Dosing Algorithm Research Articles

Abstract 13308: Characteristics of Participants Enrolled in a Randomized Controlled Trial Evaluating Vitamin D Normalization on Major Adverse Cardiovascular-Related Events: The Target-D Trial

Background: Many observational studies have reported associations between low 25-hydroxyvitamin D (25[OH] vit D) levels and adverse cardiovascular (CV) outcomes. Recent randomized clinical trials have not found vitamin (vit) D 3 supplementation to reduce CV risk. However, these trials gave “blanket” doses regardless of 25[OH] vit D level, with none having "treated to target". We are performing the first study where subjects are prescribed vit D 3 supplementation based on individual 25[OH] vit D levels. Methods: The purpose of Target-D (NCT: 02996721) is to determine if achieving 25[OH] vit D levels >40 ng/mL among myocardial infarction (MI) patients will result in a reduction of adverse CV events. Subjects meeting study criteria from April 2017 to May 2023 were randomized to receive standard of care or clinical management of 25[OH] vit D levels. The treatment arm received targeted vit D 3 supplementation based on a dosing algorithm. Target-D is an event-driven trial that will conclude when 104 primary outcome events (composite of death, MI, heart failure hospitalization, and stroke) have occurred. Results: The study consort diagram is shown in Figure. Patient characteristics are shown in Table. Baseline 25[OH] vit D levels were 26.8±12.7 (median: 25) and were similar between arms (Table). Among those randomized to the treatment arm with a 25[OH] vit D levels < 40, 58.5% of patients began vit D 3 dosing at 5000 IU. Conclusions: Average baseline 25[OH] vit D levels were deficient in Target-D participants. The discrepancy in findings between observational and randomized trials of vit D may be due to the inability of vit D 3 supplementation doses to achieve therapeutic 25[OH] vit D thresholds. Target-D will evaluate if it is the level of serum 25[OH] vit D achieved that provides CV benefit and not necessarily the vit D 3 supplementation dose that is received. It is anticipated that the number of events needed for this event-driven trial will be achieved by May 2024.

Effect of Obesity on the Pharmacokinetics and Pharmacodynamics of Anticancer Agents.

An objective of the Precision Medicine Initiative, launched in 2015 by the US Food and Drug Administration and National Institutes of Health, is to optimize and individualize dosing of drugs, especially anticancer agents, with high pharmacokinetic and pharmacodynamic variability. The American Society of Clinical Oncology recently reported that 40% of obese patients receive insufficient chemotherapy doses and exposures, which may lead to reduced efficacy, and recommended pharmacokinetic studies to guide appropriate dosing in these patients. These issues will only increase in importance as the incidence of obesity in the population increases. This publication reviews the effects of obesity on (1) tumor biology, development of cancer, and antitumor response; (2) pharmacokinetics and pharmacodynamics of small-molecule anticancer drugs; and (3) pharmacokinetics and pharmacodynamics of complex anticancer drugs, such as carrier-mediated agents and biologics. These topics are not only important from a scientific research perspective but also from a drug development and regulator perspective. Thus, it is important to evaluate the effects of obesity on the pharmacokinetics and pharmacodynamics of anticancer agents in all categories of body habitus and especially in patients who are obese and morbidly obese. As the effects of obesity on the pharmacokinetics and pharmacodynamics of anticancer agents may be highly variable across drug types, the optimal dosing metric and algorithm for difference classes of drugs may be widely different. Thus, studies are needed to evaluate current and novel metrics and methods for measuring body habitus as related to optimizing the dose and reducing pharmacokinetic and pharmacodynamic variability of anticancer agents in patients who are obese and morbidly obese.

Generalisation of radiotherapy dose calculation for Monte Carlo algorithm combined with 3D Swin-Unet: a multi-institutional IMRT evaluation

Objective. Accurate dose calculations are essential prerequisites for precise radiotherapy. The integration of deep learning into dosimetry could consider computational accuracy and efficiency and has potential applicability to clinical dose calculation. The generalisation of a deep learning dose calculation method (hereinafter referred to as TERMA-Monte Carlo network, T-MC net) was evaluated in clinical practice using intensity-modulated radiotherapy (IMRT) plans for various human body regions and multiple institutions, with the Monte Carlo (MC) algorithm serving as a benchmark. Approach. Sixty IMRT plans were selected from four institutions for testing the head and neck, chest and abdomen, and pelvis regions. Using the MC results as the benchmark, the T-MC net calculation results were used to perform three-dimensional dose distribution and dose-volume histogram (DVH) comparisons of the entire body, planning target volume (PTV) and organs at risk (OARs), respectively, and calculate the mean ±95% confidence interval of gamma pass rate (GPR), percentage of agreement (PA) and dose difference ratio (DDR) of dose indices D95, D50, and D5. Main results. For the entire body, the GPRs of 3%/3 mm, 2%/2 mm, 2%/1 mm, and the PA were 99.62 ± 0.32%, 98.50 ± 1.09%, 95.60 ± 2.90% and 97.80 ± 1.12%, respectively. For the PTV, the GPRs of 3%/3 mm, 2%/2 mm, 2%/1 mm and the PA were 98.90 ± 1.00%, 95.78 ± 2.83%, 92.23 ± 4.74% and 98.93 ± 0.62%, respectively. The absolute value of average DDR was less than 1.4%. Significance. We proposed a general dose calculation framework based on deep learning, using the MC algorithm as a benchmark, performing a generalisation test for IMRT treatment plans across multiple institutions. The framework provides high computational speed while maintaining the accuracy of MC and may become an effective dose algorithm engine in treatment planning, adaptive radiotherapy, and dose verification.

