Abstract
Background In the new era of risk-based prostate cancer (PCa) screening, stratifying men by their individual risk is crucial. Our study aims to integrate anamnestic, life-style and molecular data to minimize unnecessary second-level investigations (SLI) and improve detection of clinically significant PCa (ISUP>1).
Methods Since 2021, within the DP3 study, we collected plasma samples, recent clinical history, familiarity for cancer, and life-style information from: 421 men who underwent PSA testing and digital rectal examination (longitudinal sampling every 6 months), 421 men with suspected PCa, before diagnostic biopsy and 62 men with PCa diagnosis, before radical prostatectomy. Fifty sphingolipids were tested in the plasma of 393 men by targeted lipidomics. Univariable logistic regression analysis was applied to identify the variables associated with PCa. Multivariable penalized logistic regression analysis with 10-fold cross validation was run on different subgroups of the total cohort to build each time 10 models on 9/10 of the samples and assess their performance on the 10 test sets.
Results PSA levels, age, cardiovascular diseases (CVD), number of medications, hypertension, sedentariness, and three sphingolipids (Gb3.24, Cer.20, Cer.24.1) showed a significant association with clinically significant PCa in univariable analysis of the entire cohort. Penalized logistic regression modelling consistently selected hypertension, CVD, PSA, age and five sphingolipids (HexCer.20, Cer.20, HexCer.24.1, GM3.24.1, DHCer.24) as key variables to accurately classify PCa (average ROC AUC on the 10 test sets: 0.92 vs 0.85 for PSA). In men recalled for SLI, PSA showed poor discriminatory ability, but PSA, age, CVD, SM.16, HexCer.20, HexCer.24.1, DHS1P, and DHCer.24 were consistently selected by 10/10 models (average ROC AUC: 0.83 vs 0.65 for PSA). Cer.20 and CVD or CVD alone were key variables also for the discrimination of ISUP>1 PCa within the entire cohort or within men recalled for SLI, respectively.
Conclusions The evaluation of individual risk factors and circulating sphingolipids may allow for a more accurate identification of PCa in the context of tailored screening.
Introduction
Prostate cancer (PCa) is the most commonly diagnosed cancer and ranks as the fifth leading cause of cancer-related deaths among men in Italy and Austria (https://gco.iarc.who.int/en; Globocan 2022, version 1.1). Digital rectal examination (DRE) has low sensitivity in identifying PCa (1). Measuring ematic prostate-specific antigen (PSA) levels enables early PCa detection, but presents some limitations: aggressive tumors can be missed in men with PSA levels below 4 ng/ml and higher PSA levels do not automatically correspond to a PCa diagnosis. This last aspect leads to the significant downside of unnecessary invasive biopsies for a large fraction of men (2). Multiparametric magnetic resonance imaging (MRI) offers high accuracy in detecting clinically significant PCa and can reduce the need for unnecessary biopsies. However, it is resource-intensive, requiring experienced radiologists, considerable time, and significant costs. A set of biomarkers able to complement PSA and DRE in the identification of patients who require further clinical examinations is mandatory to spare patients without PCa from unpleasant, costly, and invasive procedures.
Starting from the analysis of plasma samples collected before diagnostic biopsy from nearly 400 men, we previously reported two circulating microRNAs associated with PCa, especially when PSA was lower than 4 ng/ml (3). In parallel, we uncovered a possible role of ceramides in discriminating PCa from benign prostatic hyperplasia (BPH) when PSA was in the so-called “grey zone”, with 40% reduction of PSA false positives (4). Ceramides are structurally defined as sphingoid bases, typically sphingosine with 18 carbons, attached to a fatty acyl chain of variable length (14 to 26 carbons), the most common one being 16 carbons in mammalian cells (5). Ceramides are essential intermediates in the biosynthesis and metabolism of all sphingolipids and regulate many pathways, such as apoptosis, cell cycle, cell senescence and differentiation, through the synthesis and degradation of different sphingolipids (6). Their homeostasis is crucial and their role in regulating biological processes is still partly unexplored, with some species associated with a tumor suppression and other to a tumor promoting function (7). Interestingly, aberrations in circulating species of the sphingolipid pathways have been consistently found in patients with localized PCa that clinically progressed or with metastatic/castration therapy resistant PCa (8–10), in PCa resistant to docetaxel (11) or to androgen receptor signalling inhibitors (12), as well as in neuroendocrine PCa (13) and in prostate cells treated with fatty acid synthase inhibitor, highlighting a role also in mediation of lipid damage (14).
