Application of machine learning algorithms to predict osteoporosis in postmenopausal women with type 2 diabetes mellitus

PURPOSE

The screening and diagnosis of osteoporosis in patients with type 2 diabetes mellitus (T2DM) based on bone mineral density remains challenging because of the limited availability and accessibility of dual-energy X-ray absorptiometry. We aimed to develop and validate models to predict the risk of osteoporosis in postmenopausal women with T2DM based on machine learning (ML) algorithms.

METHODS

This retrospective study included 303 postmenopausal women with T2DM. To develop prediction models for osteoporosis, we applied nine ML algorithms combined with demographic, clinical, and laboratory data. The least absolute shrinkage and selection operator were used to perform feature selection. We used the bootstrap resampling technique for model training and validation. To test the performance of the models, we calculated indices including the area under the receiver operating characteristic curve (AUROC), accuracy, sensitivity, specificity, positive predictive value, negative predictive value, F1 score, calibration curve, and decision curve analysis. Furthermore, we conducted fivefold cross-validation for parameter optimization and model validation. Feature importance was assessed using the SHapley additive explanation (SHAP).

RESULTS

We identified 10 independent predictors as the most valuable features. An AUROC of 0.616-1.000 was observed for nine ML algorithms. The extreme gradient boosting (XGBoost) model exhibited the best performance, outperforming conventional risk assessment tools and registering 0.993 in the training set, 0.798 in the validation set, and 0.786 in the test set for fivefold cross-validation. Using SHAP, we found that the explanatory variables contributed to the model and their relationship with osteoporosis occurrence. Furthermore, we developed a user-friendly tool for calculating the risk of osteoporosis.

CONCLUSIONS

With the integration of demographic and clinical risk factors, ML algorithms can accurately predict osteoporosis. The XGBoost model showed ideal performance. With the incorporation of these models in the clinic, patients may benefit from early osteoporosis diagnosis and treatment.

Novel role of estrogen receptor-α on regulating chondrocyte phenotype and response to mechanical loading

OBJECTIVE

In knee cartilage from patients with osteoarthritis (OA), both preserved cartilage and damaged cartilage are observed. In this study, we aim to compare preserved with damaged cartilage to identify the molecule(s) that may be responsible for the mechanical loading-induced differences within cartilage degradation.

METHODS

Preserved and damaged cartilage were harvested from the same OA knee joint. RNA Sequencing was performed to examine the transcriptomic differences between preserved and damaged cartilage cells. Estrogen receptor-α (ERα) was identified, and its function of was tested through gene knockin and knockout. The role of ERα in mediating chondrocyte response to mechanical loading was examined via compression of chondrocyte-laded hydrogel in a strain-controlled manner. Findings from the studies on human samples were verified in animal models.

RESULTS

Level of estrogen receptor α (ERα) was significantly reduced in damaged cartilage compared to preserved cartilage, which were observed in both human and mice samples. Knockdown of ESR1, the gene encoding ERα, resulted in an upregulation of senescence- and OA-relevant markers in chondrocytes. Conversely, knockin of ESR1 partially reversed the osteoarthritic and senescent phenotype of OA chondrocytes. Using a three-dimensional (3D) culture model, we demonstrated that mechanical overload significantly suppressed ERα level in chondrocytes with concomitant upregulation of osteoarthritic phenotype. When ESR1 expression was suppressed, mechanical loading enhanced hypertrophic and osteogenic transition.

CONCLUSION

Our study demonstrates a new estrogen-independent role of ERα in mediating chondrocyte phenotype and its response to mechanical loading, and suggests that enhancing ERα level may represent a new method to treat osteoarthritis.

Current assessment of parental and health professional perception of the colour of neonatal vomiting: Results of a scoping survey

PURPOSE

To determine current perceptions of doctors, nurses and parents for the colour of a neonatal vomit which should prompt an urgent surgical review.

METHODS

A voluntary scoping survey of parents/guardians of patients and non-surgical healthcare professionals was conducted with respondents asked to choose from 8 different selections in a colour swatch from pale yellow to dark green. A control group consisted of 13 paediatric surgeons. Data were analysed using the paired t test, Fishers exact test. A p value of < 0.05 was considered to be significant.