Commissioning of Mobius3D/FX for patient-specific quality assurance: The CUIMC-NewYork Presbyterian Hospital experience.

To present the process undertaken by our institute in commissioning Mobius3D (M3D) for patient-specific quality assurance. 168 plans were randomly selected to compare dose distribution measured with ArcCheck and dose calculated from M3D, both compared against the treatment planning system (TPS). The gamma criteria for measurement and M3D are 3%/2mm with 10% and 50% dose thresholds, respectively. The effect of tissue inhomogeneity was investigated on 11 plans by recalculating the dose in M3D on a homogeneous phantom. Tolerance and action limits were established following the AAPM Task Group 218 recommendations. The M3D showed high variability in gamma passing rate compared to the measurement. Twenty-three out of 168 plans had false negative dose comparisons. These plans fall under high tissue inhomogeneity like lung and metal implants, small field targets, and breast plans planned with high energy. One false negative case (0.6%) was observed. A single tolerance limit of 91% and 92% gamma passing rate for the M3D and measurement-based PSQA were established, respectively. Against the expectation, recalculating plans on the homogeneous phantom in M3D did not necessarily increase the gamma passing rate. These plans have a duty cycle >4.2, and the small field sizes combined with differences in slice thickness contributed to observed dose differences in the homogeneous phantom comparisons. Following the commissioning, M3Dis adopted in our institute. Currently, the gamma criteria used for measurement and M3Dare 3%/2mm, 40% dose threshold, with gamma passing rates of 92% and 95%, respectively. A higher passing rate for M3D is adopted until more data is available. The combined effect of plan modulation, the field sizes, the tissue inhomogeneity, the dose algorithm, and the volume averaging effect from differences in slice thickness can contribute to the differences in dose in M3Dand TPS.

The effect of the modified fat-protein unit algorithm compared with that of carbohydrate counting on postprandial glucose in adults with type-1 diabetes when consuming meals with differing macronutrient compositions: a randomized crossover trial

BackgroundThe optimization of glucose control in type-1 diabetes is challenged by postprandial glycemic variability. This study aimed to compare the postprandial glycemic effects of carbohydrate counting and the modified fat-protein unit (FPU) algorithms following meals with different protein and fat emphases in adults with type-1 diabetes.MethodsThirty adults with type-1 diabetes aged 18 to 45 years participated in a randomized crossover trial. In a random order, participants consumed four test meals with equivalent energy and different macronutrient emphases on four separate mornings. The modified FPU algorithms and carbohydrate counting were used to determine the insulin dose for the test meals. A continuous glucose monitoring system was used to measured postprandial glycemia.ResultsCompared with carbohydrate counting, the modified FPU algorithm significantly decreased the late postprandial mean glucose levels (p = 0.026) in high protein-fat meals. The number of hypoglycemia episodes was similar between insulin dosing algorithms for the high protein-fat meals; hypoglycemic events were considerably higher for the modified FPU in the normal protein-fat meal (p = 0.042).ConclusionsThe modified FPU algorithm may improve postprandial glycemic control after consuming high protein-fat meals in adults with type-1 diabetes but may result in increased hypoglycemia risk when used with a normal protein-fat meal.

FRI654 Concentrated Regular U-500 Insulin In An Automated Insulin Delivery System: A Case Report