Recent advancements in metabolomic profiling have evidenced a dysregulation of amino acids and sphingolipid metabolism during the development of PCa (15). Case-control studies from the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort (16–18) have highlighted patterns of circulating metabolites, including sphingolipids, associated with PCa cancer risk or with the development of aggressive PCa. However, to our knowledge no study has yet exploited samples from men undergoing longitudinal PCa screening or diagnostic biopsy to measure such molecules, alone or in combination with PSA and individual risk factors.
In this work, we took advantage of a prospective collection of plasma samples and questionnaire records from men undergoing organized PSA testing and DRE, prostate biopsy for suspected PCa, or radical prostatectomy. Information on lifestyle, cancer familiarity and comorbidities was analyzed in association with PCa diagnosis, to highlight the most important risk factors. We then combined them with PSA, and circulating sphingolipid levels to create models to better identify men who need SLIs and discriminate those who harbor intermediate and high grade PCa (International Society of Urological Pathology – ISUP – higher than 1) from the others.
Material and methods
Cohort recruitment, sample and data collection
421 men (age range: 50-69) undergoing organized PSA screening at Fondazione Edo ed Elvo Tempia (FEET) were prospectively recruited between September 2021 and October 2023. Blood sampling was carried out at recruitment and every six months, for at least three times. All the recruited subjects were checked by a urologist and underwent DRE. Those recalled for SLI were followed-up according to the European Association of Urology (EAU) guidelines (https://uroweb.org/guidelines/prostate-cancer). Sixty-two PCa patients were prospectively recruited at the Biella Hospital (BH) between October 2022 and April 2024. Blood sampling was carried out before surgery and at each follow-up, according to the EAU guidelines. 421 men were prospectively recruited at Molinette Hospital (MH) before diagnostic biopsy for suspected PCa, and blood sampling was carried out at recruitment. Subjects with negative DRE and no alteration of PSA for at least 4 consecutive evaluations were named healthy donors, “HD”; these, together with subjects with negative MRI/biopsy, were named “nonPCa”.
Self-reported questionnaires on lifestyle (physical activity, body mass index (BMI), diet, smoke and alcohol consumption), familiarity, past prostate biopsies and concurrent pathologies, as well as PSA levels and DRE outcomes, have been collected for each subject recruited at FEET, BH and MH. An online password protected database (https://www.keyform.eu) was used to upload the electronic Case Report Forms (eCRF) including questionnaires, urological visit report, PSA dosage and lab analyses at each follow-up, as well as histological and clinical information for PCa. Information on selected records can be downloaded to carry out statistical analyses. Each patient was pseudoanonymized and signed an informed consent.
The observational clinical protocol called DP3 coordinated by FEET was approved by the Novara ethical committee (AOU Maggiore della Carità) in September 2020 (Prot. N° 968/CE 07/09/2020, integration Prot. N°263/CE, 10/03/2021), in accordance with the Declaration of Helsinki Ethical Principles for Medical Research involving Human Participants. The MH ethical committee (AOU Città della Salute e della Scienza) in Turin approved the DP3 study in January 2022 (Prot. N° 521/2021, 10/01/2022).
A summary of the cohorts available is depicted in Table 1.
Blood sample processing
Whole blood was collected in 3 EDTA K2 tubes (10 ml each, D.P. MEDICAL s.r.l., Centallo (TO), Italy) by venipuncture. Plasma was isolated within 1 hour from blood sampling. FEET and BH samples were processed at the Genomics Lab of FEET while MH samples were processed at MH. For plasma separation, tubes were centrifuged at 2500 rpm for 10 min at 4°C; supernatant was then collected in new 15 ml tubes and centrifuged at 2500 rpm for 10 min at 4°C again; plasma was then aliquoted in barcoded tubes and stored at −80°C. Frozen plasma samples from MH were then sent to the Genomics Lab of FEET.
PSA evaluation
PSA concentration was centrally evaluated at BH for all the men recruited at FEET and BH using Elecsys total PSA kit IVD (Cobas, Roche Diagnostics, Monza, Italy). For the men recruited at MH, pre-biopsy PSA records were shared with FEET, according to the study protocol and the data transfer agreement.
