Category: News

CT Angiography Scans Help Target Statin Use in Higher-Risk Patients

Posted on

CT Angiography Can Help Target Statin Use in Higher-Risk Patients

Damian McNamara, MA

November 10, 2025

 

Statin therapy remains a cornerstone for primary and secondary prevention of major adverse cardiac events (MACEs), but prescribing based on patient phenotype identified through imaging may boost its effectiveness, according to a new study.

“While population-level primary-prevention trials have established the efficacy of statins, it remains unclear whether their benefit depends on the extent of underlying atherosclerotic disease. Our work addresses this evidence gap by assessing whether the treatment effect varies with disease characteristics,” lead investigator Bálint Szilveszter, MD, PhD, a researcher at the Semmelweis University Heart and Vascular Centre in Budapest, Hungary, wrote in an email to Medscape Medical News.

“Clarifying this relationship could enable more personalized and also intensified therapy,” Szilveszter added.

Each 10% increase in statin use was associated with a decreased risk for MACE, for example, among those with obstructive coronary artery disease (CAD), high-risk plaque, a calcium score of ≥ 400, or a segment involvement score > 4. In contrast, there was no link between patients with “any CAD” and lower risk for MACE.

By contrast, “in asymptomatic individuals, routine CT screening is not recommended,” Szilveszter said.

The study was published online on August 5, 2025, in JACC: Cardiovascular Imaging.

More Than 8 Years Follow-Up

Szilveszter and colleagues studied 11,026 consecutive adults with stable chest pain referred for clinically indicated coronary CT angiography to assess CAD between January 1, 2013, and December 31, 2020. The population had a mean age of 59 years and 55% were men.

The MACE composite endpoint included all-cause mortality, acute myocardial infarction, or revascularization for unstable angina.

Statin use was based on prescription fills in Hungary’s National Health Service database. Sixty-six percent of patients were treated with a statin during the study period. Median follow-up was 8.3 years from the date of first statin use to an event or end of the study.

Key Findings

MACE was detected in 4.4% of patients, myocardial infarction alone in 0.9%, and all-cause mortality in 3.1%. Following coronary CT angiography, 14% of total patients stopped using statins, 11% started statin therapy, and 40% continued their statin regimen. The remaining patients were not treated with statins.

With each 10% increase in statin use, the risk for MACE was lowered in the presence of:

  • Obstructive CAD (adjusted hazard ratio [aHR], 0.91; 95% CI, 0.85-0.97; P = .006)
  • High-risk plaque (aHR, 0.82; 95% CI, 0.68-0.98; P = .026)
  • Calcium score of ≥ 400 (aHR, 0.93; 95% CI, 0.87-0.99; P = .024)
  • Segment involvement score of ≥ 4 (aHR, 0.89; 95% CI, 0.84-0.95; P < .001)

In contrast, results revealed no significant MACE reduction for “any CAD” (aHR, 0.95; 95% CI: 0.85-1.07; P = .411).

“Our message is not to restrict therapy but to better target and support statin use when anatomic disease is demonstrated,” Szilveszter said.

Whether coronary CT angiography can guide therapy in asymptomatic individuals remains under investigation, he added.

“Given radiation and contrast exposure and costs, we need to define the true cost–benefit and select subgroups in whom use may be justified in the future,” he said.

The Protective Role of Fitness

“Statins are widely prescribed. While it is clear that statin users improved some aspects of their lipid profile, the effects of health outcomes are not so obvious and there are some gaps in the knowledge,” Claudio Gil Araújo, MD, PhD, dean of Research and Education at CLINIMEX — Clínica de Medicina do Exercício in Rio de Janeiro, Brazil, wrote in an email to Medscape Medical News.

There has never been a randomized controlled trial comparing statins, exercise, a combination of both, and controls, according to Abraço.

“And unfortunately, due to other interests, it will never happen. However, observational data suggest that those who are fitter will have modest or no benefit in reducing MACE from taking statins,” he said, citing a 2013 study in The Lancet by Peter Kokkinos, PhD, and colleagues that supports that finding.

“It’s better to be highly fit or at least fit than not being fit and taking a statin,” Kokkinos, lead author and director of the Center for Exercise and Aging and professor at Rutgers-New Brunswick School of Arts and Sciences in New Brunswick, New Jersey, said in an interview.

Kokkinos and colleagues classified 10,043 veterans with dyslipidemia into four strata of fitness. They found significant protective effects against MACE with greater fitness levels, and statins conferred the most benefit among those who were least fit. Median follow-up was 10 years, during which 2318 patients died. The risk for death was 18.5% among those taking statins vs 27.7% among those not taking statins (P < .0001).

Therefore, even though the research shows an interaction between statins and reduced MACE risk, “once you go up to fit or highly fit, I’m not sure it matters,” Kokkinos said.

One potential limitation of the study by Kokkinos and colleagues is that the median age of the veterans was 59 years, so the interaction of statins, exercise, and risk for MACE later in life was not addressed in this study.

“I think it would have been similar; we had some people in their seventies and eighties. But, yes, we do not know this — I don’t have that data,” Kokkinos said.

Not a ‘One-Size-Fits-All’ Therapy

Asked to comment on the findings by Szilveszter and colleagues, Kokkinos said, “If you have some degree of CAD or a higher calcium score, then statins increase your survival rate.”

Kokkinos added that screening patients using coronary CT angiography makes sense. “Of course, the more you know about the severity of disease, the better off you are as a physician. As we get more and more technology, more high-resolution results, we are able to do that better and discriminate the different phenotypes.”

Regarding the controversial public health proposal that statins are so beneficial that they should be added to public water supplies, Kokkinos said, “I’ve heard that, but my research shows that we should try to have everyone exercise instead of putting statins in the drinking water.”

“Statins are not for everyone and should not be used as one-size-fits-all therapy,” he added. Instead, prescription of statins “should be treated more skeptically when it comes to people with no symptoms, no major coronary artery disease, or other risk factors.”

The JACC: Cardiovascular Imagingstudy was independently supported. Szilveszter, Abraço, and Kokkinos reported no relevant financial relationships.

Damian McNamara is a freelance contributor to Medscape Medical News. He worked full-time for Medscape and WebMD from 2018 to 2024. McNamara has a BA in chemistry and an MA in science, health, and environmental reporting/journalism. He works out of a home office in Miami, with a 100-pound chocolate lab known to snore under his desk during work hours.

Stress CMR Helps Pinpoint Diagnosis in Angina With Nonobstructive Arteries

Posted on

Stress CMR Helps Pinpoint Diagnosis in Angina With Nonobstructive Arteries

Ted Bosworth

November 17, 2025

NEW ORLEANS — An evaluation of myocardial blood flow with stress perfusion cardiac magnetic resonance (CMR) imaging after a negative angiogram changed the presumed diagnosis and treatment in the majority of patients in a randomized double-blind trial.

“Endotyping-informed therapy in the MRI-guided intervention arm led to a reduction in angina burden and an improvement in health-related quality of life,” reported the principal investigator Colin Berry, MBChB, PhD, a professor of cardiology and imaging at the University of Glasgow, Glasgow, Scotland.

The results of the late-breaking CorCMR trial were presented at American Heart Association (AHA) 2025 Scientific Sessions 2025. The findings were simultaneously published in Nature Medicine.

CMR Reclassified Diagnosis in More Than Half of the Patients

For the study, Berry and colleagues enrolled 250 patients presenting with chest pain who did not have obstructed coronary arteries on angiogram at three centers in the UK. All underwent a stress perfusion CMR with myocardial blood flow mapping. The randomization took place after angiography ruled out obstructive disease.

In almost all patients (97.6%), the presumed diagnosis after the negative angiogram was noncardiac chest pain, but CMR led to a reclassification in 53% of them (95% CI, 46.6-59.3).

In the intervention group, the management team was provided with the CMR results to guide care. For the control group, the CMR results were withheld, and subsequent management was performed with usual care.

After CMR, the diagnosis of noncardiac chest pain was upheld in only 47% of them. Fifty-one percent of patients were diagnosed with microvascular angina and 0.4% had vasospastic angina. Cardiomyopathy and myocarditis were incidental findings in two patients in each group.

The change in the initial diagnosis was the primary outcome of this diagnostic study. The secondary outcomes included control of angina, measured by the Seattle Angina Questionnaire (SAQ), and change in quality of life, measured by the five-dimension EuroQoL-5 assessment, at 12 months relative to baseline and between study groups.

At 12 months, the SAQ score improved by a mean of 21.7 points in the intervention group but remained essentially unchanged, down 0.8 points, in the control group. Differences in medical therapy are the likely explanation, according to Berry.

For those in the intervention group relative to the control group, anti-anginal therapies (84% vs 64%; P = .001), aspirin (77% vs 56%; P < .001), and statins (82% vs 62%; P = .001) were all offered more frequently.

The high diagnostic yield of CMR for patients who otherwise would have been discharged with a diagnosis of noncardiac chest pain was consistent across patient groups, Berry said.

“These results apply equally by sex but are perhaps particularly relevant to women because microvascular angina associates more commonly with women presenting with chest pain,” he said.

Data Said to Have Immediate Clinical Relevance

The data from CorCMR have immediate clinical relevance, according to Robert A. Harrington, MD, a cardiologist currently serving as dean of Weill Cornell Medicine and provost for Medical Affairs at Cornell University in New York City. He noted the study employed a “clever study design” to address “a common and important clinical problem.”

“This is a strategy we can use routinely for those with INOCA [ischemia with nonobstructed coronary arteries],” Harrington said.

As the second most common reason for emergency department visits among adults, chest pain is an important target for better and more efficient diagnostic strategies, according to Berry. He cited data suggesting that only half of the patients with chest pain are diagnosed with a cardiac cause, but CorCMR data suggest many of these are likely sent home despite cardiovascular pathology.

Based on the CorCMR data, “coronary angiography should include a functional test either invasively or noninvasively” to capture these patients, he maintained.

Janet Wei, MD, co-director of the Stress Echocardiography Lab at Cedars-Sinai Medical Center in Los Angeles, suggested these data coupled with those from other studies are “changing the paradigm of how we manage patients who present with ischemia.”

