TL;DR
This large international study found that biparametric MRI (bpMRI), a shorter protocol without contrast agents, is as effective as the standard multiparametric MRI (mpMRI) for diagnosing clinically significant prostate cancer. This suggests bpMRI could streamline prostate cancer diagnosis, increasing patient throughput and access to this crucial imaging modality.
What Problem Does This Paper Address?
Prostate cancer diagnosis often relies on magnetic resonance imaging (MRI). The standard approach, multiparametric MRI (mpMRI), involves multiple sequences, including dynamic contrast-enhanced (DCE) imaging. However, DCE imaging adds time, cost, and patient discomfort, and its role is increasingly debated. Biparametric MRI (bpMRI), which omits DCE, has shown promise in smaller studies, but its generalizability and noninferiority to mpMRI in a large, diverse, international setting remained to be confirmed. This study aimed to definitively assess whether bpMRI is a viable, non-inferior alternative to mpMRI for diagnosing clinically significant prostate cancer (csPCa).
Methods
The researchers conducted a large-scale, international, multireader, multicase observer study. 62 radiologists from 45 centers evaluated 400 prostate MRI examinations. Crucially, they performed a paired, sequential assessment: first, they evaluated the images using a biparametric protocol (T2-weighted and diffusion-weighted imaging), assigning PI-RADS scores. Then, they were unblinded to the dynamic contrast-enhanced (DCE) sequence and reassessed the images using the full multiparametric protocol. This design allowed for a direct comparison of diagnostic performance between the two protocols. The reference standard for diagnosing csPCa was histopathology combined with at least three years of follow-up data. Statistical analyses focused on comparing the area under the receiver operating characteristic curve (AUROC), sensitivity, and specificity at various PI-RADS thresholds.
Datasets
The study utilized a dataset of 400 prostate MRI examinations collected between 2015 and 2021 from four European care centers: Radboud University Medical Center, Ziekenhuisgroep Twente, Prostaat Centrum Noord-Nederland, and St. Olav's Hospital. These examinations were from men with a suspicion of prostate cancer, characterized by elevated PSA levels and/or abnormal digital rectal examination findings. Patients with prior prostate treatment or prior csPCa findings were excluded.
Key Insights and Ideas
- bpMRI is a strong contender: The most significant takeaway is that bpMRI performs comparably to mpMRI for diagnosing clinically significant prostate cancer. This challenges the necessity of DCE in many routine cases.
- DCE's limited added value for non-experts: While DCE might offer some benefit, it appears to be marginal, especially for radiologists with less experience. This suggests that training and standardization of bpMRI interpretation could be a more impactful area for improving diagnostic accuracy across the board.
- Thresholds matter for decision-making: Decision curve analyses revealed that the optimal imaging pathway (bpMRI vs. mpMRI) and PI-RADS threshold for biopsy decisions depend on the patient's risk tolerance and willingness to undergo biopsies. In cancer-averse scenarios, mpMRI might offer a slight edge, but bpMRI becomes more favorable when prioritizing fewer biopsies.
Key Results
- Noninferiority confirmed: bpMRI demonstrated noninferior diagnostic performance to mpMRI in terms of AUROC (0.853 vs. 0.859, a difference of -0.6%).
- Similar sensitivity and specificity: At the PI-RADS ≥3 threshold, bpMRI showed sensitivity of 88.6% and specificity of 58.6%, very close to mpMRI's 89.4% sensitivity and 57.7% specificity.
- Trade-offs at different thresholds: At higher PI-RADS thresholds (e.g., ≥4), mpMRI offered slightly higher sensitivity but at the cost of increased false positives compared to bpMRI.
- Expert vs. Non-expert performance: Expert readers generally performed better than non-experts for both bpMRI and mpMRI, but the difference in performance gain from DCE was minimal between the groups.
Limitations and Future Work
The authors acknowledge several limitations:
- Verification bias: Biopsies were guided by mpMRI, which could introduce bias in the reference standard.
- Sequential reading: The sequential reading of bpMRI followed by mpMRI might have introduced a bias, though it mimics clinical practice.
- Exclusion criteria: Patients with prior treatment or severe artifacts were excluded, potentially underestimating DCE's value in these specific scenarios.
- Reader expertise: The majority of readers were experts, and studies were often from high-throughput centers, which might limit generalizability to less experienced readers or lower-volume centers.
Future work should focus on:
- Prospective studies: Large-scale, prospective, randomized trials are needed to validate these findings and assess the impact of bpMRI on clinical decision-making, particularly regarding biopsy strategies.
- PI-RADS v2.1 evaluation: Further evaluation of the PI-RADS v2.1 scoring system for bpMRI, potentially integrating quantitative MRI and AI, could enhance its accuracy.
- Real-world implementation: Investigating the practical implementation of bpMRI in routine clinical practice, including its impact on patient throughput and cost-effectiveness, is crucial.
Summary updated