RESULTS

365 participants responded: 36% (131/365) parents, 18% (64/365) nurses and 46% (166/365) doctors. 4/365 (1%) did not state their role. 343 participants completed all questions and responses were analysed using total responses for each question. 82% (121/148) of doctors and 78% (50/64) of nurses had more than 3 years of post-graduate experience. Overall, 63% (227/361) of participants (100% paediatric surgeons, 78% other doctors, 75% nurses/midwives & 30% parents) considered dark and light green vomits to be a sign of intestinal obstruction. 67% (242/361) of participants (100% paediatric surgeons, 72% other doctors, 56% nurses/midwives and 62% parents) believed dark and light green vomiting needed an urgent surgical referral. There were significant differences between the control group and other groups in terms of whether the neonate could wait until the next day for a review; nursing staff (p = 0.0002), postnatal/midwifery (p =  < 0.0001), emergency medicine (p = 0.04), general practice (p = 0.002), neonatal (p = 0.0001) and paediatricians (p = 0.005). Only the neonatologists (p = 0.04), nursing staff (p = 0.001) and postnatal/midwifery (p = 0.004) believed that the neonate could have safe observation.

CONCLUSION

Although the perception that green vomiting is potentially serious is acknowledged by the majority of healthcare professionals surveyed, there is still a requirement for more targeted educational practices in nursing, midwifery and medical staff.

Discontinuation Patterns and Cost Avoidance of a Pharmacist-Driven Methicillin-Resistant Staphylococcus aureus Nasal Polymerase Chain Reaction Testing Protocol for De-escalation of Empiric Vancomycin for Suspected Pneumonia

A pharmacist-driven methicillin-resistant Staphylococcus aureus (MRSA) nasal polymerase chain reaction (PCR)-based testing protocol with a 70% acceptance rate for vancomycin discontinuation within 24 hours of negative results significantly reduced unnecessary vancomycin use with an estimated cost avoidance of $40 per vancomycin course. We found high concordance (141 of 147, 96%) of culture-based versus PCR-based MRSA nasal screening.

Perioperative intravenous contrast administration and the incidence of acute kidney injury after major gastrointestinal surgery: prospective, multicentre cohort study

BACKGROUND

This study aimed to determine the impact of preoperative exposure to intravenous contrast for CT and the risk of developing postoperative acute kidney injury (AKI) in patients undergoing major gastrointestinal surgery.

METHODS

This prospective, multicentre cohort study included adults undergoing gastrointestinal resection, stoma reversal or liver resection. Both elective and emergency procedures were included. Preoperative exposure to intravenous contrast was defined as exposure to contrast administered for the purposes of CT up to 7 days before surgery. The primary endpoint was the rate of AKI within 7 days. Propensity score-matched models were adjusted for patient, disease and operative variables. In a sensitivity analysis, a propensity score-matched model explored the association between preoperative exposure to contrast and AKI in the first 48 h after surgery.

RESULTS

A total of 5378 patients were included across 173 centres. Overall, 1249 patients (23·2 per cent) received intravenous contrast. The overall rate of AKI within 7 days of surgery was 13·4 per cent (718 of 5378). In the propensity score-matched model, preoperative exposure to contrast was not associated with AKI within 7 days (odds ratio (OR) 0·95, 95 per cent c.i. 0·73 to 1·21; P = 0·669). The sensitivity analysis showed no association between preoperative contrast administration and AKI within 48 h after operation (OR 1·09, 0·84 to 1·41; P = 0·498).

CONCLUSION

There was no association between preoperative intravenous contrast administered for CT up to 7 days before surgery and postoperative AKI. Risk of contrast-induced nephropathy should not be used as a reason to avoid contrast-enhanced CT.

FRI0450 MEASURES OF DISEASE SEVERITY PREDICT DISABILITY AND QUALITY OF LIFE DIFFERENTLY IN RHEUMATOID ARTHRITIS AND CHRONIC CHIKUNGUNYA DISEASE

Background:

Chronic rheumatological manifestations similar to those of rheumatoid arthritis (RA) have been described after chikungunya virus infection. However, the clinical significance of the symptoms and disease severity in the two conditions has not been directly compared.

Objectives:

To compare, using identical measures of disease severity and patient outcomes, the impact of disease severity measures and symptoms on outcomes in RA and chronic chikungunya disease.