The importance of separating patients with a negative angiogram who have cardiac disease from those who do not is that INOCA “is not a benign condition,” Wei said. “It is associated with increased major adverse cardiac events as well as increased healthcare costs related to repeat testing.”

Because there has been no systematic approach to identify causes of noncardiac chest pain, delays in identifying the true cause of INOCA can often be measured in years, Wei reported.

“INOCA is associated with significant impairments in quality of life, whether it is the physical health, mental health, or social health,” she said.

This study has not resolved which protocol is best for establishing INOCA and its cause in patients presenting with chest pain, according to Wei, but it does show that taking an extra step to look for INOCA results in improvements in angina symptoms and quality of life.

The CorCMR trial was an investigator-initiated study without industry funding. Berry reported financial relationships with Abbott Vascular, AskBio, AstraZeneca, Boehringer Ingelheim, CorFlow, Edwards Lifesciences, MAIA Pharmaceuticals, Merck, Novartis, Servier, Xylocor, and Zoll Medical. Harrington reported financial relationships with Atropos, Azuma, Basking Biosciences, Bitterroot Bio, Bristol-Myers Squibb, Bridge Bio, Chiesi, CSL Behring, Edwards Lifesciences, Element Science, Foresight, and Merck. Wei reported a financial relationship with Abbott Vascular.

Contrast-enhanced photon-counting detector CT for discriminating local recurrence from postoperative changes after resection of pancreatic ductal adenocarcinoma

Posted on

Original article

Contrast-enhanced photon-counting detector CT for discriminating local recurrence from postoperative changes after resection of pancreatic ductal adenocarcinoma

European Radiology Experimental volume 9, Article number: 26 (2025) Cite this article

Abstract

Background

We evaluated the diagnostic capability of photon-counting detector computed tomography (PCD-CT) spectral variables in late arterial phase (LAP) and portal venous phase (PVP) to discriminate between local tumor recurrence (LTR) and postoperative changes (POC) after pancreatic ductal adenocarcinoma (PDAC) resection.

Methods

Seventy-three consecutive PCD-CT scans in 73 patients with postoperative soft-tissue lesions (PSLs) were included, 42 with POC and 31 with LTR. Regions of interest were drawn in each PSL, and spectral variables were calculated: iodine concentration (IC), normalized IC (NIC), fat fraction, attenuation at 40, 70, and 90 keV, and slope of the spectral curve between 40–90 keV. Multivariable binary logistic regression models were constructed. Diagnostic performance was assessed for LAP and PVP using receiver operating characteristic analysis.

Results

In LAP, all variables except fat fraction showed significant differences between LTR and POC (p ≤ 0.025). In PVP, all variables except NIC and fat fraction demonstrated significant differences between LTR and POC (p ≤ 0.005). Logistic regression analysis included NIC and 70 keV in the LAP-based model and IC and 90 keV in the PVP-based model. Both models achieved a higher area under the curve (AUC) than individual spectral variables in each phase. The LAP-based model achieved an AUC of 0.919 with 94% sensitivity, 84% specificity, and 87% accuracy, while the PVP-based model reached 0.820, 71%, 88%, and 81%, respectively.

Conclusion

Spectral variables from PCD-CT help distinguish between LTR and POC in LAP and PVP post-PDAC resection. Multivariable logistic regression improves diagnostic performance, especially in LAP.

Relevance statement

Measuring normalized iodine concentration and attenuation at 70 keV in late arterial phase, or iodine concentration and attenuation at 90 keV in portal venous phase, and incorporating these values into a logistic regression model can help differentiate between local tumor recurrence and postoperative changes after pancreatic ductal adenocarcinoma resection.

Key Points

  • Distinguishing recurrence from postoperative changes on CT after pancreatic ductal adenocarcinoma resection is challenging.
  • PCD-CT spectral variable values differed significantly between local tumor recurrence (LTR) and postoperative changes (POC).
  • Logistic regression of spectral variables can help distinguish LTR from POC.
  • The late arterial phase-based model reached an AUC of 0.919 with 94% sensitivity and 84% specificity.

Graphical Abstract

Background

Pancreatic ductal adenocarcinoma (PDAC) carries a dismal prognosis attributable to its locally aggressive growth and early systemic spread [1]. Most patients present with advanced disease, with radical surgical resection feasible for only about 30% [2]. The proportion of patients eligible for surgical resection has risen following the introduction of neoadjuvant therapy for locally advanced tumors. In particular, neoadjuvant therapy with the chemotherapy protocol Folfirinox has achieved resectability rates of 60% in patients with initially unresectable tumors [3]. As neoadjuvant treatments have advanced, so too have surgical resection techniques, which continue to serve as the cornerstone in any curative treatment strategy for PDAC [4].

Isolated local recurrence occurs in up to 30% of the patients, and the treatment options are chemoradiotherapy, stereotactic body radiation therapy, and re-resection [5]. It has been shown that median survival after resection of isolated tumor recurrence was significantly longer compared to exploration without resection: 26.0 months versus 10.8 months [6]. Hence, it is crucial to detect local recurrences early through follow-up measures. Studies have demonstrated that patients diagnosed with asymptomatic PDAC recurrence on surveillance computed tomography (CT) had better survival outcomes compared to symptomatic patients with recurrence [7, 8].

The primary challenge in detecting PDAC local tumor recurrence (LTR) on surveillance CT is to differentiate it from postoperative changes (POC). This presents a diagnostic challenge for radiologists as LTR and POC appear morphologically very similar as postoperative soft-tissue lesions (PSLs) [9]. Often, differentiation is only possible by observing progression of PSL on follow-up CT [9] or by performing an 18F-fluorodeoxyglucose positron emission tomography (PET)-CT, which has a higher sensitivity and specificity for LTR compared to contrast-enhanced CT [10]. Dual-energy (DE)-CT perfusion has only shown a non-significant lower trend of perfusion values of LTR compared to POC [11]. Furthermore, there is a significant correlation between perfusion CT parameters and iodine concentration (IC) in PDAC [12]. One study has investigated the potential of dual-energy computed tomography (DECT) to differentiate LTR from POC based on IC and CT numbers (HU) with promising results [13].

Not long ago, a photon-counting detector-CT (PCD-CT) system was approved for clinical use [14]. In a PCD, direct conversion occurs from x-ray photon to electrical signal, where ideally, each photon generates a separate signal. In a conventional energy-integrating detector (EID), however, indirect conversion occurs where the resulting electrical signal is an integrated signal generated from the energy deposition of all incident x-ray photons. Since the signal is proportional to the energy that is deposited by the incident photon, high-energy photons dominate the EID signal [15]. Due to the photoelectric effect, high-Z materials like iodine will exhibit a relatively high difference in attenuation from high to low photon energies compared to low-Z materials, such as soft tissues, which have a relatively energy-independent attenuation [16]. Consequently, contrast information carried by low-energy photons will be lost in EIDs. Since PCDs equally count low- and high-energy photons, a PCD-CT will reach a higher contrast for high-Z materials compared to an EID-CT [17]. Furthermore, PCDs provide energy information for every counted x-ray photon. By setting the energy threshold for photon counts above that of electronic noise, significant noise reduction can be achieved [15].

Early clinical experience of the inherent spectral capabilities of PCD-CT has demonstrated clinical benefits from iodine maps and virtual non-contrast images for quantification of myocardial extracellular volume, adrenal adenoma assessment, quantification of emphysema, and anemia detection [18]. Apart from certain technical limitations of PCDs related to count rate and dead time [19], the improved contrast and noise properties of PCDs could theoretically provide more accurate spectral capabilities compared to EID DECT systems, for instance, more reliable IC measurement. Phantom studies have demonstrated that IC measurement with PCD-CT is more accurate compared to DECT [20, 21]. To our knowledge, no previous study has investigated if PCD-CT-derived spectral data can differentiate between PDAC LTR and POC, as has been demonstrated with DECT [13]. Hence, the primary objective of this study was to determine if PCD-CT spectral variables, particularly IC and attenuation values on virtual monoenergetic images at different keV levels, could help discriminate between LTR and POC after PDAC resection. The secondary aim was to evaluate if a combination of different PCD-CT spectral variables could increase the diagnostic performance for predicting LTR.

Methods

This retrospective study has received approval from the Swedish ethical review authority (Dnr 2023-01724-01, 2023-06715-02), and informed consent was waived owing to the retrospective nature of the study design.

Patients

At our tertiary referral center at Karolinska University Hospital, Stockholm, Sweden, during the period 1 July 2022 to 31 October 2023, we retrospectively included 133 consecutive PCD-CT scans with PSLs around the peripancreatic vessels or at the operative site after resection of PDAC. Thirty-eight PCD-CT scans were excluded because they lacked spectral imaging data format (Spectral Post-Processing data). Sixteen PCD-CT scans were excluded because they were repeated on the same patient. Thus, 79 consecutive PCD-CT scans (79 patients) with PSL were included.

In congruence with a prior DECT study, LTR was diagnosed by observing a ≥ 20% increase in the size of PSL along at least one dimension within a timeframe of < 9 months before and/or after the PCD-CT scan (n = 29) [13]. Alternatively, in instances where the PCD-CT scan was the only available surveillance CT scan, LTR was diagnosed by the presence of PSL in combination with elevated tumor marker (CA 19-9 > 37 kU/L) (n = 2). POC was diagnosed if the PSL was stable (< 20% progression in all dimensions) over a timeframe of ≥ 6 months before and/or after the PCD-CT scan, without concurrent administration of chemotherapy or radiotherapy. One PCD-CT scan was excluded because the patient was chemotherapy-free < 6 months, and five PCD-CT scans were excluded because the observation period was < 6 months. Finally, we included 31 PCD-CT scans with LTR and 42 with POC. Among the 31 PCD-CT scans with LTR, 16 were dual-phase scans, and 15 were in portal venous phase (PVP) only. The 16 late arterial phase (LAP) series from the dual-phase scans formed a group with LTR in LAP. The 16 PVP series from the dual-phase scans were combined with the 15 PVP-only series to form a group with LTR in PVP. Of the 42 PCD-CT scans with POC, 31 were dual-phase scans, and 11 were in PVP only. The 31 LAP series from the dual-phase scans formed a group with POC in LAP. The 31 PVP series from the dual-phase scans were combined with the 11 PVP-only series to form a group with POC in PVP. Binary regression models predicting LTR in LAP were constructed using variables from the LAP groups. Models for PVP utilized variables from the PVP groups. Additionally, combined models incorporating both LAP and PVP data were constructed using variables from the dual-phase scans. The patient inclusion process is illustrated in Fig. 1. An abdominal radiologist with 20 years’ experience in pancreatic imaging verified the diagnoses of LTR and POC, respectively.