Abstract Disclosure: J. Aurora: None. A. Saraswat: None. Background: Concentrated regular U-500 (U-500R) insulin is commonly used in patients with T2DM who are highly insulin-resistant, requiring more than 200 units of insulin per day. Utilizing U-500R reduces the number of daily injections, decreases injection volume, and improves cost effectiveness. There have been clinical trials using U-500R for continuous subcutaneous insulin infusion (CSII), particularly in Omnipod which has been shown to improve glycemic control and patient satisfaction without an increase in hypoglycemia (1). In addition, U-500R administration via multiple daily injections (MDI) and an investigational U-500R CSII, have been compared and demonstrated improved glycemic control with both modalities, but an increase in nocturnal hypoglycemia with CSII (2). Automated insulin delivery (AID) systems, or hybrid closed loop systems, have not formally been studied in patients with T2DM using U-500R. We present a case of a patient with T2DM with high insulin resistance using U-500R with Omnipod 5 AID system. Clinical Case: A 54-year-old female with T2DM with high insulin resistance was initially on Omnipod Dash with U-100 insulin lispro with baseline A1c 12.1%. She was transitioned to Omnipod 5 in 2022, and over 4 months A1c improved to 8.8%, GMI 8.8%, TIR 38%, TAR 62%, TBR 0%, CV 35.9%, and automated mode 97%. However, her total daily dose was 134 units leading to pod changes every 1.5 days. She expressed interest in switching to U-500R to reduce pod changes while utilizing her current CSII technology. We switched to U-500R, continued automated mode, increased her 24-hour BG target from 120 mg/dL to 130 mg/dL to reduce hypoglycemia risk, and increased her active insulin time from 3 to 6 hours to reflect U-500R kinetics. After 4 weeks, her GMI improved to 8.0%, TIR 47%, TAR 52%, TBR 1%, CV 45.9%, and automated mode 94%. Although nocturnal BG readings were normal, ranging from 70-130 mg/dL, she was having nightly symptoms of relative hypoglycemia. We then adjusted her BG target overnight to 150 mg/dL, which kept BG between 80-130 mg/dL with resolution of her symptoms. Conclusion: U-500R proved to be safe and effective with Omnipod 5 AID. Despite setting a higher BG target, the algorithm continuously kept BG well below this, particularly overnight. However, the safety mechanisms incorporated with the CSII suspended insulin delivery for predicted hypoglycemia. Further research will be required to adapt the automated dosing algorithm to U-500R to reduce the risk of nocturnal hypoglycemia.

Comparative analysis of dose calculation algorithms for CyberKnife-based stereotactic radiotherapy in lung cancer.

The accuracy of dose calculation is the prerequisite for CyberKnife (CK) to implement precise stereotactic body radiotherapy (SBRT). In this study, CK-MLC treatment planning for early-stage non-small cell lung cancer (NSCLC) were compared using finite-size pencil beam (FSPB) algorithm, FSPB with lateral scaling option (FSPB_LS) and Monte Carlo (MC) algorithms, respectively. We concentrated on the enhancement of accuracy with the FSPB_LS algorithm over the conventional FSPB algorithm and the dose consistency with the MC algorithm. In this study, 54 cases of NSCLC were subdivided into central lung cancer (CLC, n=26) and ultra-central lung cancer (UCLC, n=28). For each patient, we used the FSPB algorithm to generate a treatment plan. Then the dose was recalculated using FSPB_LS and MC dose algorithms based on the plans computed using the FSPB algorithm. The resultant plans were assessed by calculating the mean value of pertinent comparative parameters, including PTV prescription isodose, conformity index (CI), homogeneity index (HI), and dose-volume statistics of organs at risk (OARs). In this study, most dose parameters of PTV and OARs demonstrated a trend of MC > FSPB_LS > FSPB. The FSPB_LS algorithm aligns better with the dose parameters of the target compared to the MC algorithm, which is particularly evident in UCLC. However, the FSPB algorithm significantly underestimated the does of the target. Regarding the OARs in CLC, differences in dose parameters were observed between FSPB and FSPB_LS for V10 of the contralateral lung, as well as between FSPB and MC for mean dose (Dmean) of the contralateral lung and maximum dose (Dmax) of the aorta, exhibiting statistical differences. There were no statistically significant differences observed between FSPB_LS and MC for the OARs. However, the average dose deviation between FSPB_LS and MC algorithms for OARs ranged from 2.79% to 11.93%. No significant dose differences were observed among the three algorithms in UCLC. For CLC, the FSPB_LS algorithm exhibited good consistency with the MC algorithm in PTV and demonstrated a significant improvement in accuracy when compared to the traditional FSPB algorithm. However, the FSPB_LS algorithm and the MC algorithm showed a significant dose deviation in OARs of CLC. In the case of UCLC, FSPB_LS showed better consistency with the MC algorithm than observed in CLC. Notwithstanding, UCLC's OARs were highly sensitive to radiation dose and could result in potentially serious adverse reactions. Consequently, it is advisable to use the MC algorithm for dose calculation in both CLC and UCLC, while the application of FSPB_LS algorithm should be carefully considered.

310.4: Developing an algorithm for islet transplant dosing that considers both pancreas donor and islet recipient factors

310.4: Developing an algorithm for islet transplant dosing that considers both pancreas donor and islet recipient factors

Sensitivity and specificity of Monte Carlo based independent secondary dose computation for detecting modulation-related dose errors in intensity modulated radiotherapy