Targeted sphingolipid analysis and processing
Frozen plasma aliquots were sent to the Department of Health Sciences, University of Milan, where 50 species of the sphingolipid metabolism (Ceramides, dihydroCeramides, sphingomyelins, hexosylCeramides, lactosylCeramides, globotriaosylCeramides, gangliosides GM3, sphingosines – see complete list in Supplementary table 1) have been analyzed in a subset of 393 plasma samples by mass spectrometry. Briefly, from 25 μL of plasma, sphingolipids were isolated by monophasic extraction (methanol/chloroform 2:1, Merk Life Science S.r.l., Milan, Italy) coupled to alkaline methanolysis. The extracts were resuspended in pure methanol and injected in a LC-MS/MS apparatus: Dionex 3000 UPLC, (ThermoFisher Scientific, Monza, (MB), Italy), coupled to 3200 QTRAP, (AB Sciex S.r.l. Milan, Italy). The chromatographic separation was attained on a C8 column for complex sphingolipids and a C18 column for sphingoid bases.
Four species were undetectable in most of the samples and 46 sphingolipids were retained for the analysis. Z-score standardization was performed on raw data, and batch effect was corrected using the ComBat function from the sva package in R.
Lifestyle score calculation
A lifestyle score reflecting concordance with the World Cancer Research Fund (WCRF) Recommendations (19) and the indications of the Italian Guidelines for a healthy diet (20) was defined. Starting from questionnaire data available, five parameters were considered: BMI, physical activity, dietary habits, smoking habit and alcohol consumption. A value between −1 and +1, based on a lesser or greater adherence to the WCRF Recommendations and the indications of the Italian Guidelines, was assigned to each parameter, as detailed below. For BMI score, the value of 1 was given to each subject whose BMI was in the range of healthy weight (18.5-24.9), 0.5 for underweight (< 18.5), 0 for overweight (25.0-29.9), −1 for obesity (≥ 30). For physical activity, self-reported engagement in high level physical activity (3 or more times a week) was assigned a value of 1, moderate physical activity (1-2 times a week) a value of 0, while sedentariness a value of −1. For smoking, 1 was assigned to those subjects who reported not smoking or quitting more than 15 years ago, 0 to those who had quit less than 15 years prior and −1 to current smokers. For alcohol, 1 was given for no consumption, 0.5 was given for alcohol consumption frequency of 1-2 times a week, 0 for a frequency of 3-4 times a week and −1 for a daily alcohol consumption. For diet, the definition of a score was more complex because of the variety of food items and the requirement to assess the overall dietary pattern of each subject. Calculation details are reported in the supplementary materials.
Statistical analyses
The Shapiro-Wilk normality test was applied to continuous variables in order to test the normality of their distributions. Log transformation was applied when the normality assumption was not satisfied. Univariable and multivariable logistic regression analysis were applied to identify features associated with PCa diagnosis. Least absolute shrinkage and selection operator (LASSO) penalized logistic regression, with 10-fold cross-validation, was applied to retrieve the variables associated with PCa in 10/10 penalized regression models. The performance of the 10 models on the test sets was compared with the performance of PSA alone or of the multivariable models combining PSA and age.
The area under the Receiver Operator Curve (ROC AUC) was calculated to assess the discriminatory ability of the variables or models, with reported 95% confidence intervals (C.I.), or range when the 10 AUCs were averaged. Paired DeLong test was used to compare the discrimination among different variables/models. Youden index was used as the criterion to derive a cut-off point on the predicted probabilities and compute true/false positives/negatives, sensitivity and specificity.
Wilcoxon and Chi-square tests were applied to compare continuous and categorical covariates between groups, respectively.
All analyses were performed using the glmnet, pROC, ggpubr, and stats packages within the statistical computing environment R (version 4.2.1). Associations were considered statistically significant if the p-value of the univariable logistic regression analysis was below 0.001. Differences were considered statistically significant if the p-value of the Wilcoxon, Chi-square or DeLong tests were below 0.01.
Results
Sample characteristics
In total, baseline plasma samples were available from 904 men in the DP3 cohorts (FEET+BH+MH) (Table 1). Within the MH cohort, 260 out of 421 subjects (61.8%) had a positive biopsy. Within the FEET cohort, 13 out of 40 subjects who underwent SLI were classified as PCa (32.5%): of them, three are under active surveillance (AS), one was treated with radiotherapy (RT), three had prostatectomy in other hospitals than BH, six were surgically treated at BH and were followed-up like the 62 PCa patients enrolled before surgery at BH. The clinical, lifestyle and demographic characteristics of the men enrolled in the DP3 study who filled the questionnaire on life-style habits, anamnestic information and risk factors are reported in Table 2.