Fig. 1
figure 1

Flowchart of patient inclusion. Red framed boxes indicate groups in LAP (group POC_LAP/group LTR_LAP), blue framed boxes indicate groups in PVP (group POC_PVP/group LTR_PVP), and green framed boxes indicate groups in dual phase (group POC_COMB/group LTR_COMB). Each group pair was used to construct binary regression models predicting local tumor recurrence in the late arterial phase, portal venous phase, and dual phase, respectively. LAP, Late arterial phase; LTR, Local tumor recurrence; PCD-CT, Photon-counting detector computed tomography; POC, Postoperative changes; PSL, Postoperative soft-tissue lesion; PVP, Portal venous phase; SPP-files, Spectral post-processing files

Full size image

PCD-CT protocol

The scans were obtained utilizing a dual-source PCD-CT (NAEOTOM Alpha, Siemens Healthineers, Forchheim, Germany) with either a single PVP acquisition (70 s post contrast injection), or with a biphasic acquisition comprising a LAP (35–40 s post contrast injection) and then a PVP. The contrast medium iodixanol (Visipaque® 320 mgI/mL, GE Healthcare, Princeton, NJ, USA) was intravenously administered at a dose of 0.5 g of iodine per kg of body weight with a fixed injection time of 25 s. The maximum dosage weight was 100 kg for men and 80 kg for women.

We implemented the following scanning parameters: multi-energy scan mode (QuantumPlus, Siemens Healthineers); tube potential, 120 or 140 kV; automatic exposure control (CARE Dose4D, Siemens Healthineers); pitch, 0.8; rotation time, 0.5 s; collimation, 144 × 0.4 mm; kernel, Qr36 or Qr44; iterative reconstruction algorithm, quantum iterative reconstruction level 3. The CT series were reconstructed with a slice thickness of 0.6 mm.

Post-processing of spectral data and quantitative analysis

Post-processing of spectral data, encompassing material decomposition and quantification of IC, fat fraction, and calculation of the slopes of HU curves, was conducted utilizing an imaging software (syngo.via, version VB80B, Siemens Healthineers). Within this software, the “Liver Virtual Non-Contrast” application was utilized to measure both the IC (mg/mL) and fat fraction (%). This application employs a modified three-material decomposition algorithm, with the base materials comprising liver tissue, fat, and iodine [22]. One radiologist with 10 years’ experience in abdominal imaging manually drew freehand regions of interest (ROIs) in each LTR and POC on five consecutive slices. The ROI was drawn as large as possible without extending beyond the confines of the lesion, while ensuring that vessels, calcifications, surgical clips, and artifacts, were excluded from the ROI (Figs. 2 and 3). For biphasic PCD-CT scans, the ROIs were placed in the identical location. From these ROIs the IC (mg/mL) and the fat fraction (%) were obtained. To mitigate any potential inter-patient variability stemming from injection rate, contrast agent dosage, and cardiac output, the IC was normalized relative to the IC within the abdominal aorta at the level of the superior mesenteric artery. This normalization was achieved by placing a circular ROI of maximal size within the aortic lumen, while carefully avoiding wall plaques. The normalized IC (NIC) was calculated by the formula: NIC = IC of lesion/IC of abdominal aorta.

Fig. 2
figure 2

Axial contrast-enhanced abdominal PCD-CT scan in LAP in the first row (a, b, c), and in PVP in the second row (d, e, f). In the right column (c, f) are zoomed-in images of the ROIs. The images are of a 77-year-old male with LTR of PDAC post total pancreatectomy. Mean NIC in LAP was 0.08, and mean attenuation at 70 keV was 60.64 HU. According to the LAP model this yields a probability for LTR of 0.812 (81.2%). Mean IC in PVP was 1.03 mg/mL, and mean attenuation at 90 keV was 52.73 HU. According to the PVP model this yields a probability for LTR of 0.700 (70.0%). IC, Iodine concentration; LAP, Late arterial phase; LTR, Local tumor recurrence; NIC, Normalized iodine concentration; PCD-CT, Photon-counting detector computed tomography; PDAC, Pancreatic ductal adenocarcinoma; PVP, Portal venous phase; ROIs, Regions of interest

Full size image
Fig. 3
figure 3

Axial contrast-enhanced abdominal PCD-CT scan in LAP in the first row (a, b, c), and in PVP in the second row (d, e, f). In the right column (c, f) are zoomed-in images of the ROIs. The images are of a 72-year-old male with POC post total pancreatectomy. Mean NIC in LAP was 0.03, and mean attenuation at 70 keV was 35.91 HU. According to the LAP model this yields a probability for LTR of 0.025 (2.5%). Mean IC in PVP was 0.80 mg/mL, and mean attenuation at 90 keV was 35.70 HU. According to the PVP model this yields a probability for LTR of 0.170 (17.0%). IC, Iodine concentration; LAP, Late arterial phase; LTR, Local tumor recurrence; NIC, Normalized iodine concentration; PCD-CT, Photon-counting detector computed tomography; POC, Postoperative changes; PVP, Portal venous phase; ROIs Regions of interest

Full size image

In the syngo.via application “Monoenergetic Plus” only circular ROIs were available for HU-measurements. An as large as possible circular ROI was drawn within each abovementioned freehand ROI without exceeding the size of the latter. HU-values were obtained for virtual monoenergetic images ranging from 40 to 190 keV. We defined the slope of the spectral HU curve (λHU) as the difference between the mean HU-value at 40 keV and the mean HU-value at 90 keV divided by the difference in energy (50 keV), according to the following formula:

We chose the range of 40 to 90 keV because the curve above 90 keV was nearly flat. Hence, the range from 40 kV to 90 keV offered a steeper slope with increased sensitivity to changes in attenuation.

The measurements obtained from the ROIs on the five consecutive slices were averaged by applying a 20% trimmed mean where the minimal and maximal values of the five measurements were excluded, and the remaining three values were averaged [23, 24].

The dimensions of LTR and POC were assessed by measuring their size in three perpendicular axes on the CT series in PVP, and the greatest percentage size difference along any axis was calculated.

Accurate positioning of ROIs and measurements was validated by an abdominal radiologist with 20 years’ experience.

Statistical analysis

The statistical software IBM SPSS (v.28, Chicago, IL, USA) and R Studio (v. 2024.04.1+748, Posit Software, PBC) with the package “coin” (v. 1.4.3), “glmnet” (v. 4.1.8) and pROC (v. 1.18.5) were used to perform the data analysis. A stepwise approach was used to develop the prediction model for each contrast phase, as well as a combined model integrating variables from both phases. First, univariate analysis of seven PCD-CT spectral variables in LAP and PVP was performed comprising: IC, NIC, fat fraction, 40 keV (attenuation at 40 keV), 70 keV (attenuation at 70 keV), 90 keV (attenuation at 90 keV), and λHU. The difference in each variable was compared between the LTR and POC groups. Levene’s test was used to evaluate the equality of variances (i.e., homoscedasticity) of the data in the LTR and POC groups, while the Shapiro-Wilk test was used to test for normality. Data exhibiting homoscedasticity and normality was compared between the groups using the independent samples t-test. Heteroscedastic normal data was compared between the groups with the Welch test (unequal variance t-test). Non-normal data was compared between the groups using a permutation-based two-sample test, implemented via the ‘independence_test()’ function from the “coin” package in R, with 108 resampling iterations (‘nresample = 1e+08’) to approximate the null distribution [25, 26]. A two-sided significance level of 0.05 was set, and the p-values were Bonferroni-corrected. The continuous variables were presented as mean ± standard deviation. Second, all variables, including those not showing significant differences in the univariate analysis, were included in the least absolute shrinkage and selection operator (LASSO) regression analysis with ten-fold cross-validation to identify the most relevant predictors. Variables that were statistically significant in the univariate analysis and retained non-zero coefficients after LASSO regression were selected for further logistic regression analysis. Based on these methods, several binary logistic regression models were proposed. The Corrected Akaike Information Criterion score was utilized to identify the best-fitting models ensuring minimization of overfitting. Multicollinearity among predictor variables was checked by calculating the variance inflation factor. Predictor variables demonstrating a variance inflation factor > 10 were considered to exhibit problematic collinearity, and models containing such variables were excluded. Models that included predictor variables or an intercept that were not statistically significant (p > 0.05) were excluded. The linearity of the logit assumption for each model was tested by plotting the logit of the predicted probabilities against each predictor variable and visually assessing the relationship (Supplementary Fig. S1). The Hosmer-Lemeshow test was employed to evaluate the goodness-of-fit for the logistic regression models. A p-value > 0.05 was considered indicative of a good model fit.

Receiver operating characteristic (ROC) curve analysis was performed on the predictive models utilizing predicted probability values. Additionally, ROC curve analysis was conducted on variables that demonstrated statistical significance in the univariate analysis. The Youden index was calculated to suggest an optimal cutoff value that optimizes the balance between sensitivity and specificity for LTR detection [27]. Areas under the ROC curves (AUCs) were calculated, and the difference between the final models’ AUC and that of each PCD-CT spectral variable was evaluated with the DeLong test.