BackgroundThe recent availability of Monte Carlo based independent secondary dose calculation (ISDC) for patient-specific quality assurance (QA) of modulated radiotherapy requires the definition of appropriate, more sensitive action levels, since contemporary recommendations were defined for less accurate ISDC dose algorithms. PurposeThe objective is to establish an optimum action level and measure the efficacy of a Monte Carlo ISDC software for pre-treatment QA of intensity modulated radiotherapy treatments. MethodsThe treatment planning system and the ISDC were commissioned by their vendors from independent base data sets, replicating a typical real-world scenario. In order to apply Receiver-Operator-Characteristics (ROC), a set of treatment plans for various case classes was created that consisted of 190 clinical treatment plans and 190 manipulated treatment plans with dose errors in the range of 1.5–2.5%. All 380 treatment plans were evaluated with ISDC in the patient geometry. ROC analysis was performed for a number of Gamma (dose-difference/distance-to-agreement) criteria. QA methods were ranked according to Area under the ROC curve (AUC) and optimum action levels were derived via Youden’s J statistics. ResultsOverall, for original treatment plans, the mean Gamma pass rate (GPR) for Gamma(1%, 1 mm) was close to 90%, although with some variation across case classes. The best QA criterion was Gamma(2%, 1 mm) with GPR > 90% and an AUC of 0.928. Gamma criteria with small distance-to-agreement had consistently higher AUC. GPR of original treatment plans depended on their modulation degree. An action level in terms of Gamma(1%, 1 mm) GPR that decreases with modulation degree was the most efficient criterion with sensitivity = 0.91 and specificity = 0.95, compared with Gamma(3%, 3 mm) GPR > 99%, sensitivity = 0.73 and specificity = 0.91 as a commonly used action level. ConclusionsISDC with Monte Carlo proves highly efficient to catch errors in the treatment planning process. For a Monte Carlo based TPS, dose-difference criteria of 2% or less, and distance-to-agreement criteria of 1 mm, achieve the largest AUC in ROC analysis

A first small step toward personalized immunosuppression

Lloberas etal. provide further evidence for the benefits of an individualized tacrolimus dosing algorithm based on population pharmacokinetics and pharmacogenetics. Better tacrolimus dosing could prevent underexposure and overexposure and potentially save costs. Most important, this could be the start of precision medicine in kidney transplantation, incorporating improved immunologic and donor quality assessments, advanced biopsy readouts, innovative pharmacogenomics for drug safety, and novel diagnostic and outcome algorithms to guide a truly personalized therapy.

Commissioning of Aktina SRS cones and dosimetric validation of the RayStation photon Monte Carlo dose calculation algorithm

Clinical implementation of SRS cones demands particular experimental care and dosimetric considerations in order to deliver precise and safe radiotherapy to patients. The purpose of this work was to present the commissioning data of recent Aktina cones combined with a 6MV flattened beam produced by an Elekta VersaHD linear accelerator. Additionally, the modelling process, and an assessment of dosimetric accuracy of the RayStation Monte Carlo dose calculation algorithm for cone based SRS was performed. There are currently no studies presenting beam data for this equipment and none that outlines the modelling parameters and validation of dose calculation using RayStation’s photon Monte Carlo dose engine with cones. Beam data was measured using an SFD and a microDiamond and benchmarked against EBT3 film for cones of diameter 5–39 mm. Modelling was completed and validated within homogeneous and heterogeneous phantoms. End-to-end image-guided validation was performed using a StereoPHAN™ housing, an SRS MapCHECK and EBT3 film, and calculation time was investigated as a function of statistical uncertainty and field diameter. The TPS calculations agreed with measured data within their estimated uncertainties and clinical treatment plans could be calculated in under a minute. The data presented serves as a reference for others commissioning Aktina stereotactic cones and the modelling parameters serve similarly, while providing a starting point for those commissioning the same TPS algorithm for use with cones. It has been shown in this work that RayStation’s Monte Carlo photon dose algorithm performs satisfactorily in the presence of SRS cones.

Halfway to Automated Feeding of Chinese Hamster Ovary Cells.

This paper presents a comprehensive study on the development of models and soft sensors required for the implementation of the automated bioreactor feeding of Chinese hamster ovary (CHO) cells using Raman spectroscopy and chemometric methods. This study integrates various methods, such as partial least squares regression and variable importance in projection and competitive adaptive reweighted sampling, and highlights their effectiveness in overcoming challenges such as high dimensionality, multicollinearity and outlier detection in Raman spectra. This paper emphasizes the importance of data preprocessing and the relationship between independent and dependent variables in model construction. It also describes the development of a simulation environment whose core is a model of CHO cell kinetics. The latter allows the development of advanced control algorithms for nutrient dosing and the observation of the effects of different parameters on the growth and productivity of CHO cells. All developed models were validated and demonstrated to have a high robustness and predictive accuracy, which were reflected in a 40% reduction in the root mean square error compared to established methods. The results of this study provide valuable insights into the practical application of these methods in the field of monitoring and automated cell feeding and make an important contribution to the further development of process analytical technology in the bioprocess industry.

A "Bundle of Care" to Improve Anticoagulation Control in Patients Receiving Warfarin in Uganda and South Africa: Protocol for an Implementation Study.

