Evaluating the performance of machine-learning regression models for pharmacokinetic drug-drug interactions.

Combination therapy or concomitant drug administration can be associated with pharmacokinetic drug-drug interactions, increasing the risk of adverse drug events and reduced drug efficacy. Thus far, machine-learning models have been developed that can classify drug-drug interactions. However, to enable quantification of the pharmacokinetic effects of a drug-drug interaction, regression-based machine learning should be explored. Therefore, this study investigated the use of regression-based machine learning to predict changes in drug exposure caused by pharmacokinetic drug-drug interactions. Fold changes in exposure relative to substrate drug monotherapy were collected from 120 clinical drug-drug interaction studies extracted from the Washington Drug Interaction Database and SimCYP compound library files. Drug characteristics (features) were collected such as structure, physicochemical properties, in vitro pharmacokinetic properties, cytochrome P450 metabolic activity, and population characteristics. Three different regression-based supervised machine-learning models were then applied to the prediction task: random forest, elastic net, and support vector regressor. Model performance was evaluated using fivefold cross-validation. Strongest performance was observed with support vector regression, with 78% of predictions within twofold of the observed exposure changes. The results show that changes in drug exposure can be predicted with reasonable accuracy using regression-based machine-learning models trained on data available early in drug discovery. This has potential applications in enabling earlier drug-drug interaction risk assessment for new drug candidates.

Pre-diagnostic prescription patterns in bladder and renal cancer: a longitudinal linked data study.

BACKGROUND: Understanding pre-diagnostic prescribing activity could reveal windows during which more timely cancer investigation and detection may occur. AIM: To examine prescription patterns for common urological clinical features prior to renal and bladder cancer diagnoses. DESIGN AND SETTING: A retrospective cohort study was performed using electronic primary care and cancer registry data on patients with bladder and renal cancer, who received their diagnosis between April 2012 and December 2015 in England. METHOD: Primary care prescriptions up to 2 years pre- diagnosis were analysed for five groups of clinical features (irritative urological symptoms, obstructive symptoms, urinary tract infections [UTIs], genital infections, and atrophic vaginitis). Poisson regressions estimating the inflection point from which the rate of prescriptions increased from baseline were used to identify the start of diagnostic windows during which cancer could be detected. RESULTS: A total of 48 094 prescriptions for 5322 patients were analysed. Inflection points for an increase in UTI prescriptions were identified 9 months pre- diagnosis for renal (95% confidence interval [CI] = 5.3 to 12.7) and bladder (95% CI = 7.4 to 10.6) cancers. For bladder cancer, the change in UTI antibiotic prescription rates occurred 4 months earlier in females (11 months pre- diagnosis, 95% CI = 9.7 to 12.3) than in males (7 months pre-diagnosis, 95% CI = 5.4 to 8.6). For other clinical features, no inflection points were identified and, as such, no diagnostic windows could be defined. CONCLUSION: Prescription rates for UTIs increased 9 months before bladder and renal cancer diagnoses, indicating that there is potential to expedite diagnosis of these cancers in patients presenting with features of UTI. The greatest opportunity for more timely diagnosis may be in females with bladder cancer, who experienced the earliest increase in UTI prescription rate.

Comparing the applications of machine learning, PBPK, and population pharmacokinetic models in pharmacokinetic drug-drug interaction prediction.

The gold-standard approach for modeling pharmacokinetic mediated drug-drug interactions is the use of physiologically-based pharmacokinetic modeling and population pharmacokinetics. However, these models require extensive amounts of drug-specific data generated from a wide variety of in vitro and in vivo models, which are later refined with clinical data and system-specific parameters. Machine learning has the potential to be utilized for the prediction of drug-drug interactions much earlier in the drug discovery cycle, using inputs derived from, among others, chemical structure. This could lead to refined chemical designs in early drug discovery. Machine-learning models have many advantages, such as the capacity to automate learning (increasing the speed and scalability of predictions), improved generalizability by learning from multicase historical data, and highlighting statistical and potentially clinically significant relationships between input variables. In contrast, the routinely used mechanistic models (physiologically-based pharmacokinetic models and population pharmacokinetics) are currently considered more interpretable, reliable, and require a smaller sample size of data, although insights differ on a case-by-case basis. Therefore, they may be appropriate for later stages of drug-drug interaction assessment when more in vivo and clinical data are available. A combined approach of using mechanistic models to highlight features that can be used for training machine-learning models may also be exploitable in the future to improve the performance of machine learning. In this review, we provide concepts, strategic considerations, and compare machine learning to mechanistic modeling for drug-drug interaction risk assessment across the stages of drug discovery and development.

Pre-diagnostic clinical features and blood tests in patients with colorectal cancer: a retrospective linked-data study.

BACKGROUND: The majority of colorectal cancer is diagnosed in patients following symptomatic presentation in the UK. AIM: To identify windows of opportunity for timely investigations or referrals in patients presenting with colon and rectal cancer-relevant symptoms or abnormal blood tests. DESIGN AND SETTING: A retrospective cohort study was undertaken using linked primary care and cancer registry data for patients with colorectal cancer diagnosed in England between 2012 and 2015. METHOD: Monthly consultation rates for relevant clinical features (change in bowel habit, rectal bleeding, abdominal pain, abdominal mass, constitutional symptoms, and other bowel symptoms) and abnormal blood test results (low haemoglobin, high platelets, and high inflammatory markers) up to 24 months pre-diagnosis were calculated. Poisson regression adjusted for age, sex, and relevant comorbidities was used to estimate the most likely month when consultation rates increased above baseline. RESULTS: In total, 5033 patients with colon cancer and 2516 with rectal cancer were included. Consultations for all examined clinical features and abnormal blood tests increased in the year pre-diagnosis. Rectal bleeding was the earliest clinical feature to increase from the baseline rate: at 10 months (95% confidence interval [CI] = 8.3 to 11.7) pre-diagnosis for colon cancer and at 8 months (95% CI = 6.1 to 9.9) pre-diagnosis for rectal cancer. Low haemoglobin, high platelets, and high inflammatory markers increased from as early as 9 months pre-diagnosis. CONCLUSION: This study found evidence for an early increase in rates of consultation for relevant clinical features and abnormal blood tests in patients with colorectal cancer, suggesting that earlier instigation of cancer-specific investigations or referrals may be warranted in some patients who were symptomatic.

Harnessing Clinical Trial and Real-World Data Towards an Understanding of Sex Effects on Drug Pharmacokinetics, Pharmacodynamics and Efficacy.

Increasing clinical data on sex-related differences in drug efficacy and toxicity has highlighted the importance of understanding the impact of sex on drug pharmacokinetics and pharmacodynamics. Intrinsic differences between males and females, such as different CYP enzyme activity, drug transporter expression or levels of sex hormones can all contribute to different responses to medications. However, most studies do not include sex-specific investigations, leading to lack of sex-disaggregated pharmacokinetic and pharmacodynamic data. Based available literature, the potential influence of sex on exposure-response relationship has not been fully explored for many drugs used in clinical practice, though population-based pharmacokinetic/pharmacodynamic modelling is well-placed to explore this effect. The aim of this review is to highlight existing knowledge gaps regarding the effect of sex on clinical outcomes, thereby proposing future research direction for the drugs with significant sex differences. Based on evaluated drugs encompassing all therapeutic areas, 25 drugs demonstrated a clinically meaningful sex differences in drug exposure (characterised by ≥ 50% change in drug exposure) and this altered PK was correlated with differential response.