Results

Patients’ characteristics

The demographic characteristics of the patient population are depicted in Table 1. Thirty-one patients were diagnosed with LTR, 29 of which were identified based on the size progression of PSL according to the abovementioned criteria. For LTR cases, the mean observation period during which the progression of PSL was detected was 4.0 months, and the mean greatest size increase of PSL along any axis was 53.3%. In two patients, where the PCD-CT scan was the only available surveillance CT scan, tumor recurrence was identified based on the presence of a PSL combined with elevated tumor marker (one of the patients had a CA 19-9 of 1356 kU/L, and the other had a CA 19-9 of 376 kU/L). For cases with POC, the mean longest observation period without chemo- or radiotherapy during which the PSL was stable (< 20% size increase) was 23.6 months. Thirty POC cases demonstrated a decrease in the size of the PSL with a mean greatest size decrease along any axis of 17.0%. The mean longest lesion diameter at the start of the observation period was 31.3 mm for LTR and 27.3 mm for POC, without significant difference (p = 0.117). At the end of the observation period, the mean longest diameter was 39.8 mm for LTR and 25.8 mm for POC, a difference that was statistically significant (p < 0.001). The mean ROI areas were larger for LTR than POC, a finding that was statistically significant for all ROI areas except for freehand ROIs in LAP (freehand ROI LAP: 2.00 cm2versus 1.43 cm2, p = 0.078; circular ROI LAP: 0.91 cm2 versus 0.51 cm2, p < 0.001; freehand ROI PVP: 2.27 cm2 versus 1.47 cm2, p = 0.031; circular ROI PVP: 0.96 cm2 versus 0.53 cm2, p = 0.031). There was a lower proportion of microscopically negative resection margins (R0) in LTR cases (12.9%) compared to POC cases (38.1%). Ten LTR cases and three POC cases had ongoing chemotherapy at the time of the PCD-CT. None of the LTR cases underwent biopsy to verify the LTR histologically.

Table 1 Demographic characteristics of the study population, in the LAP and PVP groups
Full size table

Univariate analysis and LASSO regression

Of the seven variables in LAP, all but the variable fat fraction demonstrated significant differences between the LTR_LAP and POC_LAP groups. Of the seven variables in PVP, all but the variables NIC and fat fraction exhibited significant differences between the LTR_PVP and POC_PVP groups (Table 2). Of the 14 variables in the dual phase, seven were significantly different between LTR_COMB and POC_COMB (Supplementary Table S1). LASSO regression identified the optimal penalty parameter lambda (λ), which minimizes model error as log(λ) = -2.449 (λ = 0.08637233) for LAP and log(λ) = -5.991 (λ = 0.0025004) for PVP (Fig. 4). Statistically significant variables from the univariate analysis that retained non-zero coefficients after LASSO regression included three variables each for LAP (NIC, 70 keV, and 90 keV), PVP (IC, 70 keV, and 90 keV), and the dual phase (NIC in LAP, 70 keV in LAP, and 90 keV in PVP) (Fig. 4). Further logistic regression analysis, aimed at balancing goodness of fit and model complexity, resulted in the inclusion of the same selected variables in both the LAP model and the combined model. Hence, combining PCD-CT spectral variables from LAP and PVP did not improve the diagnostic performance of the model (Supplementary Table S2). The following two models offered the best diagnostic value for each contrast phase, respectively:

Table 2 Comparison of quantitative spectral PCD-CT variables between LTR and POC in LAP and PVP
Full size table
Fig. 4
figure 4

a, c, e Plots of LASSO regularization paths (a in LAP, c in PVP, and e in dual phase) showing the log(λ) associated with the minimum cross-validated error (deviance) as indicated by the vertical dashed red line. The continuous line represents the mean and dashed lines on either side of the continuous line indicate one standard error of the mean. b, d, f Plots of LASSO coefficient paths (bin LAP, d in PVP, and f in dual phase) where the individual coefficients are plotted as functions of log(λ). The vertical dashed red line indicates the log(λ) that minimizes model error. The variables that had non-zero coefficients at this log(λ) value and were statistically significant in the univariate analysis were included for further logistic regression analysis (NIC, 70 keV, and 90 keV in LAP; IC, 70 keV, and 90 keV in PVP; and NIC LAP, 70 keV LAP, and 90 keV PVP in dual phase). IC, Iodine concentration, LAP, Late arterial phase; LASSO, Least absolute shrinkage and selection operator; NIC, Normalized iodine concentration; PVP, Portal venous phase; λHU, Slope of the spectral HU curve from 40 to 90 keV (mean HU40keV - mean HU90keV)/(90keV - 40 keV)

Full size image

ROC analysis

The ROC analysis revealed that the AUC for the LAP model was 0.919 (95% CI 0.815–1.000), while for the PVP model, it was 0.820 (95% CI 0.720–0.921) (Fig. 5 and Table 3). The AUCs of the models were compared with those of the variables that showed statistical significance in the univariate analysis. In LAP, the AUC of the model was significantly higher compared to the AUCs of IC (p = 0.034), NIC (p = 0.013), and λHU (p = 0.018). No significant differences were observed in LAP between the AUCs of the variables NIC and 70 keV (p = 0.365), which served as predictors in the LAP model. In PVP, there were no significant differences between the AUC of the model and the AUCs of the individual variables.

Fig. 5
figure 5

ROC analyses of spectral PCD-CT variables that were statistically significant in the univariate analysis and of the logistic regression models for LAP (a) and PVP (b), respectively, for predicting LTR post PDAC resection. Model LAP and Model PVP had the highest AUC values in comparison to any individual variable in LAP and PVP, respectively. AUC, Area under the curve; IC, Iodine concentration; LAP, Late arterial phase; LTR, Local tumor recurrence; NIC, Normalized iodine concentration; PCD-CT, Photon-counting detector computed tomography; PDAC, Pancreatic ductal adenocarcinoma; PVP, Portal venous phase; λHU, Slope of the spectral HU curve from 40 to 90 keV (mean HU40keV - mean HU90keV)/(90keV - 40 keV); ROC, Receiver operating characteristic

Full size image
Table 3 Performance of individual quantitative parameters and binary logistic regression models derived from PCD-CT spectral data for the differentiation between LTR and POC
Full size table

For the LAP model, at a cutoff predicted probability of 0.281, the sensitivity, specificity, and accuracy were 93.8%, 83.9%, and 87.2%, respectively. For the PVP model, at a cutoff predicted probability of 0.494, the sensitivity, specificity, and accuracy were 71.0%, 88.1%, and 80.8%, respectively (Fig. 5 and Table 3).

Discussion

In our study, PCD-CT-derived spectral data has been used for the first time to discriminate between LTR and POC following resection of PDAC. LTR demonstrated significantly higher mean IC in both contrast phases and significantly higher mean NIC in LAP. In LAP, LTR cases exhibited more than double the mean NIC of POC cases. This finding is in line with a previous DECT study that demonstrated significantly higher iodine uptake in LTR compared to POC in “early venous phase” (images acquired 30 s after trigger time point) [13], which corresponds to our LAP (35–40 s after contrast injection).

As far as we know, this study represents the first evaluation of spectral CT variables in a later contrast phase (i.e., PVP, 70 s after contrast injection) for discriminating between PDAC LTR and POC. However, our findings reveal that the diagnostic performance of spectral variables in PVP is comparatively modest when compared to the corresponding variables in LAP. Furthermore, while significant differences in mean IC were evident between LTR and POC during the venous phase, normalization of these values (NIC) weakened the statistical significance, unlike the LAP, where normalization increased statistical significance. The LAP is an earlier phase and, therefore, more sensitive than the PVP to variations in cardiac output and blood pressure.

The observed increase in statistical significance of differences after normalization of IC in LAP indicates that LTR truly has a higher contrast uptake than POC during this phase. Conversely, the non-significant difference in NIC between LTR and POC in PVP suggests that the significant difference in absolute IC values may be attributed to variability in systemic patient factors. Nevertheless, IC in PVP was significantly different between LTR and PVP and was also selected as a predictor variable by the LASSO regression analysis.

Furthermore, the attenuation values were significantly higher for LTR compared to POC in both contrast phases, with the greatest attenuation difference between LTR and POC observed in LAP. This observation, coupled with the significant difference in NIC in LAP, could be due to the already established fact that the contrast enhancement in PDAC increases most during 0–30 s (corresponding to the LAP) and then plateaus after 60 s (corresponding to the PVP) [28].

Given the previously established significant correlation between perfusion CT parameters (blood volume and permeability) and IC in PDAC [12], our findings suggest that vascularization is higher in LTR compared to POC, as also proposed in the previous DECT study [13]. This contrasts with the findings from the initial DECT perfusion study that presented lower perfusion values of LTR compared to POC, indicating a poorer vascular supply of LTR [11], and this discrepancy was also pointed out by the authors of the previous DECT study [13]. The reason for these contradicting results is unclear. Possible sources of measurement error when assessing relatively small lesions near larger vessels on an abdominal CT perfusion scan are breathing artifacts, even though the authors tried to mitigate these by instructing the patients to implement shallow breathing [11].

It is well established that PDAC has a low microvascular density compared to other malignant tumors. Furthermore, the pronounced fibro-inflammatory reaction (desmoplastic reaction) causes vessel collapse. This results in a hypoxic microenvironment that is profibrogenic, but that also triggers pancreatic stellate cells to produce angiogenic factors [29]. Less is known about the morphology and composition of intra-abdominal POC. To our knowledge, nothing has yet been published about the morphology of postoperative retroperitoneal changes. We have found three studies on peritoneal adhesions that challenge previous beliefs that adhesions are merely composed of avascular fibrous scar tissue. These studies demonstrated histologically that apart from containing collagen bundles, peritoneal adhesions are cellular and vascularized structures [30,31,32]. Histological comparison of microvascular density between PDAC and POC is a topic that requires further investigation.

We found that the fat fraction was lower for LTR compared to POC in both contrast phases; however, this difference was not statistically significant. This is in line with the previous DECT study [13].

We have demonstrated that integrating the two LAP variables, NIC and 70 keV, in the LAP model resulted in a higher AUC compared to the AUCs obtained from each variable independently. The increase was statistically significant for the variable NIC. In line with the previous DECT study, we did not observe any significant difference in the AUCs between the NIC and 70 keV variables in LAP. We chose to evaluate the 70 keV virtual monoenergetic images since they are equivalent to 120 kVp images [33]. Also, the mean energy of the photons at 120 kVp is around 70 keV [34].

