for uncomplicated patients. Screening tool: STOP-BANG. On outcomes, PAP delivers only modest blood-pressure reduction (24-h SBP −1.5 mm Hg [CI −2.3 to −0.7]; DBP −1.6 mm Hg [CI −2.2 to −0.9]), and RCTs have not shown cardiovascular event prevention in non-sleepy patients (SAVE). Comprehensive lifestyle intervention improved AHI by 8.5 events/h (CI −10.8 to −6.25).

EndocrinologyNotable · restates June guideline

Care of Patients Receiving GLP-1 Receptor Agonists

Key learning point. A practical synopsis of the ACP obesity pharmacotherapy guideline (16 June 2026): semaglutide or tirzepatide first-line with lifestyle modification for adults with obesity (BMI ≥30), moderate-certainty evidence; phentermine–topiramate second-line (low certainty), then liraglutide, then naltrexone–bupropion. For overweight (BMI ≥27 to <30) with T2DM, dyslipidemia, hypertension, OSA, or CVD: semaglutide or tirzepatide first-line, liraglutide second.

Health PolicyLower yield · operational

The Role of Telemedicine on Interhospital Transfer Outcomes: A Systematic Review

Key learning point. 33 studies, 609,188 patients. Telemedicine was associated with lower transfer rates in most adult (13/17) and pediatric (4/5) studies, across rural (9/16) and urban (9/10) settings, with lower or unchanged mortality in 15 of 17 studies reporting it. Evidence is almost entirely observational with substantial heterogeneity — a structured narrative synthesis was used rather than meta-analysis.

Also this month — observational, modeling, and health-services work; no PICO.

GLP-1 RAs and ischemic optic neuropathy — two independent cohorts, same answer. A U.S. claims target trial emulation found 18-month ION risk of 8.5 vs 5.5 per 10,000 vs SGLT2i (risk difference 3.0 [CI 0.4–5.7]) and 7.8 vs 4.2 vs DPP4i (RD 3.6 [CI 1.1–6.1]) — number needed to harm 3333 and 2778; events concentrated in men (70.3%) and age >50 (85.2%). Independently, a Swedish nationwide cohort (107,518 vs 185,898 initiators) found AION in 0.04% vs 0.02% at 1 year (RR 1.93 [CI 1.00–3.73]) and 0.12% vs 0.07% at 5 years (RR 1.69 [CI 0.95–3.01]). Both attenuated substantially when restricted to metformin users, and both author groups conclude the signal may reflect residual confounding. Absolute risk is very low — counseling material, not a reason to withhold therapy. (10.7326/ANNALS-25-00860 · 10.7326/ANNALS-25-02096)

Same-day COVID-19 + influenza vaccination is safe. VA target trial emulation, 705,124 co-administered vs 1,813,205 influenza-only, Sept 2022–Aug 2025 across bivalent, XBB- and KP-adapted eras. 90-day risk was similar for all three composite tiers — tier 1 (serious) RR 1.03 (CI 0.99–1.09), tier 2 RR 0.99, tier 3 RR 0.99. Of 46 individual adverse events only syncope (RR 1.09) and tinnitus (RR 0.95) were nominally significant, neither surviving multiple-comparison correction. (10.7326/ANNALS-26-00217)

Olorofim for disseminated coccidioidomycosis. Novel dihydroorotate dehydrogenase (DHODH) inhibitor antifungal; single-group phase 2b in 41 patients with limited or no treatment options at 10 U.S. sites (73.2% CNS disease, 95.1% non-immunosuppressed). Clinical success 75.6% (CI 59.7–87.6) at day 42 and 73.2% (CI 57.1–85.8) at day 84; main toxicity was hepatic biochemistry elevation in 21.9%. No comparator arm. (10.7326/ANNALS-26-00103)

Carvedilol beat other nonselective β-blockers in cirrhosis. U.S. claims cohort (2013–2025), IPTW over 129 covariates. 6-month major decompensation: vs nadolol RD −3.69 pp (CI −5.33 to −2.09; RR 0.80) and vs propranolol RD −2.88 pp (CI −4.29 to −1.49; RR 0.83), including less variceal hemorrhage and less ascites/SBP/hepatorenal syndrome. Supports the existing carvedilol preference without randomization. (10.7326/ANNALS-25-04010)

Adding a GLP-1 RA does not get patients off insulin. VA target trial emulation, 8,869 matched sets on basal insulin. Over 3 years insulin discontinuation occurred in 16.7% of GLP-1 RA initiators vs 17.9% (SGLT-2i) and 17.1% (DPP-4i) — RR 0.93 (CI 0.86–1.01) and 0.98 (CI 0.87–1.09); no subgroup favored GLP-1 RAs. A useful expectation-setter. (10.7326/ANNALS-25-05216)

Lung cancer risk-prediction models underperform in non-White groups. 641,830 participants, 12 U.S. cohorts, 16 models. Models substantially underestimated risk in non-Hispanic Black participants (expected–observed <0.75 in 11/16) and discriminated worse in Asian participants (13/16). Holding eligibility to the USPSTF-2021 size (38.0%), all risk-based strategies were more efficient and narrowed racial/ethnic gaps; PLCOm2012 and LYFS-CT performed best (mean number needed to screen 36.5 and 40.1). No strategy optimized eligibility, sensitivity, and efficiency simultaneously. (10.7326/ANNALS-25-03816)

Epidemiology & health services. Benzodiazepine prescribing to U.S. adults ≥65, 2015–2024, including long-term use and long-term-care dispensing before vs after the pandemic (10.7326/ANNALS-25-05594) · onsite vs referral-only prenatal care at U.S. federally qualified health centers and variation by maternity-desert status (10.7326/ANNALS-26-00381) · duration of U.S. residence and obesity prevalence among foreign-born adults, NHIS 2019 + 2021–2024 (10.7326/ANNALS-25-05002).

No Beyond the Guidelines teaching case ran in this window.

Summaries are original paraphrases prepared for personal educational use; figures are drawn from the published full text. Read the full articles at acpjournals.org (subscription required). Window: articles posted online 23 June – 25 July 2026; Ann Intern Med 2026, Vol. 179. Trial registrations: NCT03445988 (opioid tapering), NCT06256549 (bofanglutide), NCT02960776 and NCT02835261 (sleep restriction), NCT03583164 (olorofim). DOIs: 10.7326/ANNALS-25-04784 (editorial 10.7326/ANNALS-26-02208), 10.7326/ANNALS-25-01660, 10.7326/ANNALS-25-04623, 10.7326/ANNALS-25-05493, 10.7326/ANNALS-26-02124, 10.7326/ANNALS-26-02496, 10.7326/ANNALS-25-05342, 10.7326/ANNALS-25-00860, 10.7326/ANNALS-25-02096, 10.7326/ANNALS-26-00217, 10.7326/ANNALS-26-00103, 10.7326/ANNALS-25-04010, 10.7326/ANNALS-25-05216, 10.7326/ANNALS-25-03816, 10.7326/ANNALS-25-05594, 10.7326/ANNALS-26-00381, 10.7326/ANNALS-25-05002.