Incidental coronary artery calcium (CAC) seen on chest computed tomography (CT) scans done for reasons other than atherosclerotic cardiovascular (ASCVD) risk assessment identifies high-risk individuals.1,2 Artificial intelligence (AI) algorithms can quantify CAC on non-dedicated CT scans.1 In NOTIFY-1, we demonstrated that notifying statin-naïve patients without ASCVD and their clinicians of AI-identified incidental CAC (AI-CAC) increased statin prescription rates.3 The median notification time in NOTIFY-1 was >2 years after CT scan.3 The feasibility and impact of using AI-CAC in real-time care in broader populations are unknown. We designed NOTIFY-Picture of Incidental Calcium To Understand Risk Estimate (NOTIFY-PICTURE) (NCT05588895) to investigate whether notification of AI-CAC in near real-time would impact statin prescriptions across a broader spectrum of clinicians and patients during routine care.
NOTIFY-PICTURE was a single-center, prospective trial that randomized lipid-lowering-naïve patients 1:1 to notification or usual care, stratified by ASCVD history. The trial was approved by the Stanford University IRB with a waiver of informed consent. The trial protocol, statistical analysis plan, and de-identified data are available from the corresponding author on reasonable request.
Eligible patients were aged ≥18 and <85 years who underwent a non-contrast, non-ECG-gated chest CT at Stanford. CT scans were screened using an available AI algorithm4 (FDA-cleared, 510(k) K230223) from 01/2024 through 09/2024. Those with AI-CAC score >0 underwent eligibility screening. Exclusions included baseline lipid-lowering therapy, statin allergy/intolerance, language other than English, Spanish, Vietnamese, Cantonese, or Mandarin, advanced cancer, or an acute/life-threatening condition.
For patients randomized to notification, the patient’s cardiologist or primary care clinician was notified with an image of the patient’s CAC and American College of Cardiology/American Heart Association (ACC/AHA) guideline recommendations on statin use. Clinician and patient notification processes were similar to NOTIFY-1.2 The primary outcome was statin prescription within 6 months.
A sample size of 200 provided 95% power to detect an absolute 20% statin prescription difference in the Notification arm (30%) compared with usual care (10%), using a two-sided α of 0.05. Analysis was conducted with the intention-to-treat principle. The primary analysis compared statin prescription rates using Fisher’s exact test. Treatment effect heterogeneity was assessed in pre-specified subgroups.
Of 6,188 patients screened, 202 patients were randomized after application of exclusion criteria. On average, patients were 67.9 years old; 116 (57.4%) were female, 91 (45.0%) were non-White, median AI-CAC was 100, and 15 (7.5%) spoke a non-English language. A total of 172 (85.1%) were without baseline ASCVD—the median 10-year predicted ASCVD risk was 7.0% for PREVENT (interquartile range 4.7–10.1) and 13.3% for ACC/AHA (interquartile range 7.1–23.5%), respectively. While CAC was mentioned in 196 (97.0%) of radiology reports, only 8 (4.0%) mentioned CAC in the impression (Table).
Table.
Patient Characteristics and Outcomes
| Baseline Patient Characteristics | ||
|---|---|---|
| Characteristics | Notification (n=100) | Usual Care (n=102) |
| Age, y | 69.3 (9.3) | 66.6 (10.3) |
| Female Sex | 57 (57.0) | 59 (57.8) |
| Race or Ethnicity | ||
| Non-Hispanic White | 56 (56.0) | 55 (53.9) |
| Asian | 20 (20.0) | 23 (22.5) |
| Black | 8 (8.0) | 9 (8.8) |
| Hispanic | 11 (11.0) | 6 (5.9) |
| Other | 5 (5.0) | 9 (8.8) |
| Preferred Language | ||
| English | 92 (92.0) | 95 (93.1) |
| Non-English | 8 (8.0) | 7 (6.9) |
| ASCVD† History | 14 (14.0) | 16 (15.7) |
| PREVENT ASCVD 10-Year Risk | ||
| <5% | 20 (20.0) | 24 (23.5) |
| ≥5% to <10% | 32 (32.0) | 27 (26.5) |
| ≥10% | 24 (24.0) | 25 (24.5) |
| Unavailable | 24 (24.0) | 26 (25.5) |
| PCE 10-Year Risk | ||
| <7.5% | 19 (19.0) | 23 (22.5) |
| ≥7.5% to <20% | 33 (33.0) | 31 (30.4) |
| ≥20% | 26 (26.0) | 26 (25.5) |
| Unavailable | 22 (22.0) | 22 (21.6) |
| Chest CT Scan Indication ‡ | ||
| Pulmonary nodule evaluation | 27 (27.0) | 43 (42.2) |
| Pulmonary | 40 (40.0) | 28 (27.5) |
| Cancer-related | 25 (25.0) | 20 (19.6) |
| Other | 8 (8.0) | 11 (10.8) |
| Primary Outcome § | ||
| Statin Prescription at 6 Months | 44 (44.0) | 10 (9.8) |
| Key Secondary Outcomes | ||
| Healthcare Utilization | ||
| Follow-up primary care visits per participant | 1.97 (2.3) | 1.42 (1.7) |
| Follow-up cardiology visits per participant | 0.38 (1.0) | 0.22 (0.6) |
| Cardiovascular Testing || | 41 (41.0) | 28 (27.5) |
Table Notes. Continuous variables displayed as mean (standard deviation) and categorical variables displayed as frequency (percentage), unless noted otherwise. Bolded outcomes indicate significant differences with p-value <0.05.