While further validation is necessary, the LAP model, in its current form, demonstrates higher sensitivity and specificity compared to conventional contrast-enhanced CT without spectral information, which, according to a meta-analysis, showed a pooled sensitivity of 70% and specificity of 80%. Our LAP model also outperforms 18F-fluorodeoxyglucose PET-CT in terms of sensitivity, as reported in the same study where 18F-fluorodeoxyglucose PET-CT had a pooled sensitivity of 88% and specificity of 89% [10]. Compared to other 18F-fluorodeoxyglucose PET-CT studies, however, our model exhibited inferior diagnostic performance, with one study demonstrating a sensitivity of 97.6% and an accuracy of 90% [35], and another study reporting a sensitivity of 90.9%, a specificity of 100.0% and an accuracy of 92.3% [36].

There are several limitations of our study. First, we have a relatively small sample size. However, PDAC is a rare diagnosis, which, coupled with our strict inclusion criteria, further diminishes the pool of eligible patients. Second, none of the PSLs were biopsied. This is because we follow PSLs at our center with CT, and in cases of progression, we only consider biopsy for resectable lesions. None of the LTR cases in our cohort were considered for re-resection (25 cases had metastatic disease, 5 cases had no metastases but inoperable LTR, and 1 case received the best supportive care due to high age (85 years)). Third, 10 of 31 patients with LTR and 3 of 42 patients with POC had ongoing chemotherapy at the time of the PCD-CT scan. Despite this, we decided to include these patients, given the routine administration of adjuvant chemotherapy in most cases. Fourth, in some cases, PSLs were in the immediate vicinity of central abdominal vessels, introducing measurement uncertainty. We tried to mitigate this by making sure that the freehand ROIs are drawn within the confines of the PSL, excluding any vessels, and by applying a 20% trimmed mean of measurements from the five consecutive slices to minimize the influence of outliers. Fifth, the discrepancy in case numbers between the PVP groups (31 LTR cases, 42 POC cases) and LAP groups (16 LTR cases, 31 POC cases) hinders a direct comparison of diagnostic performance between the LAP model and the PVP model. However, given the rarity of this disease, we included the additional PVP-only cases to boost the statistical power of the PVP model, as this later contrast phase has not been assessed in this context in the literature. Furthermore, we created groups comprising patients scanned in both LAP and PVP (group LTR_COMB, n = 16; and group POC_COMB, n = 31), where univariate and logistic regression analysis yielded identical predictor variables as those in the LAP model (NIC in LAP, and 70 keV in LAP). Given the absence of PVP variables in the final model of the combined phase groups, we abstained from further direct comparisons between LAP and PVP variables.

Our study has several advantages over previous research. First, we performed quantitative measurements on CT series with a 0.6-mm slice thickness, unlike the previous DECT study [13], which used a 5-mm slice thickness, thus mitigating partial volume effects. Second, we normalized the IC of the PSL to the IC of the abdominal aorta to minimize inter-patient variations in physiology and contrast administration. The inclusion of NIC as a predictor variable in the LAP model likely increases the robustness and generalizability of this model. We chose not to normalize the attenuation values (variables 40 keV, 70 keV, and 90 keV) to those of the aorta, as it would introduce an additional layer of complexity to the models, and we are unaware of any studies that have calculated the slope of the spectral curve based on normalized HU-values. However, we are planning to perform the normalization of these variables as well in a forthcoming study. We also intend to include a larger number of patients who have undergone PCD-CT in both contrast phases in a future study to further validate the models.

In conclusion, quantitative variables derived from PCD-CT spectral data facilitate the discrimination of LTR from POC in LAP and PVP following PDAC resection. Combining relevant variables in a multivariable binary logistic regression model enhances diagnostic performance, particularly in LAP, where the model at the optimal cutoff value achieves a 94% sensitivity, an 84% specificity, and an 87% accuracy.

Data availability

The datasets used in this study are available from the corresponding author upon reasonable request.

Abbreviations

AUC:
Area under the curve
CT:
Computed tomography
DECT:
Dual-energy computed tomography
EID:
Energy-integrating detector
IC:
Iodine concentration
LAP:
Late arterial phase
LASSO:
Least absolute shrinkage and selection operator
LTR:
Local tumor recurrence
NIC:
Normalized iodine concentration
PCD:
Photon-counting detector
PCD-CT:
Photon-counting detector computed tomography
PDAC:
Pancreatic ductal adenocarcinoma
PET-CT:
Positron emission tomography-computed tomography
POC:
Postoperative changes
PSL:
Postoperative soft-tissue lesion
PVP:
Portal venous phase
ROC:
Receiver operating characteristic
ROI:
Region of interest

References

  1. Kolbeinsson HM, Chandana S, Wright GP, Chung M (2023) Pancreatic cancer: a review of current treatment and novel therapies. J Invest Surg 36:2129884. https://doi.org/10.1080/08941939.2022.2129884

    Article PubMed Google Scholar

  2. Nienhüser H, Büchler MW, Schneider M (2022) Resection of recurrent pancreatic cancer: Who can benefit? Visc Med 38:42–48. https://doi.org/10.1159/000519754

    Article PubMed Google Scholar

  3. Hackert T, Sachsenmaier M, Hinz U et al (2016) Locally advanced pancreatic cancer: neoadjuvant therapy with Folfirinox results in resectability in 60% of the patients. Ann Surg 264:457–463. https://doi.org/10.1097/sla.0000000000001850

    Article PubMed Google Scholar

  4. Schneider M, Hackert T, Strobel O, Büchler MW (2021) Technical advances in surgery for pancreatic cancer. Br J Surg 108:777–785. https://doi.org/10.1093/bjs/znab133

    Article CAS PubMed Google Scholar

  5. Groot VP, van Santvoort HC, Rombouts SJ et al (2017) Systematic review on the treatment of isolated local recurrence of pancreatic cancer after surgery; re-resection, chemoradiotherapy and SBRT. HPB (Oxford) 19:83–92. https://doi.org/10.1016/j.hpb.2016.11.001

    Article PubMed Google Scholar

  6. Strobel O, Hartwig W, Hackert T et al (2013) Re-resection for isolated local recurrence of pancreatic cancer is feasible, safe, and associated with encouraging survival. Ann Surg Oncol 20:964–972. https://doi.org/10.1245/s10434-012-2762-z

    Article PubMed Google Scholar

  7. Nordby T, Hugenschmidt H, Fagerland MW, Ikdahl T, Buanes T, Labori KJ (2013) Follow-up after curative surgery for pancreatic ductal adenocarcinoma: asymptomatic recurrence is associated with improved survival. Eur J Surg Oncol 39:559–566. https://doi.org/10.1016/j.ejso.2013.02.020

    Article CAS PubMed Google Scholar

  8. Tzeng CW, Fleming JB, Lee JE et al (2012) Yield of clinical and radiographic surveillance in patients with resected pancreatic adenocarcinoma following multimodal therapy. HPB (Oxford) 14:365–372. https://doi.org/10.1111/j.1477-2574.2012.00445.x

    Article PubMed Google Scholar

  9. Javadi S, Karbasian N, Bhosale P et al (2018) Imaging findings of recurrent pancreatic cancer following resection. Abdom Radiol (NY) 43:489–496. https://doi.org/10.1007/s00261-017-1397-8

    Article CAS PubMed Google Scholar

  10. Daamen LA, Groot VP, Goense L et al (2018) The diagnostic performance of CT versus FDG PET-CT for the detection of recurrent pancreatic cancer: a systematic review and meta-analysis. Eur J Radiol 106:128–136. https://doi.org/10.1016/j.ejrad.2018.07.010

    Article PubMed Google Scholar

  11. Fritz F, Skornitzke S, Hackert T et al (2016) Dual-energy perfusion-CT in recurrent pancreatic cancer – preliminary results. Rofo 188:559–565. https://doi.org/10.1055/s-0042-105765

    Article CAS PubMed Google Scholar

  12. Skornitzke S, Mayer P, Kauczor HU, Stiller W (2023) Evaluation of optimal acquisition delays of DECT iodine maps in pancreatic adenocarcinoma: a potential alternative to the Patlak model of CT perfusion. Heliyon 9:e14726. https://doi.org/10.1016/j.heliyon.2023.e14726

    Article CAS PubMed PubMed Central Google Scholar

  13. Mathy RM, Fritz F, Mayer P et al (2021) Iodine concentration and tissue attenuation in dual-energy contrast-enhanced CT as a potential quantitative parameter in early detection of local pancreatic carcinoma recurrence after surgical resection. Eur J Radiol 143:109944. https://doi.org/10.1016/j.ejrad.2021.109944

    Article PubMed Google Scholar

  14. U.S. Food and Drug Administration (2021) FDA news release, FDA clears first major imaging device advancement for computed tomography in nearly a decade. https://www.fda.gov/news-events/press-announcements/fda-clears-first-major-imaging-device-advancement-computed-tomography-nearly-decade. Accessed 18 Feb 2024

  15. McCollough CH, Boedeker K, Cody D et al (2020) Principles and applications of multienergy CT: report of AAPM task group 291. Med Phys 47:e881–e912. https://doi.org/10.1002/mp.14157

    Article CAS PubMed Google Scholar

  16. Jost G, McDermott M, Gutjahr R, Nowak T, Schmidt B, Pietsch H (2023) New contrast media for K-edge imaging with photon-counting detector CT. Invest Radiol 58:515–522. https://doi.org/10.1097/rli.0000000000000978

    Article CAS PubMed PubMed Central Google Scholar

  17. Gutjahr R, Halaweish AF, Yu Z et al (2016) Human imaging with photon counting-based computed tomography at clinical dose levels: contrast-to-noise ratio and cadaver studies. Invest Radiol 51:421–429. https://doi.org/10.1097/rli.0000000000000251

    Article CAS PubMed PubMed Central Google Scholar

  18. Sartoretti T, Wildberger JE, Flohr T, Alkadhi H (2023) Photon-counting detector CT: early clinical experience review. Br J Radiol 96:20220544. https://doi.org/10.1259/bjr.20220544

