Toward a validated diagnostic test with machine learning algorithm for interstitial cystitis

Risk Classification for Interstitial Cystitis/Bladder Pain Syndrome Using Machine Learning Based Predictions

OBJECTIVE

To improve diagnosis of interstitial cystitis (IC)/bladder pain syndrome(IC) we hereby developed an improved IC risk classification using machine learning algorithms.

METHODS

A national crowdsourcing resulted in 1264 urine samples consisting of 536 IC (513 female, 21 male, 2 unspecified), and 728 age-matched controls (318 female, 402 male, 8 unspecified) with corresponding patient-reported outcome (PRO) pain and symptom scores. In addition, 296 urine samples were collected at three academic centers: 78 IC (71 female, 7 male) and 218 controls (148 female, 68 male, 2 unspecified). Urinary cytokine biomarker levels were determined using Luminex assay. A machine learning predictive classification model, termed the Interstitial Cystitis Personalized Inflammation Symptom (IC-PIS) Score, that utilizes PRO and cytokine levels, was generated and compared to a challenger model.

RESULTS

The top-performing model using biomarker measurements and PROs (area under the curve [AUC]=0.87) was a support vector classifier, which scored better at predicting IC than PROs alone (AUC=0.83). While biomarkers alone (AUC=0.58) did not exhibit strong predictive performance, their combination with PROs produced an improved predictive effect.

CONCLUSION

IC-PIS represents a novel classification model designed to enhance the diagnostic accuracy of IC/bladder pain syndrome by integrating PROs and urine biomarkers. The innovative approach to sample collection logistics, coupled with one of the largest crowdsourced biomarker development studies utilizing ambient shipping methods across the US, underscores the robustness and scalability of our findings.

Temporally complex inflammatory networks in an animal model reveal signatures for interstitial cystitis and bladder pain syndrome phenotype

INTRODUCTION AND OBJECTIVE

Interstitial cystitis and bladder pain syndrome (IC/BPS) presents with symptoms of debilitating bladder pain and is typically a diagnosis of exclusion. The cystoscopic detection of Hunner's lesions increases the likelihood of detecting tissue inflammation on bladder biopsy and increases the odds of therapeutic success with anti-inflammatory drugs. However, the identification of this subgroup remains challenging with the current lack of surrogate biomarkers of IC/BPS. On the path towards identifying biomarkers of IC/BPS, we modeled the dynamic evolution of inflammation in an experimental IC/BPS rodent model using computational biological network analysis of inflammatory mediators (cytokines and chemokines) released into urine. The use of biological network analysis allows us to identify urinary proteins that could be drivers of inflammation and could therefore serve as therapeutic targets for the treatment of IC/BPS.

METHODS

Rats subjected to cyclophosphamide (CYP) injection (150 mg/kg) were used as an experimental model for acute IC/BPS (n = 8). Urine from each void was collected from the rats over a 12-h period and was assayed for 13 inflammatory mediators using Luminex™. Time-interval principal component analysis (TI-PCA) and dynamic network analysis (DyNA), two biological network algorithms, were used to identify biomarkers of inflammation characteristic of IC/BPS over time.

RESULTS

Compared to vehicle-treated rats, nearly all inflammatory mediators were elevated significantly (p < 0.05) in the urine of CYP treated rats. TI-PCA highlighted that GRO-KC, IL-5, IL-18, and MCP-1 account for the greatest variance in the inflammatory response. At early time points, DyNA indicated a positive correlation between IL-4 and IL-1β and between TNF-α and IL-1β. Analysis of TI-PCA and DyNA at later time points showed the emergence of IL-5, IL-6, and IFNγ as additional key mediators of inflammation. Furthermore, DyNA network complexity rose and fell before peaking at 9.5 h following CYP treatment. This pattern of inflammation may mimic the fluctuating severity of inflammation associated with IC/BPS flares.

CONCLUSIONS

Computational analysis of inflammation networks in experimental IC/BPS analysis expands on the previously accepted inflammatory signatures of IC by adding IL-5, IL-18, and MCP-1 to the prior studies implicating IL-6 and GRO as IC/BPS biomarkers. This analysis supports a complex evolution of inflammatory networks suggestive of the rise and fall of inflammation characteristic of IC/BPS flares.

Performance of urinary biomarkers in differentiating dysfunctional voiding in women with overactive bladder syndrome: a prospective pilot study

PURPOSE

Dysfunctional voiding (DV) in women is a common disorder that causes bladder outlet obstruction (BOO) and may aggravate overactive bladder (OAB) symptoms. The diagnosis of DV relies on clinical alertness and videourodynamic study (VUDS). In this pilot study, we aimed to explore the diagnostic potential of urinary biomarkers in women with DV.

METHODS

Women who visited our institute for the management of OAB were enrolled in this prospective study. Urine samples were collected before treatment and urinary proteins were quantified using enzyme-linked immunosorbent assay (ELISA) kits. Patients with DV and patients with normal tracings were included in the study and control group, respectively. The patients with DV were further divided into two subgroups: those with and those without detrusor overactivity (DO). VUDS parameters and urinary protein levels were compared between each subgroup.

RESULTS

A total of 67 women were enrolled. There were 47 patients with VUDS-diagnosed DV and 20 patients in the control group. Urinary 8-OHdG level (32.65 vs. 15.59, p < 0.001) and urinary TNF-α level (1.43 vs. 0.83, p = 0.031) were significantly higher in the DV group, and urinary IL-2 level (0.29 vs. 0.83, p < 0.001) was significantly lower in the DV group. Among the DV patients, urinary PGE2 level was higher in the DO group (240.4 vs 149.6, p = 0.049).

CONCLUSIONS

In women with DV, urinary 8-OHdG level is elevated compared to women with normal VUDS. Urinary PGE2 level is elevated in patients with DV and DO. These finding suggests increased reactive-oxidative stress response and detrusor hyperactivity in DV patients. Urinary biomarkers can serve as potential non-invasive diagnostic tools in women with OAB.

Using social media to crowdsource collection of urine samples during a national pandemic

The COVID-19 pandemic and subsequent lockdown had a substantial impact on normal research operations. Researchers needed to adapt their methods to engage at-home participants. One method is crowdsourcing, in which researchers use social media to recruit participants, gather data, and collect samples. We utilized this method to develop a diagnostic test for Interstitial Cystitis/Bladder Pain Syndrome (IC/BPS). Participants were recruited via posts on popular social-media platforms, and enrolled via a website. Participants received and returned a mail kit containing bladder symptom surveys and a urine sample cup containing room-temperature preservative. Using this method, we collected 1254 IC/BPS and control samples in 3 months from all 50 United States. Our data demonstrate that crowdsourcing is a viable alternative to traditional research, with the ability to reach a broad patient population rapidly. Crowdsourcing is a powerful tool for at-home participation in research, particularly during the lockdown caused by the COVID-19 pandemic.