Abbreviations: ASCVD, atherosclerotic cardiovascular disease; CT, computed tomography; PCE, Pooled Cohort Equations; PREVENT, Predicting Risk of Cardiovascular Disease EVENTs.
ASCVD was defined using relevant ICD-9 and ICD-10 codes for coronary artery disease, cerebrovascular disease, and peripheral arterial disease
Pulmonary indication included shortness of breath and pulmonary embolism; Cancer-related includes lung cancer screening and all cancer follow up
Difference in proportion for the primary end point is 34%, 95% confidence interval of (23%,46%)
Cardiology testing includes electrocardiogram, echocardiogram, coronary artery calcium testing, coronary computed tomography angiography, stress testing and invasive angiography
Statin prescriptions occurred in 44 (44%) Notification arm patients and 10 (9.8%) Usual Care patients (p<0.001). Results were consistent across prespecified subgroups (age, sex, AI-CAC) except baseline ASCVD. Statin prescription among those without ASCVD was greater (47% in Notification vs. 6% in Usual Care) than those with ASCVD (29% in Notification vs. 31% Usual Care) (p-interaction=0.003). There was no increase in healthcare utilization following notification.
NOTIFY-PICTURE demonstrated that AI-CAC notification soon after a CT scan for non-cardiac indications increases statin prescriptions among patients without ASCVD, not among those with ASCVD.
Extending our prior findings from NOTIFY-13, NOTIFY-PICTURE showed that statin prescriptions increase even when notifications are delivered in real-time, across more representative patients, including those speaking non-English languages, and that clinician notifications can effectively expand beyond primary care clinicians. Although nearly three-fourths of patients met guideline criteria for lipid-lowering therapy5 and almost all had a comment of CAC in radiology reports, we found that notification was a potent motivator for statin prescriptions. This study demonstrates that making findings actionable using timely image-based notification of AI-CAC is not only effective but may also be the missing link in translating incidental CAC findings into improved care.
A finding that warrants further investigation is that Notification patients with ASCVD were less likely to receive a statin prescription compared with those without ASCVD, while Usual Care patients with ASCVD were more likely to receive a statin prescription than those without ASCVD. We hypothesize that patients with ASCVD may have already been recommended to take statin therapy; thus, additional notification was ineffective.
This study has several limitations, including: 1) a single-center randomized trial may not be generalizable to other health systems, 2) lack of power to detect subgroup differences and low-density-lipoprotein cholesterol changes, 3) statin adherence was not assessed, and 4) lack of assessment of what notification approach (patient, clinician or both) with or without imaging would be most impactful.
In this pragmatic, prospective clinical trial, real-time notification of clinicians and patients of CAC identified by AI on non-ECG-gated chest CTs led to a significant increase in statin prescriptions compared with usual care, and may be a scalable method to implement lipid-lowering therapy in high-risk patients.
Acknowledgments:
RD, SSJ, ATS, DJM, and FR designed the study. DM, SX, and SN were involved in trial execution. AF provided statistical support and analysis. DE, NK, AC, CL were involved in technical workflows regarding AI and radiology for trial design and execution. DS is a patient representative, integral to the design and execution of the study. RD, SSJ wrote the manuscript with other authors providing revisions and comments.
Sources of Funding:
This study was supported by the Doris Duke Foundation (Grant #2022051). Dr. Rodriguez also receives funding from the NIH National Heart, Lung, and Blood Institute (R01HL168188; R01HL167974; R01HL169345).