    Article PubMed PubMed Central Google Scholar

  19. Hsieh SS, Rajbhandary PL, Pelc NJ (2018) Spectral resolution and high-flux capability tradeoffs in CdTe detectors for clinical CT. Med Phys 45:1433–1443. https://doi.org/10.1002/mp.12799

    Article CAS PubMed Google Scholar

  20. Sartoretti T, Mergen V, Jungblut L, Alkadhi H, Euler A (2022) Liver iodine quantification with photon-counting detector CT: accuracy in an abdominal phantom and feasibility in patients. Acad Radiol 30:461–469. https://doi.org/10.1016/j.acra.2022.04.021

    Article PubMed Google Scholar

  21. Vrbaski S, Bache S, Rajagopal J, Samei E (2023) Quantitative performance of photon-counting CT at low dose: Virtual monochromatic imaging and iodine quantification. Med Phys 50:5421–5433. https://doi.org/10.1002/mp.16583

    Article CAS PubMed Google Scholar

  22. Siemens Healthineers (2024) syngo.CT DE Virtual Unenhanced. https://www.siemens-healthineers.com/se/computed-tomography/options-upgrades/clinical-applications/syngo-ct-de-virtual-unenhanced. Accessed 4 June 2024

  23. Wilcox RR, Keselman HJ (2003) Modern robust data analysis methods: measures of central tendency. Psychol Methods 8:254–274. https://doi.org/10.1037/1082-989X.8.3.254

    Article PubMed Google Scholar

  24. Ascenti G, Sofia C, Mazziotti S et al (2016) Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma. Clin Radiol 71:938.e1-9. https://doi.org/10.1016/j.crad.2016.05.002

    Article PubMed Google Scholar

  25. Hothorn T, Hornik K, van de Wiel MA, Zeileis A (2006) A Lego System for Conditional Inference. Am Stat 60:257–263. https://doi.org/10.1198/000313006X118430

    Article Google Scholar

  26. Hothorn T, Hornik K, van de Wiel MA, Zeileis A (2008) Implementing a class of permutation tests: the coin package. J Stat Softw 28:1–23. https://doi.org/10.18637/jss.v028.i08

    Article Google Scholar

  27. Perkins NJ, Schisterman EF (2006) The inconsistency of “optimal” cutpoints obtained using two criteria based on the receiver operating characteristic curve. Am J Epidemiol 163:670–675. https://doi.org/10.1093/aje/kwj063

    Article PubMed Google Scholar

  28. Hata H, Mori H, Matsumoto S et al (2010) Fibrous stroma and vascularity of pancreatic carcinoma: correlation with enhancement patterns on CT. Abdom Imaging 35:172–180. https://doi.org/10.1007/s00261-008-9460-0

    Article PubMed Google Scholar

  29. Longo V, Brunetti O, Gnoni A et al (2016) Angiogenesis in pancreatic ductal adenocarcinoma: a controversial issue. Oncotarget 7:58649–58658. https://doi.org/10.18632/oncotarget.10765

    Article PubMed PubMed Central Google Scholar

  30. Epstein JC, Wilson MS, Wilkosz S, Ireland G, O’Dwyer ST, Herrick SE (2006) Human peritoneal adhesions show evidence of tissue remodeling and markers of angiogenesis. Dis Colon Rectum 49:1885–1892. https://doi.org/10.1007/s10350-006-0747-3

    Article PubMed Google Scholar

  31. Herrick SE, Mutsaers SE, Ozua P (2000) Human peritoneal adhesions are highly cellular, innervated, and vascularized. J Pathol 192:67–72. https://doi.org/10.1002/1096-9896(2000)9999:9999

    Article CAS PubMed Google Scholar

  32. Binnebösel M, Klinge U, Rosch R, Junge K, Lynen-Jansen P, Schumpelick V (2008) Morphology, quality, and composition in mature human peritoneal adhesions. Langenbecks Arch Surg 393:59–66. https://doi.org/10.1007/s00423-007-0198-x

    Article PubMed Google Scholar

  33. Rajiah P, Parakh A, Kay F, Baruah D, Kambadakone AR, Leng S (2020) Update on multienergy CT: physics, principles, and applications. Radiographics 40:1284–1308. https://doi.org/10.1148/rg.2020200038

    Article PubMed Google Scholar

  34. Sauter AP, Shapira N, Kopp FK et al (2020) CTPA with a conventional CT at 100 kVp vs. a spectral-detector CT at 120 kVp: comparison of radiation exposure, diagnostic performance and image quality. Eur J Radiol Open 7:100234. https://doi.org/10.1016/j.ejro.2020.100234

    Article PubMed PubMed Central Google Scholar

  35. Jung W, Jang JY, Kang MJ et al (2016) The clinical usefulness of 18F-fluorodeoxyglucose positron emission tomography-computed tomography (PET-CT) in follow-up of curatively resected pancreatic cancer patients. HPB (Oxford) 18:57–64. https://doi.org/10.1016/j.hpb.2015.06.001

    Article PubMed Google Scholar

  36. Rayamajhi S, Balachandran A, Katz M, Reddy A, Rohren E, Bhosale P (2018) Utility of (18) F-FDG PET/CT and CECT in conjunction with serum CA 19-9 for detecting recurrent pancreatic adenocarcinoma. Abdom Radiol (NY) 43:505–513. https://doi.org/10.1007/s00261-017-1316-z

    Article PubMed Google Scholar

Download references

Acknowledgements

LLMs were not used for this manuscript.

Funding

Open access funding provided by Karolinska Institute.

Author information

Authors and Affiliations

  1. Department of Diagnostic Radiology, Karolinska University Hospital, 171 76, Stockholm, Sweden

    Zlatan Alagic, Carlos Valls Duran, Anders Svensson-Marcial & Seppo K. Koskinen

  2. Department of Clinical Science, Intervention, and Technology (CLINTEC), Karolinska Institutet, 171 77, Stockholm, Sweden

    Zlatan Alagic, Carlos Valls Duran, Anders Svensson-Marcial & Seppo K. Koskinen

Contributions

Conceptualization: ZA, CVD, ASM, SKK. Methodology: ZA, CVD, ASM, SKK. Data preparation: ZA, CVD. Data collection: ZA. Data analysis: ZA. Writing original draft: ZA. Writing—review and editing: ZA, CVD, ASM, SKK. All authors reviewed, read, and approved the final manuscript.

Corresponding author

Correspondence to Zlatan Alagic.

Ethics declarations

Ethics approval and consent to participate

This retrospective study has been approved by the Swedish ethical review authority (Dnr 2023-01724-01, 2023-06715-02), and informed consent was waived due to the study’s retrospective design.

Consent for publication

Not applicable.

Study subjects or cohorts overlap

Eight study subjects are also included in another study cohort with a different objective.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Additional file 1 of Contrast-enhanced photon-counting detector CT for discriminating local recurrence from postoperative changes after resection of pancreatic ductal adenocarcinoma

Eur Radiol Exp (2025) Alagic Z, Valls Duran C, Svensson–Marcial A, Koskinen SK.
Contrast–enhanced photon–counting detector CT for discriminating local recurrence from postoperative
changes after resection of pancreatic ductal adenocarcinoma
ELECTRONIC SUPPLEMENTARY MATERIAL
Supplementary Table S1. Comparison of quantitative spectral PCD–CT variables between LTR and POC in dual phase
LTR POC Uncorrected
p–value
Bonferroni–
corrected p–
value
IC (mg/mL) in LAP 0.90 ± 0.34 0.61 ± 0.25 0.001 0.017
NIC in LAP 0.12 ± 0.10 0.05 ± 0.03 < 0.001 0.005
Fat fraction (%) in LAP 17.14 ± 7.05 24.07 ± 8.47 0.007 ns (0.102)
Attenuation at 40 keV (HU) in LAP 94.72 ± 35.25 55.90 ± 23.54 < 0.001 0.009
Attenuation at 70 keV (HU) in LAP 55.87 ± 14.10 38.43 ± 13.01 < 0.001 0.001
Attenuation at 90 keV (HU) in LAP 47.47 ± 11.00 34.22 ± 12.03 < 0.001 0.007
λHU (HU/keV) in LAP 0.95 ± 0.57 0.43 ± 0.33 0.004 ns (0.051)
IC (mg/mL) in PVP 1.12 ± 0.43 0.86 ± 0.26 0.034 ns (0.477)
NIC in PVP 0.24 ± 0.09 0.20 ± 0.08 ns (0.137) ns (1)
Fat fraction (%) in PVP 17.38 ± 6.81 23.02 ± 8.14 0.022 ns (0.308)
Attenuation at 40 keV (HU) in
PVP 112.23 ± 42.11 78.35 ± 21.25 0.007 ns (0.097)
Attenuation at 70 keV (HU) in
PVP 60.82 ± 15.15 45.65 ± 11.17 < 0.001 0.004
Attenuation at 90 keV (HU) in
PVP 50.27 ± 10.76 38.28 ± 10.02 < 0.001 0.006
λHU (HU/keV) in PVP 1.24 ± 0.70 0.80 ± 0.32 0.028 ns (0.394)
IC Iodine concentration, LAP Late arterial phase, LTR Local tumor recurrence, NIC Normalized iodine concentration, ns Not significant, POC Postoperative
changes, PVP Portal venous phase, λHU Slope of the spectral HU curve from 40 to 90 keV (mean HU40keV − mean HU90keV)/(90keV − 40keV)