Methods:

Forty patients with chronic chikungunya arthralgia two years post-infection and 40 matched patients with RA were enrolled in Roraima, Brazil. Twenty-eight joints were assessed for tenderness and swelling, a pain intensity visual analogue scale, musculoskeletal stiffness questionnaire, modified Health Assessment Questionnaire and the EuroQol EQ5D-5L quality of life assessment were completed. The importance of the various measures of disease severity were analysed using Spearman’s rank correlation and regression analysis.

Results:

Tender and swollen joint counts, pain and stiffness were all predictive of the HAQ disability index in RA, but only stiffness was significantly associated with disability in chikungunya patients (Table 1). Tender and swollen joint counts, pain and stiffness were predictive for all EQ5D quality of life domains (except anxiety/depression) in RA patients. In contrast, in chikungunya disease, tender joint counts were predictive only of usual daily activities; pain was predictive of impaired mobility, self-care and discomfort, while stiffness was predictive for the mobility and anxiety/depression domains (Figure 1). Swollen joint counts were not associated with any of the patient outcomes in chikungunya disease. Linear regression analysis confirmed (p=0.003) that the effect of swollen joint count on the HAQ disability index depends on the underlying disease.

Table 1.

Association of disease severity with HAQ disability index in rheumatoid and CHIKV+ arthritis

Severity measureRheumatoid arthritisCHIKV+ arthritisr (p)r (p)Tender joint count0.56 (0.0002)0.24 (0.14)Swollen joint count0.60 (<0.0001)0.002 (0.99)Joint pain (VAS)0.55 (0.0002)0.29 (0.07)Stiffness severity0.57 (0.0001)0.38 (0.02)Figure 1.

Association of disease severity with quality of life domains in rheumatoid and CHIKV+ arthritis

Conclusion:

The value of all the disease severity measures tested in RA were confirmed, but tender joint counts may have more limited value in the assessment of chronic chikungunya disease. Joint swelling appears to have little impact for chikungunya patients, while stiffness appears to be an important metric to quantify chikungunya arthritis disease severity.

Disclosure of Interests:

Hugh Watson Shareholder of: Sanofi, Employee of: Sanofi, Ramão Luciano Nogueira-Hayd: None declared, Maony Rodrigues-Moreno: None declared, Felipe Naveca: None declared, Giulia Calusi: None declared, Richard Amdur: None declared, Karol Suchowiecki: None declared, Gary S. Firestein: None declared, Gary Simon: None declared, Aileen Chang: None declared

0303 Heart Rate and Systolic Blood Pressure Increase During Experimental Sleep Restriction

Introduction

Experimental sleep restriction is associated with elevated daytime cardiac activity, including heart rate (HR) and blood pressure (BP). However, some studies have found changes in systolic (SBP) but not diastolic blood pressure (DBP) or found changes in neither. Although findings are mixed, there may be a dose-response effect of cumulative sleep loss on daytime cardiac activity, such that HR and BP increase above basal levels with additional nights of insufficient sleep. This study examined changes in cardiac activity during experimental sleep restriction.

Methods

We used multilevel models with random effects for individuals to analyze data from 15 healthy males (M=22.3 years old, SD=2.8) in an 11-day inpatient protocol consisting of three nights of 10-hour/night baseline sleep opportunity, five nights of sleep restriction (5-hour/night sleep opportunity), and then two recovery nights (10-hour/night sleep opportunity). HR and BP were measured approximately every two hours during wake.

Results

HR increased 0.75 beats/minute with each successive night of sleep restriction (SE=0.18, p&lt;0.001). HR was 5.13 beats/minute higher during the recovery condition than during baseline or sleep restriction (SE=1.05, p&lt;0.001). During sleep restriction only, HR was lower in the later morning and evening compared to the earliest morning timepoint of the day, F(10, 743)=10.44, p&lt;0.001. SBP increased 0.33 mmHg following each successive night of sleep restriction (SE=0.16, p=0.041); however, SBP was only marginally higher during the sleep restriction condition than during baseline (b=1.90, SE=1.09, p=0.082).

Conclusion

Our findings suggest that HR and SBP increase with each additional day of experimental sleep restriction, even after accounting for diurnal effects on HR and SBP. HR did not recover to baseline levels following a night of recovery sleep, suggesting that longer recovery sleep may be necessary to recover from a week of sleep restriction.

Support

Grant UL1TR000127 (Chang PI), Clinical and Translational Science Institute; College of Health and Human Development at Pennsylvania State University.