The quality of warfarin anticoagulation among Sub-Saharan African patients is suboptimal. This is due to several factors, including a lack of standardized dosing algorithms, difficulty in providing timely international normalized ratio (INR) results, a lack of patient feedback on their experiences with treatment, a lack of education on adherence, and inadequate knowledge and training of health care workers. Low quality of warfarin anticoagulation, expressed as time in therapeutic range (TTR), is associated with higher adverse event rates, including bleeding and thrombosis, and ultimately, increased morbidity and mortality. Processes and interventions that improve this situation are urgently needed. This study aims to evaluate the implementation of the "warfarin bundle," a package of interventions to improve the quality of anticoagulation and thereby clinical outcomes. The primary outcome for this study is TTR over the initial 3 months of warfarin therapy. Patients aged 18 years or older who are newly initiated on warfarin for venous thromboembolism, atrial fibrillation, or valvular heart disease will be enrolled and followed up for 3 months at clinics in Cape Town, South Africa, and Kampala, Uganda, where the warfarin bundle is implemented. A retrospective review of the clinical records of patients on warfarin treatment before implementation (controls) will be used for comparison. This study uses a mixed methods approach of the implementation of patient- and process-centered activities to improve the quality of anticoagulation. Patient-centered activities include the use of clinical dosing algorithms, adherence support, and root cause analysis, whereas process-centered activities include point-of-care INR testing, staff training, and patient education and training. We will assess the impact of these interventions by comparing the TTR and safety outcomes across the 2 groups, as well as the cost-effectiveness and acceptability of the package. We started recruitment in June 2021 and stopped in August 2022, having recruited 167 participants. We obtained ethics approval from the University of Cape Town Faculty of Health Sciences Human Research Ethics Committee, the Provincial Health Research Committees in South Africa, the Joint Clinical Research Centre Institutional Review Board, Kampala, and the University of Liverpool Research Ethics Committee. As of February 2023, data cleaning and formal analysis are underway. We expect to publish the full results by December 2023. We anticipate that the "bundle of care," which includes a clinical algorithm to guide individualized dosing of warfarin, will improve INR control and TTR of patients in Uganda and South Africa. We will use these findings to design a larger, multisite clinical trial across several Sub-Saharan African countries. DERR1-10.2196/46710.

The clinical application of beta-lactam antibiotic therapeutic drug monitoring in the critical care setting.

Critically ill patients have increased variability in beta-lactam antibiotic (beta-lactam) exposure due to alterations in their volume of distribution and elimination. Therapeutic drug monitoring (TDM) of beta-lactams, as a dose optimization and individualization tool, has been recommended to overcome this variability in exposure. Despite its potential benefit, only a few centres worldwide perform beta-lactam TDM. An important reason for the low uptake is that the evidence for clinical benefits of beta-lactam TDM is not well established. TDM also requires the availability of specific infrastructure, knowledge and expertise. Observational studies and systematic reviews have demonstrated that TDM leads to an improvement in achieving target concentrations, a reduction in potentially toxic concentrations and improvement of clinical and microbiological outcomes. However, a small number of randomized controlled trials have not shown a mortality benefit. Opportunities for improved study design are apparent, as existing studies are limited by their inclusion of heterogeneous patient populations, including patients that may not even have infection, small sample size, variability in the types of beta-lactams included, infections caused by highly susceptible bacteria, and varied sampling, analytical and dosing algorithm methods. Here we review the fundamentals of beta-lactam TDM in critically ill patients, the existing clinical evidence and the practical aspects involved in beta-lactam TDM implementation.

P-619 Is the algorithm derived, maximum daily dose of 12µg follitropin delta indeed sufficient, even for women considered at risk for underexposure?