Non-standard Abbreviations and Acronyms
- ACC
American College of Cardiology
- AHA
American Heart Association
- AI
artificial intelligence
- AI-CAC
AI-identified incidental coronary artery calcium
- ASCVD
atherosclerotic cardiovascular disease
- CAC
coronary artery calcium
- CT
computed tomography
- PICTURE
Picture of Incidental Calcium To Understand Risk Estimate
Footnotes
Disclosures:
Dr. Dudum reports consulting fees from Novartis. Dr. Jain has received research funding from the American Heart Association and Janssen Research and Development LLC, and reports consulting fees from Bristol Myers Squibb, ARTIS Ventures, and Broadview Ventures outside of the submitted work. Dr. Mastrodicasa has received research funding from the National Heart, Lung, and Blood Institute; has provided consulting services to Segmed Inc, and has equity interest in Segmed Inc. Mr. Eng and Mr. Khandwala are employees of Bunkerhill. Dr. Chaudhari receives research support from NIH, ARPA-H, GE Healthcare, Philips, Microsoft, Amazon, Google, NVIDIA, Stability; has provided consulting services to Patient Square Capital, Chondrometrics GmbH, and Elucid Bioimaging; is co-founder of Cognita Imaging; has equity interest in Cognita Imaging, Subtle Medical, LVIS Corp, Brain Key outside of the submitted work. Dr. Sandhu has received consulting fees from Cleerly Health, Holosis Health, and Reprieve Cardiovascular, and has received research funding from the American Heart Association, NIH, Astra Zeneca, Bayer, Novartis, and Novo Nordisk outside of the submitted work. Dr. Maron reports research funding to institution from the National Heart, Lung, and Blood Institute, Cleerly, Inc, and Omada Health; advisory board with New Amsterdam; equity in Ablative Solutions, Inc.; consultant fees from HeartFlow, and Inno Med, Johnson & Johnson, Regeneron, and Scilex; DSMB Astra Zeneca. Dr. Rodriguez reports consulting fees from Novartis, NovoNordisk, Esperion Therapeutics, Movano Health, Kento Health, Edwards, Arrowhead Pharmaceuticals, HeartFlow, iRhythm, Amgen, and Cleerly Health outside the submitted work. Dr. Langlotz has received research funding to his institution from AWS, BunkerHill Health, Carestream, CARPL.ai, Clairity, GE HealthCare, Google Cloud, IBM, Kheiron, Lambda, Lunit, Microsoft, Nightingale Open Science, Philips, Siemens Healthineers, Stability.ai, Subtle Medical, VinBrain, Visiana, Whiterabbit.ai, National Institute of Biomedical Imaging and Bioengineering (NIBIB), National Heart, Lung, and Blood Institute (NHLBI), Advanced Research Projects Agency for Health (ARPA-H), and Agency for Healthcare Research and Quality (AHRQ); consultant fees from Sixth Street and Gilmartin Capital; speaking fees and honoraria from Singapore Ministry of Health, Philips Medical, Canon Medical, McKinsey & Company outside the submitted work; equity in Bunkerhill Health, Sirona Medical, Whiterabbit.ai, Galileo CDS, ADRA.ai, Cognita, TurboRadiology. The remaining authors report no relevant disclosures or competing interests.
References
- 1.Foraker R, Sperling L, Bratzke L, Budoff M, Leppert M, Razavi AC, Rodriguez F, Shapiro MD, Whelton S, Wong ND, et al. Opportunistic Detection of Coronary Artery Calcium on Noncardiac Chest Computed Tomography: An Emerging Tool for Cardiovascular Disease Prevention: A Scientific Statement From the American Heart Association. Circulation. 2025. doi: 10.1161/CIR.0000000000001382. [DOI] [PubMed] [Google Scholar]
- 2.Peng AW, Dudum R, Jain SS, Maron DJ, Patel BN, Khandwala N, Eng D, Chaudhari AS, Sandhu AT, Rodriguez F. Association of Coronary Artery Calcium Detected by Routine Ungated CT Imaging With Cardiovascular Outcomes. J Am Coll Cardiol. 2023;82:1192–1202. doi: 10.1016/j.jacc.2023.06.040. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Sandhu AT, Rodriguez F, Ngo S, Patel BN, Mastrodicasa D, Eng D, Khandwala N, Balla S, Sousa D, Maron DJ. Incidental Coronary Artery Calcium: Opportunistic Screening of Prior Non-gated Chest CTs to Improve Statin Rates (NOTIFY-1 Project). Circulation. 2023;147(9):703–714. doi: 10.1161/CIRCULATIONAHA.122.062746. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Eng D, Chute C, Khandwala N, Rajpurkar P, Long J, Shleifer S, Khalaf MH, Sandhu AT, Rodriguez F, Maron DJ, et al. Automated coronary calcium scoring using deep learning with multicenter external validation. NPJ Digit Med. 2021;4:88. doi: 10.1038/s41746-021-00460-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Grundy SM, Stone NJ, Bailey AL, Beam C, Birtcher KK, Blumenthal RS, Braun LT, de Ferranti S, Faiella-Tommasino J, Forman DE, et al. 2018 AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA Guideline on the Management of Blood Cholesterol. J Am Coll Cardiol. 2019;73:e285–e350. doi: 10.1016/j.jacc.2018.11.003. [DOI] [PubMed] [Google Scholar]