The study workflow is depicted in Figure 1.
Workflow of the study. FEET: Fondazione Edo ed Elvo Tempia; BH: Biella Hospital; MH: Molinette Hospital; SLI: Second level investigation; HD: Healthy donors. CVD: cardiovascular diseases; BMI: Body Mass Index. Black dashed borders: men recalled for SLI. Red rectangulars: PCa. Green rectangulars: non PCa
Lifestyle and risk factors
All the 904 enrolled men within the DP3 cohorts received a questionnaire on lifestyle, familiarity, past biopsies and concurrent pathologies. The analysis of questionnaires revealed that age, pharmacological treatments in general, hypertension, cardiovascular diseases (CVD) and sedentariness were significantly associated with PCa, independently of PSA. Age, CVD and the number of medications were positively associated also with ISUP>1 PCa when ISUP1 tumors were included in the reference group, independently of PSA (Table 3).
Odds ratios (OR) and p-values of the indicated variables analyzed with univariable or multivariable (with PSA adjustment) logistic regression analysis on the DP3 cohort. Med = number of medications; hypert = hypertension; CVD = cardiovascular disease; sedent = sedentariness; diab = diabetes.
Sedentary men had substantially higher recall rates than those who usually undertake moderate/high-level physical activity (p-value < 0.00001) and PSA was significantly lower in men used to regular physical activity (p-value < 0.001) compared to the others. The final diet score was negatively associated with PCa, but with low statistical significance (OR=0.80; p-value=0.035). As expected, PSA was significantly associated with PCa in the entire prospective cohort (Table 3), with an ROC AUC of 0.84 (C.I: = 0.82-0.87) (Figure 2A). Combination of age with PSA improved the discrimination (DeLong p-value = 0.0002), with an AUC of 0.87 (C.I. = 0.85-0.89) (Figure 2A).
ROC curves for the identification of PCa of any ISUP grade: PSA (red) and the logistic regression model that combines PSA and age (orange) for 904 patients (A) and for the subgroup of patients who underwent SLI (B). ROC curve for the identification of PCa of ISUP>1 among patients who underwent SLI (C): PSA (red) and the logistic regression model that combines PSA, age and CVD (blue). AUCs and their confidence intervals are reported, as well as the p-values of the DeLong test between curve pairs.
To test whether a model including the clinical covariates with significant association to PCa could improve the discriminatory ability of PSA and age, we considered the subset with complete information on PSA, age, hypertension, CVD, sedentariness and the number of medications (n=814). We applied penalized LASSO logistic regression modelling with 10-fold cross-validation and obtained 10 models, that were tested each time on the 1/10 of samples not used for model construction. The average ROC AUC on the 10 test sets was 0.89 (range: 0.84-0.93), while for PSA it was 0.85 (range: 0.81-0.91). PSA, age, hypertension and CVD were selected by 10/10 models and sedentariness by 9/10, indicating their relevance for the identification of men who really need SLI and for the implementation of targeted prevention actions.
When the same analysis was conducted including ISUP1 PCa in the reference set, we obtained an average ROC AUC on the 10 test sets of 0.89 (range: 0.83-0.94), while for PSA it was 0.85 (range: 0.79-0.94). Age and CVD were selected by 10/10 models, hypertension by 9/10, sedentariness by 6/10.
When we considered only the men who underwent SLI (n=523), the ROC AUC of PSA dropped to 0.59 (C.I. = 0.54-0.64) showing minimal predictive potential (Figure 2B). No covariate except PSA and age resulted associated with PCa in this subgroup in univariable and multivariable analysis. Instead, CVD was significantly associated with ISUP>1 PCa in multivariable analysis (p-value < 0.001) and with slightly lower significance in univariable analysis (p-value = 0.0021) (Table 2). The combination between age and PSA improved the discrimination of PCa of any ISUP grade from the rest, compared to PSA alone (DeLong p-value = 0.0091), and the inclusion of CVD allowed for a better identification of ISUP>1 PCa (DeLong p-value < 0.001). However, ROC AUCs were still poor (Figure 2B-C), indicating the need of stronger predictors.