Abstract Study question Does the follitropin delta (FD) dosing algorithm potentially underexpose high body weight (BW) women, especially those with a low ovarian reserve? Summary answer Our data indicate no decrease of ovarian response with increasing BW in women with serum AMH <15pmol/L treated with daily FD 12µg. What is known already The FD dosing algorithm stipulates a 1stcycle dose of daily 12µg FD (equivalent to approximately daily 188 IU rFSH) for patients with serum AMH <15pmol/L (2.1ng/ml), irrespective of BW. In trial and real-word study populations (ESTHER-1, 2017; Blockeel, 2022), respectively, nearly 40-50% of patients are thereby exposed to the same 12µg FD dose. In the phase III trial (ESTHER-1), however, only 5% of patients had a BW > 85kg. It has so-far been unclear, if women with high BW, e.g. large volume-of-distribution, are potentially underexposed by the algorithm dose of 12µg daily FD, especially when in the AMH 7-15pmol/L strata. Study design, size, duration A single, university center, retrospective analysis of all (n = 337) stimulation cycles with FD treatment performed from 12-June-2016 to 10-Dec-2021. FD was prescribed according to the summary of product characteristics in all patients. BW was assessed in the center on cycle day two or three. Primary outcome is the cumulus-oocyte-complex (COC) number retrieved. A target response is defined as 8-14 COCs, in line with the ESTHER-1 study. Participants/materials, setting, methods Women with an indication for IVF or ICSI, treatment naïve (e.g., 1st cycle), undergoing controlled ovarian stimulation in a GnRH-antagonist protocol with hCG or GnRH-agonist triggering for fresh or frozen-thawed transfer. There was no restriction on AMH levels, BW, cycle regularity or presence of PCOS. Data are shown as mean±standard deviation and/or median and range, as appropriate, or proportion. Main results and the role of chance Inclusion of 182 first cycles with mean age 33±4.2 years (22-44 years), AMH 25.4±16.7 pmol/L (<0.07-97.86) and BW 72.6±16.3 kg (47-135). 37 out of 182 (20%) women had a BW > 85kg. Median COC number was 7 (0-19) in women with BW < 85kg and 7 (2-20) COCs with BW ≥ 85kg receiving 12µg FD in the AMH <15pmol/L stratum (p > 0.05). A separate comparison by BW strata and AMH strata 0-7 and 7-15 pmol/L, respectively, also indicated no decrease of ovarian response with higher BW. Overall, in all maximally dosed 12 µg cycles (75/182, 41%), a target response was achieved in 32.7% of patient cycles in women with BW < 85kg and 35.7% in women with BW ≥ 85kg. In patients with AMH ≥15pmol/L, 21/70 patients (30%) had <8 COCs. While no difference in age, BMI, and mean FD dose was observed for those, they were on average taller (1.71±0.06m vs. 1.66±0.06m; p = 0.01), showed a tendency for an increased BW (73kg (48-110kg) vs. 63kg (47-119kg), p = 0.08) and body surface (1.87±0.21 vs. 1.76±0.2m2, p = 0.05) vs. patients with target response. In all subgroups, the prevalence of cycle irregularity was not different between target and hypo- or hyperresponding patients. Limitations, reasons for caution The cohort is uncontrolled and the analysis is of retrospective nature. Women with high BW are relatively rare limiting the sample size and power. High BW may be associated with alterations of follicular recruitment and ovarian physiology. Wider implications of the findings The maximum FSH dose for low ovarian reserve patients is debated (ESHRE COS guideline 2020). The FD algorithm predicts a maximum response at 12µg FD, e.g. equivalent to only approximately 188 IU FSH/day, in all women with AMH <15pmol/L. Our analysis supports this assumption. Trial registration number not applicable

P-674 Automatizing ovarian follicle counting and measurement with artificial intelligence

Abstract Study question Can ovarian follicle counting and volumetric measurement in ultrasound cine-loops be automated using deep learning methods? Summary answer A deep-learning model achieved an Area Under Precision-Recall Curve (AUC-PR) of 77.8% for identifying follicles of diverse sizes, in cine-loops from everyday clinical practice. What is known already The measurement of ovarian follicles via transvaginal ultrasound is an established procedure in infertility treatment. The high predictive value of antral follicle count (AFC) is utilized in hormonal dosing algorithms, scheduling of triggering and oocyte retrieval, as a criterion to assess hyperstimulation risk, etc. As it is repeatable and time-consuming, it is a good candidate for automatization. However, traditional methods fail to distinguish follicles from acoustically shadowed areas, ovarian cysts, or anechoic extraovarian structures. Existing approaches assume manual ovary outlining or 3D volume acquisition, which requires additional operator training. Moreover, existing evaluations rarely define criteria for correct follicle identification. Study design, size, duration A retrospective study was conducted on 331 ultrasound cine-loop videos from 100 patients (mean age 35 ± 5.6, AMH level 3.3 ± 2.9 ng/ml) undergoing an IVF or donor cycle between February and December 2021 at six IVF centers in Poland. For training the model, 1903 more cine-loops from 350 other patients were used. To reflect everyday clinical practice, there was no selection based on patient condition or video quality. Participants/materials, setting, methods Folliscan (MIM Solutions) is a model developed based on 3D neural networks architectures. It analyzes a cine-loop without any manual preprocessing and presents a list of follicles together with exact 3D outlines and confidence scores. A total of 24711 follicles were manually annotated and reviewed by sonography experts to ensure exhaustive enumeration. A detected follicle is considered correct if it sufficiently overlaps (Intersection over Union above 35%) with a single expert annotation. Main results and the role of chance The precision was 85.8% (95% Confidence Interval: 83.7–87.7) and recall was 74.2% (CI 71.7–76.6). The area under the precision-recall curve (AUC-PR) was 77.8%. We note that 19.7% of annotations were only added during review, indicating that even expert sonographers fail to annotate a certain number of follicles. For studies performed on days 7-12 of stimulation (N = 163 cine-loops), when only follicles ≥10mm were taken into account, accuracy was significantly higher: precision 95.4% (CI 93.0–96.8), recall 87.7% (CI 73.1–94.2). We observe that the smallest anechoic areas, about 1mm in diameter, often are not imaged in enough resolution to be unanimously classified by experts. Common reasons for errors in medium-sized follicles are poor image quality (acoustic shadowing, unclear boundaries between nearby follicles), ambiguous volumes (where two outlines can seem equally reasonable, but do not overlap sufficiently), non-convex follicle shapes (due to adjacent follicles). When measuring follicle diameters, the Mean Average Error (MAE) was 0.76mm, only slightly larger than the inter-observer MAE of 0.62mm (calculated on 32 cine-loops for which two experts independently measured all follicles). Moreover, 3D outlines enable a more physiological measurement than 2D diameters. Limitations, reasons for caution Follicle recognition is highly dependent on image quality and patient group. To contrast our model with manual methods, more cine-loops would need to be independently annotated by multiple experts. A future study on 3D acquisitions would be useful to compare with software that requires it, such as SonoAVC (GE Healthcare). Wider implications of the findings Automatizing follicle measurement can speed-up the process and diminish acquisition requirements (examination time, operator experience). Deep learning methods consider each follicle’s context, enabling more reliable recognition of ovarian structures. They have potential to increase the predictive value of follicle counting beyond what is already possible with human observers. Trial registration number N/A