0296 Less Self-Reported Alertness and Motivation During Sleep Restriction are Associated with Decreased Attentional Performance

Introduction

Some individuals demonstrate more performance decrements on the psychomotor vigilance task (PVT) after sleep restriction (SR). We investigated whether individuals who reported less alertness and/or less motivation after SR demonstrated poorer performance on the PVT.

Methods

Fifteen healthy men (22.3±2.8 years) participated in a 10-night inpatient protocol with three nights of 10-hour baseline time in bed (TIB), five nights of SR (5-hour TIB), then two recovery (10-hour TIB) nights. Participants completed the 10-minute PVT (Joggle Research® battery) approximately every two hours during wake. Outcomes included number of false starts (&lt;100 ms reaction time, RT) and number of lapses (≥500 ms RT). Participants reported alertness and motivation levels after each PVT. Median splits were used to characterize changes in alertness (“sleepy,” n=8, versus “alert,” n=7) and motivation (“unmotivated,” n=7, versus “motivated,” n=8) from the last day of baseline to the last day of SR. Outcomes were analyzed in mixed models with the predictor day*alertness or day*motivation, excluding the first three baseline days to preclude practice effects.

Results

There were significant interactions between day and alertness (p=.025) and day and motivation (p=.043) for false starts. False starts followed a quadratic inverted-U shape across days in sleepy (b=-0.16, p=.003) and unmotivated (b=-0.16, p=.004) participants, but not in alert or motivated participants (p&gt;.05). There was a significant interaction between day and alertness for lapses (p=.008); lapses followed a quadratic inverted-U shape across days with a stronger effect in sleepy (b=-0.43, p&lt;.001) versus alert (b=-0.15, p=.031) participants. There was no interaction between day and motivation for lapses.

Conclusion

Participants reporting less alertness were more likely to make both false starts and lapses after SR; those reporting less motivation were more likely to make false starts, but not lapses. Findings suggest greater motivation is sufficient to preserve inhibitory control but not vigilance after sleep restriction. In contrast, greater alertness despite sleep restriction was sufficient to preserve inhibitory control and resulted in lower vigilance decrements.

Support

This study was funded by grant UL1TR000127 from the Clinical and Translational Science Institute and the College of Health and Human Development at the Pennsylvania State University (Chang PI).

0085 Vulnerability to Sleep Restriction is Associated with Decreased Working Memory Performance

Introduction

We investigated whether individuals with more lapses on the psychomotor vigilance task (PVT) after sleep restriction (SR) demonstrated poorer working memory compared to those with fewer PVT lapses.

Methods

Fifteen healthy men (22.3±2.8 years) participated in a 10-night inpatient protocol with three nights of 10-hour baseline time in bed (TIB), five nights of SR (5-hour TIB), then two recovery (10-hour TIB) nights. Participants completed the Visual Object Learning Task (VOLT) and Fractal 2-Back (F2B; visual n-back) measuring working memory and the PVT (Joggle Research® battery) approximately every two hours during wake. During the VOLT, participants indicated whether presented images had been shown previously. Outcomes included number of misses and false alarms. During the F2B, participants tapped the screen when an image appeared that had been shown 2 images previously. Outcomes included sensitivity and specificity. Median split of mean PVT lapses after the last night of SR was used to categorize participants into “vulnerable” (n=8) versus “resistant” (n=7) groups. Outcomes were analyzed in mixed models with the predictor day*vulnerability, excluding the first three baseline days to preclude practice effects.

Results

There was a significant interaction between day and attentional vulnerability for VOLT misses (p&lt;.001); misses increased linearly across days in vulnerable (b=.18, p&lt;.001) but not resistant (p=.956) participants. There was no interaction between day and vulnerability for VOLT false alarms, which did not change across days. There was a significant interaction between day and attentional vulnerability for F2B sensitivity (p=.002); sensitivity increased linearly across days in resistant (b=.02, p&lt;.001) but not in vulnerable (p=.273) participants. There was no interaction between day and vulnerability for F2B specificity, which did not change across days.

Conclusion

Performance on the VOLT decreased in vulnerable participants only; performance on the F2B improved in resistant participants likely due to practice effects not seen in vulnerable participants. Findings indicate vulnerability to attentional lapses after SR is a marker of vulnerability to working memory decrements.

Support

This study was funded by grant UL1TR000127 from the Clinical and Translational Science Institute (Chang PI) and the College of Health and Human Development at the Pennsylvania State University.