Lung cancer is the leading cause of cancer-related deaths worldwide, with tobacco smoking being the primary risk factor. When detected at an early stage, the disease can often be cured with surgery or radiation - carrying a 5-year survival rate of 69-82% - but most cases are found too late for curative treatment.
Low-dose CT (LDCT) screening of high-risk heavy smokers has demonstrated mortality reductions in major clinical trials. The NLST trial in the US showed a 20% reduction in lung cancer deaths, while the European NELSON trial demonstrated a 24% reduction in men and a remarkable 33% reduction in women.
Despite this evidence, most European countries lack organized national screening programs. The European Council issued a formal recommendation in 2022 urging member states to explore feasibility and effectiveness of LDCT screening, prompting pilot programs across the continent.
Women represent an underserved group in lung cancer screening research. In Sweden, women already account for the majority of lung cancer cases, yet most screening trials have enrolled predominantly men. The Stockholm PLUS study was designed specifically to fill this evidence gap.
The Stockholm PLUS study is the first lung cancer screening pilot conducted in Sweden, run as a single-center, prospective cohort study between September 2022 and September 2024 at Karolinska University Hospital in collaboration with the Regional Cancer Center Stockholm-Gotland.
Recruitment targeted women aged 54-74 in southern Stockholm using the civil registry database. Eligibility mirrored the NELSON trial criteria: a minimum of 15 pack-years of smoking history with current daily smoking or cessation within the past 10 years. Over 34,580 invitation letters were mailed with QR codes linking to an online smoking history questionnaire.
The primary objectives were feasibility and reproducibility, measuring participation rates, detection rates, adverse events, and the proportion of unnecessary (benign) surgeries. A secondary analysis evaluated the performance of AI-assisted nodule detection and its potential to reduce radiologist workload.
Unlike annual screening programs, this pilot was limited to a single baseline LDCT scan with optional 6-12 month follow-up for intermediate findings, reflecting its exploratory feasibility purpose rather than being a full-scale screening program.
All scans were performed on two Siemens CT systems at Karolinska University Hospital using ultra-low radiation protocols, with a median effective dose of only 0.3 mSv - comparable to a few weeks of natural background radiation. Protocols were validated using a chest phantom before clinical use.
Nodule findings were categorized into five codes: Code 1 (negative, no follow-up needed), Code 2 (intermediate, nodules 6-8 mm requiring follow-up CT in 6-12 months), Code 3 (benign at follow-up, discharged), Code 4 (indeterminate at follow-up, referred for continued monitoring), and Code 5 (positive, nodules 8 mm or larger requiring diagnostic work-up).
Two experienced subspecialized radiologists independently read each scan using Fleischner Society guidelines adapted for local practice. All scans were simultaneously analyzed by the Siemens AI-Rad Companion Chest CT system for automated nodule detection and measurement, with AI results available to radiologists during their reads.
Adverse events were tracked using standardized oncology criteria, capturing any grade 3 or higher complications from the screening process itself or subsequent diagnostic procedures including biopsies and surgery.
Of the 34,580 women invited, 11,607 (33.4%) completed the online questionnaire - a participation rate comparable to major European trials. Of those responding, 1,106 (10%) met the eligibility criteria and were invited for scanning, and 990 (90%) of eligible women actually underwent the baseline LDCT.
At the baseline scan, 783 participants (79%) had negative or clearly benign findings requiring no follow-up. Another 152 (15%) had intermediate findings requiring follow-up CT, and 55 (6%) had positive findings - nodules 8 mm or larger - prompting immediate referral for diagnostic evaluation.
Among the 152 intermediate cases with completed follow-up by the censor date, 99 (67%) were confirmed benign, 41 (28%) required continued follow-up due to persistent indeterminate findings, and 8 (5%) grew to 8 mm or more and were reclassified as positive.