Targeted sphingolipid analysis
Leveraging our previous findings on ceramide-like metabolites useful for the discrimination of PCa when PSA is in the so called grey-zone (4), we decided to perform targeted sphingolipid analysis. Fifty sphingolipids, including ceramides, dihydro-ceramides, hexosyl-ceramides, sphingomyelins, dihydro-sphingosine, sphingosine and its phosphorylated form, were analyzed on plasma samples using LC-MS/MS. All these molecules are characterized by an 18-carbon sphingoid base but harbor variable number of carbons in the fatty acids, ranging from 14 to 24. 46 sphingolipids were detected in all samples and were retained for the analysis.
Gb3.24, Cer.20, and Cer.24.1 were significantly associated with PCa of any grade and with ISUP>1 PCa, in univariable analysis on the samples of the DP3 cohorts.
LASSO regression modelling with 10-fold cross-validation was applied to the subset of men with complete information on circulating molecules, lifestyle and clinical data (n=369). The average ROC AUC of the 10 models in the test sets was 0.92 (range: 0.84-0.99), compared to 0.85 (range: 0.76-0.93) and 0.88 (range: 0.80-0.98) for PSA alone and PSA combined with age, respectively (Figure 3A). Hypertension, CVD, PSA, age and five sphingolipids (HexCer.20 and Cer.20 with positive coefficients; HexCer.24.1, GM3.24.1, DHCer.24 with negative coefficient) were selected in 10/10 models. PSA, age, CVD and Cer.20 were selected in 10/10 models also when the predicted outcome was ISUP>1 PCa. Figure 3B shows the predicted probabilities of harboring ISUP>1 PCa given by the multivariable models built using PSA, age, CVD and Cer.20, with predicted probabilities increasing with increasing ISUP grade. In Figure 3C, the samples are ordered by the predicted probabilities from one of the 10 models, which includes PSA, age, CVD, hypertension and 10 sphingolipids. Some ISUP >1 PCa cases (red dots) appear below the horizontal dashed line on the right side of the figure, indicating patients with high predicted probabilities despite low PSA levels.
(A) Classification performances of the multivariable models vs PSA alone in the entire cohort with complete information on sphingolipids and lifestyle/clinical variables. (B) Predicted probabilities of harboring a PCa of ISUP grade >1 given by the multivariable model built using or PSA, age, CVD and Cer.20. (C) Representation of logPSA, logAge, sphingolipid levels, hypertension and CVD state in each subject, ordered by the predicted probabilities of harboring an ISUP>1 PCa. Grey dots correspond to nonPCa, orange to ISUP1 PCa and red to ISUP>1 PCa. (D) Classification performances of the multivariable models vs PSA alone in the subgroup of men who underwent SLI.
The same approach was applied considering only the men who underwent SLI (n=221). In this case, only CVD, PSA and age together with five sphingolipids were selected by 10/10 models. Namely, SM.16 and HexCer.20 had positive coefficients, while HexCer.24.1, DHS1P, and DHCer.24 had negative coefficients. The improvement in the average ROC AUC of the 10 models on the 10 test sets not used for model construction was even better, from 0.65 (range: 0.56-0.83) and 0.70 (range: 0.56-0.87) for PSA and PSA combined with age, respectively, to 0.83 (range: 0.57-0.92) for the model including CVD and sphingolipids (Figure 3D).
Discussion
With the implementation of PCa screening policies across European countries, such as PRAISE-U (21), there will be an exponential rise in PSA testing. While this increase enhances early detection, it also raises the risk of dispensable second-level investigations (SLI), which can lead to over diagnosis, patient anxiety, and potential complications from invasive procedures like biopsies (22). To optimize screening benefits and minimize harm, it is crucial to apply risk-stratified approaches and follow evidence-based guidelines to ensure that only high-risk individuals undergo further diagnostic assessments.
In this work, we summarized the results of a prospective multicentre study addressed to men undergoing PSA testing and digital rectal examination, sometimes followed by SLI (MRI and/or prostate biopsy), for the early detection of PCa. The study also included men candidate to prostatectomy. The goal was to improve PSA diagnostic accuracy and reduce unnecessary MRIs/biopsies by integrating lifestyle, age, comorbidity, and easily measured blood markers, ultimately informing targeted prevention strategies.