Eur Radiol Exp (2025) Alagic Z, Valls Duran C, Svensson–Marcial A, Koskinen SK.
Supplementary Table S2. Comparison between different logistic regression models for late arterial phase, portal venous phase,
and both contrast phases combined, comprising variables that were statistically significant in the univariate analysis and retained
non–zero coefficients after LASSO regression
Is there
multicollinearity?,
(highest VIF)
AICc Nagelkerke
R2
Highest p–value
among predictor
variables
p–value of
intercept
ROC
AUC
p–value of
Hosmer–
Lemeshow test
Late arterial phase
NIC + 70 keV + 90 keV Yes, (23.383) 37.510 0.679 0.861 0.003 0.944 < 0.001
NIC + 70 keV* No, (1.004) 37.450 0.643 0.005 0.002 0.919 0.073
NIC + 90 keV No, (1.001) 35.148 0.679 0.005 0.003 0.944 < 0.001
70 keV + 90 keV Yes, (18.961) 49.064 0.436 0.591 0.003 0.843 0.019
NIC N/A 52.168 0.321 0.010 0.001 0.736 0.467
70 keV N/A 47.071 0.430 0.002 0.001 0.851 < 0.001
90 keV N/A 49.861 0.372 0.003 0.002 0.821 0.006
Portal venous phase
IC + 70 keV + 90 keV Yes, (66.530) 69.917 0.548 0.004 < 0.001 0.889 0.395
IC + 70 keV No, (1.296) 82.122 0.373 0.082 < 0.001 0.801 0.158
IC + 90 keV* No, (1.078) 79.029 0.414 0.012 < 0.001 0.820 0.338
70 keV + 90 keV Yes, (15.140) 85.416 0.329 0.855 < 0.001 0.797 0.462
IC N/A 89.766 0.234 0.001 < 0.001 0.769 0.035
70 keV N/A 83.272 0.328 < 0.001 < 0.001 0.792 0.445
90 keV N/A 84.423 0.312 < 0.001 < 0.001 0.790 0.185
Late arterial and portal venous phase
NIC LAP + 70 keV LAP + 90 keV PVP No, (3.109) 38.475 0.665 0.259 0.002 0.925 0.085
NIC LAP + 70 keV LAP* No, (1.004) 37.450 0.643 0.005 0.002 0.919 0.073
NIC LAP + 90 keV PVP No, (1.002) 38.904 0.620 0.009 0.001 0.913 0.189
70 keV LAP + 90 keV PVP No, (3.011) 49.171 0.434 0.667 0.001 0.851 < 0.001
NIC LAP N/A 52.168 0.321 0.010 0.001 0.736 0.467
70 keV LAP N/A 47.071 0.430 0.002 0.001 0.851 < 0.001
90 keV PVP N/A 51.430 0.337 0.003 0.002 0.790 0.186

Additional file 1: Supplementary Table S1.

Comparison of quantitative spectral PCD-CT variables between LTR and POC in dual phase. Supplementary Table S2. Comparison between different logistic regression models for late arterial phase, portal venous phase, and both contrast phases combined, comprising variables that were statistically significant in the univariate analysis and retained non-zero coefficients after LASSO regression. Supplementary Fig. S1. Check of the linearity of the logit assumption for the LAP model (a, b) by plotting the logit of predicted probabilities against the variable NIC (a) and 70 keV (b); and for the PVP model (c, d) by plotting the logit of predicted probabilities against the variable IC (c) and 90 keV (d). The variables from each model demonstrate a satisfactory level of linearity with the logit of predicted probabilities.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Alagic, Z., Duran, C.V., Svensson-Marcial, A. et al. Contrast-enhanced photon-counting detector CT for discriminating local recurrence from postoperative changes after resection of pancreatic ductal adenocarcinoma. Eur Radiol Exp 9, 26 (2025). https://doi.org/10.1186/s41747-025-00567-0

Download citation

  • Received
  • Accepted
  • Published
  • Version of record
  • DOIhttps://doi.org/10.1186/s41747-025-00567-0

Share this article

Anyone you share the following link with will be able to read this content:

Provided by the Springer Nature SharedIt content-sharing initiative

Keywords

MRI as an alternative to CT after inconclusive ultrasound in subacute/acute abdominal pain in young women

Posted on

MRI as an alternative to CT after inconclusive ultrasound in subacute/acute abdominal pain in young women: a prospective multicenter noninferiority study

  • Emergency Radiology
  • Published:
European RadiologyAims and scope Submit manuscript

Abstract

Objective

To assess the noninferiority of MRI diagnostic accuracy to CT scan as a second-line examination of acute/subacute abdominopelvic pain in a population of young women after an inconclusive ultrasound (US).

Methods

This prospective, multicenter non-inferiority study included 18–40-year-old non-pregnant women with non-traumatic acute/subacute abdominal pain. They had an inconclusive US warranting the prescription of an additional CT scan. Within 6 h of the CT, all these women underwent abdomino-pelvic MRI. A retrospective reading of the CT and MR provided a diagnosis using a standardized list. The gold standard diagnosis, based on a 3-month follow-up, was done by a panel of experts. Statistical analysis was conducted to assess the noninferiority of the diagnostic accuracy of MRI compared to that of CT. The noninferiority margin was set at 10%. Inter-observer agreement and diagnostic performance of a conditional imaging strategy were estimated.

Results

133 participants were analyzed (median: 27 years). The most common diagnoses were non-specific pain (30.1%), ovarian cyst rupture (12.8%), and appendicitis (9.7%). MRI demonstrated non-inferiority diagnostic accuracy estimated between 60.9% (81/133) and 88% (117/133) compared to CT, estimated between 64.7% (86/133) and 83.5% (111/133). The conditional imaging strategy (MRI, followed by CT when the MRI was normal) had a diagnostic accuracy of 91%.

Conclusion

MRI diagnostic performances are not inferior to CT for acute abdominal pain in women aged 18–40. A conditional imaging strategy based on MRI would give an accuracy of 91% and might be considered a second-line imaging modality in that context.

Key Points

Question Can MRI serve as an alternative to CT as a second-line imaging modality for acute abdominopelvic pain in young women (18–40) after an inconclusive ultrasound?

Findings MRI accuracy after inconclusive US ranged from 60.9 to 88%. A conditional strategy (MRI first, CT if normal) reached 91% accuracy, avoiding 59% of CTs.

Clinical relevance MRI is not inferior to CT for diagnosing uncategorized causes of acute abdomino-pelvic pain in young non-pregnant women. A conditional imaging strategy based on MRI as a second-line imaging modality would give an accuracy of 91%.

Preliminary reports of lung cancer screening with low-dose computed tomography: a nationwide performance on the Korean population in 2019–2020

Posted on

Preliminary reports of lung cancer screening with low-dose computed tomography: a nationwide performance on the Korean population in 2019–2020

  • Computed Tomography
  • Published:
European RadiologyAims and scopeSubmit manuscript

A Commentary to this article was published on 18 June 2025

Abstract

Background

Korea introduced its National Lung Cancer Screening Program (NLCSP) with low-dose computed tomography (LDCT) in 2019. However, there are few published results of lung cancer screening at national level. We investigated the performance of LDCT for lung cancer screening and suggested recommendations for improving a nationwide population-based program.

Materials and methods

This study was a nationwide, population-based cross-sectional study. We analyzed the Korean National Health Insurance Big Data Base, which included lung cancer-related screening information between 2019 and 2020, was analyzed. Performance indicators were the number of examinations, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), overexposure, and suspected lung cancer (LCA). The performance indicators were stratified according to screening year, sex, age group, and region.

Results

A total of 149,936 examinations were done in 2019–2020. Most participants were men (98.3%) or were aged 54–59 or 60–64 (35.2% and 38.2%). The sensitivity, specificity, PPV, NPV, overexposure, and suspected LCA were 82.9%, 91.9%, 6.4%, 99.9%, 2.7%, and 4.0%, respectively. Women showed lower sensitivity than men. In 2020, there was a substantial increase in PPV (11.0%), and a substantial decrease in overexposure (47.7%) and suspected LCA (16.9%) compared to 2019. At regional level, there was a large variance in sensitivity, PPV, overexposure, and suspected LCA.

Conclusions

Korea’s NLCSP successfully demonstrated validity and reliability at national level. Sexual, and regional differences should be addressed to improve NLCSP.

Key Points

Question How has biennial low-dose CT (LDCT) lung cancer screening performed in Korea for individuals aged 54–74 years with a smoking history of ≥ 30 pack-years?

Findings The sensitivity, specificity, and positive predictive value were 82.9%, 91.9%, and 6.4%.

Clinical relevance Korea’s program demonstrated a similar level of validity and reliability of previous lung cancer screening studies at national level. It facilitated early diagnosis of lung cancer for all the relevant Korean population and laid the foundation for long-term cancer screening.

Automated coronary analysis in ultrahigh-spatial resolution photon-counting detector CT angiography: Clinical validation and intra-individual comparison with energy-integrating detector CT

Posted on

Automated coronary analysis in ultrahigh-spatial resolution photon-counting detector CT angiography: Clinical validation and intra-individual comparison with energy-integrating detector CT

a,b,c ∙ a,d ∙ a musccvi@musc.edu ∙ … ∙ a,j ∙ a,k,l ∙ a,h … Show more
Cover Image - Journal of Cardiovascular Computed Tomography, Volume 19, Issue 5

Abstract

Objectives

To evaluate a deep-learning algorithm for automated coronary artery analysis on ultrahigh-resolution photon-counting detector coronary computed tomography (CT) angiography and compared its performance to expert readers using invasive coronary angiography as reference.

Methods

Thirty-two patients (mean age 68.6 years; 81 ​% male) underwent both energy-integrating detector and ultrahigh-resolution photon-counting detector CT within 30 days. Expert readers scored each image using the Coronary Artery Disease–Reporting and Data System classification, and compared to invasive angiography. After a three-month wash-out, one reader reanalyzed the photon-counting detector CT images assisted by the algorithm. Sensitivity, specificity, accuracy, inter-reader agreement, and reading times were recorded for each method.

Results

On 401 arterial segments, inter-reader agreement improved from substantial (κ ​= ​0.75) on energy-integrating detector CT to near-perfect (κ ​= ​0.86) on photon-counting detector CT. The algorithm alone achieved 85 ​% sensitivity, 91 ​% specificity, and 90 ​% accuracy on energy-integrating detector CT, and 85 ​%, 96 ​%, and 95 ​% on photon-counting detector CT. Compared to invasive angiography on photon-counting detector CT, manual and automated reads had similar sensitivity (67 ​%), but manual assessment slightly outperformed regarding specificity (85 ​% vs. 79 ​%) and accuracy (84 ​% vs. 78 ​%). When the reader was assisted by the algorithm, specificity rose to 97 ​% (p ​< ​0.001), accuracy to 95 ​%, and reading time decreased by 54 ​% (p ​< ​0.001).