In total, 63 individuals were classified as screen-positive at baseline or follow-up and referred for diagnostic work-up. The positive predictive value (PPV) - the proportion of positive screens that were actually cancer - was 24%, meaning roughly 1 in 4 screen-positive women had lung cancer.
Fifteen women were diagnosed with lung cancer through the screening program, yielding an overall detection rate of 1.5%. Thirteen were detected at the baseline scan and two at follow-up of intermediate findings.
Strikingly, 87% of screen-detected cancers were stage IA - the earliest and most curable stage - far exceeding the proportion seen in major earlier trials: 58% in NELSON, 67% in UKLS, 54% in MILD, and 50% in NLST. All patients with early-stage disease received curative-intent surgery or stereotactic body radiotherapy.
Molecular profiling revealed actionable mutations in 4 of 15 cases (27%): two KRAS-G12C mutations, one EGFR L858R, and one ALK rearrangement. While these are primarily relevant for advanced disease, they may influence decisions about surveillance for recurrence.
Screen-detected cancers were predominantly adenocarcinoma (87%), with a median tumor size of 15 mm - consistent with patterns seen in other European trials. Lung cancer cases were on average older (median age 67) than the general screened population (median age 62), a pattern also noted in the NELSON trial.
All 990 baseline scans were analyzed by the Siemens AI-Rad Companion Chest CT system for automated detection and measurement of solid nodules. This allowed direct comparison between AI-assisted and radiologist-based classification.
The AI system missed or underclassified 29 cases that were rated positive or intermediate by the radiologists, corresponding to an 86% sensitivity. Importantly, none of the 15 lung cancers diagnosed before the censor date were missed or underclassified by the AI - all missed cases were ultimately deemed negative by updated ESTI 2025 volumetric criteria.
A key limitation of the comparison is that the study used maximum diameter as its classification threshold, while the AI system is optimized for solid nodule detection and volumetric measurement. Several AI misclassifications involved non-solid or partly solid nodules that the AI is not designed to assess.
The AI correctly identified 489 cases as negative, suggesting a potential 49% reduction in radiologist workload if AI pre-screening were implemented. The researchers note this is likely an underestimate, since volumetric AI assessment per ESTI 2025 guidelines would probably classify even more cases as clearly negative.
Screening-related harms were limited. Three grade 3 adverse events occurred - all in patients who were ultimately diagnosed with lung cancer and undergoing treatment. These included one case of post-surgery atrial fibrillation, one air leak requiring reoperation, and one case of neutropenic sepsis from chemotherapy.
Benign (unnecessary) surgeries were rare. Only one participant underwent surgery for a benign lesion out of 16 who had curative-intent surgery, a benign resection rate of 6%. This is substantially lower than in NLST (23%) and NELSON (28%), suggesting improved nodule management in contemporary practice.
Two additional cases involved non-cancer diagnostic procedures: one CT-guided biopsy prompted a diagnostic wedge resection revealing benign inflammatory lung disease, and one patient underwent lobectomy after inconclusive biopsies, with pathology showing only fibrosis.
Number needed to screen was 66 - meaning 66 high-risk women needed to be scanned to detect one lung cancer case. This compares favorably with other cancer screening programs and reflects the enrichment achieved by targeting heavy smokers.
The Stockholm PLUS study confirms that organized lung cancer screening is feasible in Sweden, with participation rates, adherence, cancer detection rates, and safety profiles consistent with other major European trials. This is the first such demonstration in Sweden.
The exclusively female cohort design addresses a genuine evidence gap, demonstrating that screening is both well-accepted and effective in women - a group bearing an increasing and often overlooked burden of lung cancer in Europe.
Key strengths include the population-based recruitment strategy, dual radiologist reading, prospective AI integration, and molecular profiling of detected cancers. Limitations include the single-center, single-round design and the absence of volume-based CT measurements, which are now recommended by European guidelines.
A follow-up study is planned to expand the cohort by 2,000 participants, invite the initial screened cohort for a second round to assess repeat participation rates, and include male participants to evaluate generalizability across genders.