P-679 Controlled ovarian stimulation using follitropin delta results in higher cumulative live birth rate compared to follitropin alfa/beta in a large prospective real world dataset

Abstract Study question Does IVF/ICSI-treatment using follitropin delta (hrFSH) result in higher pregnancy rates (PR) or live birth rates (LBR) compared to follitropin alfa/beta (recFSH)? Summary answer Real world data (RWD) from a large prospective German registry study shows that using hrFSH results in higher pregnancy rates compared to recFSH. What is known already Follitropin delta represents a new recombinant FSH derived from a human cell line. Contrary to follitropin alfa or beta, follitropin delta has a glycolysation pattern consisting of α2,3- and α2,6-linked sialic acids which is more similar to native human FSH. Follitropin delta is approved for use with an individualized dosing algorithm based on serum AMH and body weight, targeting an optimal ovarian response (8-14 oocytes). Efficacy and safety have been demonstrated in numerous clinical trials. Moreover, Follitropin delta seems to reduce the risk of OHSS although its safety with respect to ovarian hyperstimulation using RWD remains to be investigated. Study design, size, duration The German IVF Registry collects prospectively data about IVF/ICSI-treatments from 140 German IVF centers. The underlying analysis of RWD includes controlled ovarian hyperstimulation cycles that have been performed in Germany between 2017-2021. Study groups were built subject to the gonadotropin used for stimulation, irrespective of previous treatments: (1) hrFSH, n = 3002, (2) recFSH, n = 135293. “Pregnancy” was defined as clinically assessed, intrauterine pregnancy, including miscarriages. Biochemical pregnancies were not defined as “pregnant”. Ectopic pregnancies (n = 558) were excluded. Participants/materials, setting, methods PR and LBR were calculated subject to the number of embryo transfers (ET) (n = 2062 (rFSH) vs. n = 103357 (recFSH)) after excluding freeze-all cycles, (n = 529 (hrFSH) vs. n = 13682 (recFSH)) and cycles that ended without ET (n = 411 (hrFSH) vs. n = 18254 (recFSH)). The collected data were saved in compliance with the applicable data processing regulations. Statistical analysis was performed using Fishers-exact- and students-t-test and Microsoft Excel, whereas p < 0.05 was defined as significant. Main results and the role of chance There was no difference between the study groups regarding age (33.9 ± 4.00y (hrFSH) vs. 33.8y±4.34y (recFSH), p = 0.12) or infertility diagnosis. Stimulation with hrFSH resulted in a higher number of oocytes (11.15±7.2 vs. 10.39±7.01, p < 0.01). Antagonist protocol was used less often in the hrFSH-group (78% (2344/3002 vs. 80% (109217/135293), p < 0.01). PR in 2021 were higher (39.2% (222 pregnancies / 567 embryo transfers) vs. 35.2% (6837/19420), OR = 1.18 [0.99-1.41] p = 0.05) while using hrFSH compared to recFSH. The effect was even stronger when comparing patients aged 30-34y (46.2% (98/212) vs. 38.6% (2967/7865), OR = 1.41 [1.07-1.88], p = 0.01). Another quantitative difference could be observed for patients aged 30-34y in their first IVF/ICSI-cycle (PR 48.5% (63/130) vs. 39.5% (1926/4873), OR = 1.44 [1.00-2.07], p = 0.05). Including all cycles between 2017-2021 resulted in higher PR for couples treated with hrFSH (38.3% (790/2062) vs. 36.2% (37439/103357), OR = 1.09 [1.00-1.20], p = 0.049). Cumulative PR including consecutive frozen embryo transfers after the first stimulation showed significantly higher PR while using hrFSH (81.6% (422/517) vs. 71.3% (17373/24367), OR = 1.79 [1.42-2.23], p < 0.01). Finally, cumulative LBR per ET was significantly increased if hrFSH was used for ovarian stimulation (60.0% (310/517) vs. 51.9% (12648/24367), OR = 1.39 [1.16-1.66], p < 0.01) compared to recFSH. Limitations, reasons for caution Since we analyzed RWD, comparison with large clinical trials should be considered carefully. PR and LBR was calculated only using stimulations which successfully generated ET. Moreover, individual AMH values are not transmitted to the German IVF Registry. Therefore, difference in ovarian reserve between study groups can’t be excluded. Wider implications of the findings In this large, prospective RWD, higher cumulative LBR and PR using hrFSH compared to recFSH, irrespective of age or infertility diagnosis, supports the use of individualized fertility treatment approach based on hrFSH. These results are consistent with previous retrospective findings. Trial registration number Not applicable