PSA dosage remains the milestone for PCa detection, even if with its limitations. However, the predictive value of PSA is drastically reduced in those men with PCa suspicion, candidates for SLI; in this subgroup of subjects, only about 60% really have a confirmed PCa diagnosis, actually making necessary the identification of other complementary biomarkers.
One of the findings of our study is that a sedentary lifestyle and cardiovascular diseases (CVD) increase the likelihood of a PCa diagnosis. The association between CVD and PCa should not be considered in the same way that smoking is a risk factor for lung cancer, as both CVD and PCa share some common risk factors, such as age, diet, obesity and lack of exercise. However, a recent study highlighted a causative association between angina pectoris and PCa (23). We also found that combining sedentariness and CVD, along with PSA and age, improves patient selection for SLI. However, the key finding of our research is that analysis of plasma sphingolipids further enhances the precision in detecting PCa among men recalled for suspected PCa (average AUC 0.83). These results are comparable to risk calculators using MRI data (24) (AUC 0.81-0.87 for clinically significant PCa) and significantly better than models without MRI data (2) (AUC 0.66-0.79).
While pre-biopsy MRI in suspected PCa can help over 30% of men avoid immediate biopsy and improve cancer detection – identifying about 15% more clinically significant and 40% fewer clinically insignificant cancers (25) – it also increases the demand for MRI equipment, highly trained staff, manpower and (future) AI-based image analysis. Using a pre-MRI risk assessment tool in a clinical cohort could eliminate over a third of prostate MRIs (or biopsies, if MRI is unavailable) (26), with even greater potential in screening populations where the overall risk of PCa is much lower. Therefore, MRI should be prioritized for men at intermediate and high risk based on a reliable stratification tool (27). The drawback is a potential risk of missing some clinically significant PCas in men considered low-risk (26). In our cohort, the combination of clinical/demographic and laboratory covariates allowed the correct classification of ISUP>1 PCa also in some men with very low serum PSA levels.
The variables most strongly associated with PCa at enrolment were PSA levels, age, number of medications, hypertension, CVD, and sedentariness. Notably, sedentariness is a modifiable risk factor that can influence the other variables. Our observation of decreased PSA levels proportional to exercise frequency in a subgroup of sedentary men participating in a biweekly Nordic-walking program within the DP3 study (manuscript in preparation) further underscores this point. Exercise impacts overall health, including cardiovascular health, and can even influence biological age (28). These findings highlight the critical importance of stratifying patients for PCa screening and implementing risk-adapted strategies. By considering modifiable risk factors like sedentariness, alongside established factors such as PSA and age, we can more effectively identify individuals who would benefit most from further investigation, reducing unnecessary interventions in lower-risk individuals while ensuring timely diagnosis and treatment for those at higher risk. Critically, risk-adapted strategies could include the prescription of targeted exercise programs for sedentary men, addressing a key modifiable risk factor and potentially impacting both PSA levels and overall health.
Interest in circulating sphingolipids as cancer biomarkers has grown rapidly in recent years due to their role as not just structural components of cell membranes, but also active molecules involved in various cellular processes, including cell growth, differentiation, and death. Most of the previous studies on sphingolipids and PCa regard patients with existing PCa diagnoses and have identified sphingolipid signatures associated with cancer progression and therapy resistance (8–13). A prognostic blood-based sphingolipid panel has recently been proposed for men with localized PCa on active surveillance. This panel aims to identify patients at high risk of biopsy Gleason grade upgrading (29). To our knowledge, no blood-based sphingolipid panel exists for the early detection of PCa within a screening context, nor as a pre-MRI/biopsy risk assessment tool for men with suspected PCa.
As part of the abovementioned ancillary study within the DP3 project, we investigated the impact of a 6-month biweekly Nordic-Walking program (led by certified instructors) on quality of life, lifestyle, mobility parameters and circulating molecules in a subgroup of sedentary men enrolled in our study. Interestingly, DHCer24, which was negatively associated with PCa in all multivariable models, after six months proportionally increased with the frequency of activity (manuscript in preparation).
A limitation of this study is the lack of completeness regarding the different levels of information available for the subjects enrolled. We maximized the use of available data without imputing missing values. In multivariable modelling, we used a 10-fold cross-validation approach on each subset analyzed, to avoid overfitting and identify the most robust variables. We plan in the future to extend our analyses on wider datasets as well as to integrate plasma proteomic and urinary microbiome data to identify additional potential biomarkers and provide a more comprehensive understanding.