Conclusion

This deep-learning algorithm demonstrates high agreement with experts and improved diagnostic performance on photon-counting detector CT. Expert review augmented by the algorithm further increases specificity and dramatically reduces interpretation time.

Discrete non-calcified plaque is associated with increased major adverse cardiovascular events in a high cardiovascular risk population with low coronary artery calcium scores (0–100)

Posted on

Discrete non-calcified plaque is associated with increased major adverse cardiovascular events in a high cardiovascular risk population with low coronary artery calcium scores (0–100)

a ∙ b ∙ b ∙ … ∙ a ∙ c ∙ a jm3472@rwjms.rutgers.edu … Show more
Cover Image - Journal of Cardiovascular Computed Tomography, Volume 19, Issue 5

Abstract

Background

Patients with coronary artery calcium (CAC) scores of 0–100 and non-calcified plaque (NCP) on coronary computed tomography angiography (CCTA) have traditionally been considered low risk for obstructive coronary artery disease (CAD) and future adverse cardiovascular events (CVEs). In regions with high pre-test probability for CAD and negative social determinants of health, rates of adverse CVEs remain higher than in lower-risk populations.

Methods

A retrospective review from January 2019 to May 2022 of 1050 symptomatic patients without known CAD and a CAC score of 0–100 identified 385 patients (37 ​%) with discrete NCP and 665 patients (63 ​%) without NCP on CCTA. The study’s primary endpoint was to identify predictors of discrete NCP presence and future adverse CVEs (death, non-ST and ST-elevation myocardial infarction, or cerebrovascular accident) within two years.

Results

A logistic regression analysis showed the presence of discrete NCP in patients with a CAC score of 0–100 was significantly associated with hyperlipidemia (OR 1.556, 95 ​% CI [1.145–2.115], p ​< ​0.005), diabetes mellitus (OR 1.475, 95 ​% CI [1.043–2.085], p ​< ​0.028), tobacco use disorder (OR 1.372, 95 ​% CI [1.028–1.830], p ​< ​0.032), older age (OR 1.035, 95 ​% CI [1.022–1.048], p ​< ​0.001), elevated systolic blood pressure (OR 1.020, 95 ​% CI [1.011–1.028], p ​< ​0.001), and higher total CAC score (OR 1.013 95 ​% CI [1.007–1.020], p ​< ​0.001). Patients with NCP had higher cardiac risk scores (ASCVD and Morise score) and were more likely to live in rural communities (0–5000 people) (p ​< ​0.005). They also had higher rates of coronary angiography, non-ST and ST-elevation myocardial infarctions, and coronary artery bypass grafting at two years (p ​< ​0.001). The presence of discrete NCP remained an independent predictor for future adverse CVEs after adjusting for diabetes mellitus, systolic blood pressure, hyperlipidemia, total CAC score, age, female sex, body mass index, and community population size (aOR 1.882, 95 ​% CI [1.048–3.380], p ​< ​0.034). Patients with discrete NCP had a 5.70 ​% adverse CVE rate.

Conclusion

High rates of discrete NCP (37 ​%) and subsequent adverse CVEs were observed in our symptomatic, high cardiovascular risk population with CAC scores of 0–100. The presence of discrete NCP on CCTA was an independent risk factor for future adverse CVEs. Our findings emphasize the need for a more comprehensive approach to cardiovascular risk assessment in these vulnerable groups.

More Heart Fat on Coronary CTA scans Linked to Greater Risk for Atrial Fibrillation

Posted on

More Heart Fat Linked to Greater Risk for Atrial Fibrillation

Edited by Patricia McKnight

August 29, 2025

New research reveals a significant connection between having more fat around the heart and the development of atrial fibrillation (AF) in previously unaffected individuals.

In a comprehensive study of over 2200 participants aged 40 and above, researchers used cardiac CT angiography to measure epicardial fat volume. The findings were striking: People with the most fat in this area were more than twice as likely to develop AF compared to those with the least fat.

In fact, even a small increase in this heart fat raised the risk by nearly 30%. Overall, the chances of getting AF jumped from about 5% to over 11% between the lowest and highest fat groups.

The implications are significant for clinical practice. This tissue can be modified through weight loss and medication, opening new possibilities for early prevention strategies.

While the study has some limitations, including possible underestimation of AF cases and its observational nature, these findings suggest a compelling new direction for AF risk assessment and prevention.

This content was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.

Credits

Lead image: Medscape, LLC.

Medscape Medical News © 2025 WebMD, LLC
Send comments and news tips to news@medscape.net.

Cite this: More Heart Fat Linked to Greater Risk for Atrial Fibrillation – Medscape – August 29, 2025.

Rise in Late-Stage Lung Cancer in Nonsmokers Highlights Need for Awareness and Screening

Posted on

Rise in Late-Stage Lung Cancer in Nonsmokers Highlights Need for Awareness and Screening

Evra Taylor

September 05, 2025

As the rates of late-stage lung cancer in nonsmokers rise in Canada, researchers are advocating for increased awareness and screening.

According to the Canadian Cancer Society, lung cancer was estimated to account for 24.3% of all cancer deaths in 2022. This proportion was more than colorectal, pancreatic, and breast cancer deaths combined. The Society estimates that roughly a quarter of lung cancer cases in the country affect nonsmokers.

The Canadian Partnership Against Cancer noted that patients whose lung cancer is detected early have a better chance of recovering than patients who are diagnosed later. But in Canada, patients with lung cancer are more likely to be diagnosed at an advanced stage (ie, stage III or IV) than patients with breast, prostate, or colorectal cancer.

A Smoker’s Disease?

The common thinking is that lung cancer is a smoker’s disease, but that isn’t the full picture, Jessica Moffatt, PhD, vice president of programs and health system partnerships at Lung Health Foundation in Toronto, told Medscape Medical News. “The only thing it takes to get lung cancer is having a pair of lungs. All of us have a risk for lung cancer,” she said. “We still don’t have a map of genetic profiles or exposure that equates with someone having a diagnosis of lung cancer.”

Moffatt and her colleagues are working to dispel the stigma that smokers “get what they deserve.” Rosalyn Juergens, MD, professor of oncology at McMaster University in Guelph, Ontario, and president of Lung Cancer Canada, said, “If you find out someone has lung cancer, your first question shouldn’t be ‘Did you smoke?’ It should be ‘What can I do to help you along this journey?’”

“The vast majority of patients with lung cancer will enter late-stage disease at some point in their cancer journey, with nearly half being diagnosed initially at the late stages,” Geoffrey Liu, MD, senior scientist at the Princess Margaret Cancer Centre in Toronto, told Medscape Medical News. “The incidence varies by city and province, based on demographics, but it’s roughly 20%,” said Liu. He highlighted the lack of awareness among the public, patients, and physicians regarding the incidence of lung cancer in nonsmokers. Many of his nonsmoker patients with lung cancer are completely shocked by their diagnosis. When they search online, most of the information focuses on smoking and lung cancer. Liu and Juergens are working with family doctors to raise awareness of the risk in nonsmokers. “Many of them had previously thought lung cancer in nonsmokers was a rarity,” said Liu.

“Unfortunately, these cases are sometimes diagnosed late, as patients often don’t develop symptoms until the cancer has begun to grow and spread,” said Lawson Eng, MD, a medical oncologist and researcher at the Princess Margaret Cancer Centre and an assistant professor of medicine at the University of Toronto, Toronto. Another reason for late diagnosis is that symptoms such as coughing tend to go unnoticed or be ignored by patients and physicians.

Early Screening Needed

Clinicians may be less likely to investigate respiratory symptoms in nonsmokers, potentially overlooking early warning signs, according to Moffat, who previously led the Ontario Lung Screening Program. She stated that a major contributor to late-stage diagnosis is the design of current lung cancer screening programs, which primarily target older adults with a significant smoking history. As a result, nonsmokers — especially younger women — are rarely screened proactively.

“Unlike for patients who have smoked, where there is evidence and support for lung cancer screening, for patients who have never smoked, we currently don’t have any guidelines related to lung cancer screening, and this is another area of ongoing active research,” added Eng.

“We’ve realized that lung cancer isn’t one disease. Now, we’re able to do molecular testing on tumors,” said Juergens. “There are now more than a dozen molecularly defined subtypes of lung cancer. Getting that information is vitally important because if you’ve got one type of the disease, I might have a tablet for you, and even if your cancer has spread to your bones, brain, or liver, you might live for a decade or more. We can now match the subtype with the right treatment, whether it’s a tablet, chemotherapy, chemotherapy combined with immunotherapy, or immunotherapy alone,” she added.

The prognosis for these patients varies depending on the stage of the cancer, said Eng. “When caught early, lung cancer is potentially curable, but if it has spread elsewhere, often it is not curable. The prognosis for patients with advanced lung cancer depends in part on the molecular profile of the lung cancer. Other factors include how well the patient is, their ability to tolerate treatment, and their response to treatment.”

The chances of surviving 5 years with lung cancer have doubled in the past 15 years, noted Juergens. She views immunotherapy as a game-changer in improving lung cancer survival rates. “Twenty years ago, when I was giving people chemotherapy as their only option, I told them there was only a 1 in 5 chance it would shrink their tumor at all.” This led many patients to refuse the treatment. “Now, if I’ve got the right profile, with stage IV cancer I tell them they have a 1 in 3 chance of still being here 5 years later. We only got national access to next-generation sequencing in the past couple of years, during the COVID-19 pandemic. It solidified all the infrastructure needed to make this a reality,” Juergens concluded.

Moffatt, Juergens, Liu, and Eng reported having no relevant financial relationships.

Evra Taylor is a widely published freelance medical writer and reporter with 20 years’ experience covering a broad range of therapeutic sectors, including family health, cardiology, psychiatry, ophthalmology, and dermatology.