1057-P: Improvement in Hypoglycemia Rates with EndoTool IV

Health systems are utilizing insulin dosing algorithms (IDA) within the EMR to adjust IV insulin (IVI). EndoTool IV (ETIV) is IDA software for blood glucose (BG) management which evaluates BG values to recommend IV insulin/dextrose doses to adjust the BG level within a target range. ETIV was integrated with our EMR as part of a quality project to reduce hypoglycemia in IVI patients. The tool was started in two of our ICUs that treat our DKA admissions. Dashboards capture BG values defined as hypoglycemia BG less than 70 mg/dL, severe hypoglycemia BG less than 40 mg/dL, and length of hospital stay for diabetic ketoacidosis (DKA LOS). Dosing modalities were available for hyperglycemia, HHS or DKA; 90% of ordering providers selected the hyperglycemia mode. We report results on the first 437 patient runs. The population was 60 % type 2 DM, 16 % type 1 and the remainder unknown. Average BG at IVI start was 221.5 mg/dL with an average time to goal of 4.7 hours. This translated to an average time on IVI of 15.5 hours for the group studied. Adoption of ETIV was associated with an approximate 5-fold reduction in hypoglycemic BG values, elimination of observed severe hypoglycemic BG values and a reduction in length of stay for DKA LOS. We conclude that ETIV is effective in reducing hypoglycemic events in persons treated with IVI and may reduce DKA LOS. Hospital systems considering a quality improvement project to reduce inpatient hypoglycemia for IVI should consider adopting an IDA. Disclosure J.Aloi: Research Support; Abbott Diabetes, Medtronic. C.E.Price: None. C.O.Usoh: None. K.Dunn: None.

948-P: Glycemic Monitoring Using the Dexcom G6 in Cardiac Surgery

Hyperglycemia is associated with adverse post-surgical outcomes. We evaluated Dexcom G6 accuracy following cardiac surgery. A clinical glucose dataset included 1428 readings of 29 patients, Dexcom data included 45,645 measures. Clinical and Dexcom measures were restricted at least 1 hour after prior measure. Matching clinical and Dexcom measures were required within 5 min. Data were included at least 2 hr after Dexcom insertion and ICU admit. A dataset only measuring ≥24 hrs post ICU admit was created to explore device stability. Patients remained on Dexcom until discharge or 10 days postop. The population was 71% male 14% with known diabetes, 66% who went on intravenous insulin infusion. Figure 1a is Clarke error grid plot of all measures post-ICU admission: 53.5% were Zone A, 45.9% Zone B, and 0.6% (n=5) Zones D or E. Figure 1b uses restricted dataset beginning 24 hours post-ICU admission: 55.9% in Zone A, 43.9% in Zone B, and 0.2% in Zone D. MARD between clinical and Dexcom measures was 20.6% and 21.6% in post-ICU admission dataset and dataset excluding first 24 hours after ICU admission. Almost 100% of Dexcom G6 and clinical data matching points fell within areas considered as giving clinically correct decisions (Zone A) and clinically uncritical decisions (Zone B). If CGM readings are used for clinical decision-making, glucose targets or insulin dosing algorithms may need adjustment. As technology evolves, this device may limit glycemic fluctuations. Disclosure S.R.Insler: None. B.J.Wakefield: None. J.F.Bena: None. A.Debs: None. K.Brake: None. I.Nwosu: None. M.Lansang: Research Support; Dexcom, Inc., Abbott, Alertgy Inc.

Automatic Levothyroxine Dosing Algorithm for Patients Suffering from Hashimoto's Thyroiditis.

Hypothyroidism is a condition where the patient's thyroid gland cannot produce sufficient thyroid hormones (mainly triiodothyronine and thyroxine). The primary cause of hypothyroidism is autoimmune-mediated destruction of the thyroid gland, referred to as Hashimoto's thyroiditis. A patient's desired thyroid hormone concentration is achieved by oral administration of thyroid hormone, usually levothyroxine. Establishing individual levothyroxine doses to achieve desired thyroid hormone concentrations requires several patient visits. Additionally, clear guidance for the dosing regimen is lacking, and significant inter-individual differences exist. This study aims to design a digital automatic dosing algorithm for patients suffering from Hashimoto's thyroiditis. The dynamic behaviour of the relevant thyroid function is mathematically modelled. Methods of automatic control are exploited for the design of the proposed robust model-based levothyroxine dosing algorithm. Numerical simulations are performed to evaluate the mathematical model and the dosing algorithm. With the help of the developed controller thyroid hormone concentrations of patients, emulated using Thyrosim, have been regulated under the euthyroid state. The proposed concept demonstrates reliable responses amidst varying patient parameters. Our developed model provides a useful basis for the design of automatic levothyroxine dosing algorithms. The proposed robust feedback loop contributes to the first results for computer-assisted thyroid dosing algorithms.