Conclusion
This study investigated the potential of combining readily available clinical data (PSA, age, hypertension, CVD, sedentariness) with novel circulating biomarkers for improved PCa detection. The research identified specific sphingolipid panels that, when combined with clinical factors, demonstrated strong diagnostic potential for both PCa screening and pre-MRI/biopsy risk assessment. These findings suggest that incorporating sphingolipid analysis could enhance PCa early detection and reduce dispensable invasive procedures.
Future research will expand the analysis to include plasma proteomic and urinary microbiome data for an even more comprehensive understanding and identification of additional biomarkers.
Data Availability
All data produced in the present study are available upon reasonable request to the authors
DECLARATIONS
Ethics approval and consent to participate
The observational clinical protocol called DP3 coordinated by FEET was approved by the Novara ethical committee (AOU Maggiore della Carità) in September 2020 (Prot. N° 968/CE 07/09/2020, integration Prot. N°263/CE, 10/03/2021), in accordance with the Declaration of Helsinki Ethical Principles for Medical Research involving Human Participants. The MH ethical committee (AOU Città della Salute e della Scienza) in Turin approved the DP3 study in January 2022 (Prot. N° 521/2021, 10/01/2022).
Consent for publication
NA
Availability of data and materials
Upon reasonable request to the corresponding author.
Competing interests
The authors declare that they have no competing interests
Funding
The DP3 study was funded by Compagnia di San Paolo (CP-N; FG); Rete Oncologica del Piemonte e della Valle d’Aosta; the National Plan for Complementary Investments to the NRRP, project “D34H—Digital Driven Diagnostics, prognostics and therapeutics for sustainable Health care” (project code: PNC0000001), Spoke 4 financed by the Italian Ministry of University and Research (MS); Fondazione CRT (CP-N; MM-G; PO; IG; GC).
Authors’ contributions (da adattare)
CPN: Conceptualization, Methodology, Investigation, Resources, Data Curation, Writing - Review & Editing. PO: Conceptualization, Software, Validation, Formal analysis, Data Curation, Writing - Review & Editing, Visualization. MS: Formal analysis, Investigation, Resources, Data Curation, Visualization. MMG: Conceptualization, Investigation, Resources, Writing - Review & Editing. IG: Investigation, Resources. FG: Software, Writing - Review & Editing. FC: Resources. FM: Resources. MDC: Investigation. RP: Supervision. AS: Resources. FF: Resources. NT: Resources. MO: Conceptualization, Resources, Writing - Review & Editing. AZ: Conceptualization, Resources. CF: Resources, Data Curation. AG: Resources. AN: Funding acquisition, Writing - Review & Editing. LC: Data Curation. PG: Resources, Supervision. SZ: Resources, Supervision. GC: conceptualization, methodology, Writing - Original Draft and Review & Editing, Supervision, Project Administration and Funding acquisition.
All authors read and approved the final manuscript
Authors’ information (optional)
Supplementary materials
Lifestyle score calculations
The reported weekly consumption frequencies (“never”, “once”, “2-3 times”, “more than 3 times”) for each food in the questionnaire were therefore assessed, in particular: “Pasta”, “Rice”, “White bread”, “Wholegrain bread”, “White meat”, “Red meat”, “Fish”, “Fried food”, “Cereals”, “Legumes”, “Fruit”, “Vegetables”, “Sweets”, “Eggs”, “Diary products”. To each food or combination of foods, a value between −1 and +1 was given according to the consumption frequencies recommended by the Italian Guidelines and the WCRF Recommendations. By adding the values attributed, an overall score for nutrition was calculated, ranging from 0 to 14.5, where higher scores indicated more adequate dietary habits, while lower scores an unhealthy diet. This score was further categorized into 5 classes: class 1 (0≤score≤7), class 2 (7<score≤8.75), class 3 (8.75<score≤10), class 4 (10<score≤11.5), class 5 (11.5<score≤14.5). Finally, the five scores thus created for BMI, physical activity, smoke, alcohol and diet were summed to obtain the final lifestyle score: higher lifestyle scores reflected a healthier global lifestyle, while lower scores were associated with worse lifestyle habits, predisposing to an increased risk of many pathologies.
Acknowledgements
We thank all patients who agreed to participate in the DP3 study, as well as all the nurses, physicians and students supporting sample collection and processing.
References
