‘Precision medicine’ refers to the tailoring of medical treatment depending on a patient’s individual characteristics, e.g., genes, environmental factors, and lifestyle. If physicians could accurately predict an individual patient’s responses to different treatment options, the best option could be selected. For most diseases, the efficacy and safety of standard treatments are highly variable. Thus, individual-specific treatment protocols have been hailed as an emerging revolution in medicine, with the potential to improve patient care and deliver cost savings to health services [1].
Copyright by Felix Beacher via www.mdpi.com
Currently, precision medicine in real-world clinical practice is mainly associated with treatment based on cancer subtype and genotype. For example, olaparib is a monotherapy for ovarian cancer in women with BRCA1/2 mutations [2]. However, there are still few examples of real-world precision medicine. Current clinical practice still relies heavily on subjective judgment and limited individual patient data [3]. A ‘one-drug-fits-all’ approach is often used, in which a particular diagnosis leads to a specific type of treatment. Alternatively, trial-and-error practices are common, in which various treatment options are tried in the hope that one will work.
Machine learning (ML) has been described as ‘the key technology’ for the development of precision medicine [4]. ML uses computer algorithms to build predictive models based on complex patterns in data. ML can integrate the large amounts of data required to “learn” the complex patterns required for accurate medical predictions. ML has excelled in diagnostics, e.g., in neurodegenerative diseases [5], cardiovascular disease [6], and cancer [7]. ML approaches have also been used to predict treatment outcomes for a range of conditions, including schizophrenia [8], depression [9], and cancer [10].
In 2014, IBM launched ‘Watson for Oncology,’ which aimed to use ML to recommend treatment plans for cancer, based on combined inputs from research, a patient’s clinical notes, and the clinician [11]. However, this project has so far failed to deliver the kinds of commercial products which had been envisioned [12]. Other reports have used ML to predict treatment outcomes for cancer. One study used ML to predict patient survival based on microscopy images of cancer biopsy tissue and genomic markers [13]. Another study used ML to predict response to treatment in patients with cervical and head cancers based on PET images [14]. Despite such advances, there are currently no ML-based tools approved by regulators.
A significant limitation of these kinds of exploratory approaches to ML-based tools for clinical practice is that they tend to rely on types of data that are expensive to collect and may require specialist skills to analyze (e.g., genomics or MRI/PET imaging). This limitation can be a barrier to translating systems into tools for routine clinical practice. One possible way to address this is to base such systems on data from Phase III clinical trials: studies that provide the pivotal data for regulators to assess whether a new drug should be commercially approved. Phase III clinical trials are typically large enough for ML (usually 1000+ subjects) and include a wide range of data (e.g., demographic, clinical, and biochemical), which can be easily stored in tabular format. Moreover, the current trend is towards making clinical trial datasets publicly available. Thus, Phase III clinical trial data may be a good source for developing practical ML-based tools for precision medicine. This approach is relatively untried: a literature search revealed only one prior study that used clinical trial data to predict treatment responses [15]. This study used ML to predict responses to an anti-depressant after 12 weeks. This was an important first step, however the accuracy level was modest (65%). It is possible that, with different data, and with a different approach to modeling, a level of accuracy could be achieved which could lead to the development of tools for real-world clinical practice. […]
Read more: www.mdpi.com
‘Precision medicine’ refers to the tailoring of medical treatment depending on a patient’s individual characteristics, e.g., genes, environmental factors, and lifestyle. If physicians could accurately predict an individual patient’s responses to different treatment options, the best option could be selected. For most diseases, the efficacy and safety of standard treatments are highly variable. Thus, individual-specific treatment protocols have been hailed as an emerging revolution in medicine, with the potential to improve patient care and deliver cost savings to health services [1].
Copyright by Felix Beacher via www.mdpi.com
Machine learning (ML) has been described as ‘the key technology’ for the development of precision medicine [4]. ML uses computer algorithms to build predictive models based on complex patterns in data. ML can integrate the large amounts of data required to “learn” the complex patterns required for accurate medical predictions. ML has excelled in diagnostics, e.g., in neurodegenerative diseases [5], cardiovascular disease [6], and cancer [7]. ML approaches have also been used to predict treatment outcomes for a range of conditions, including schizophrenia [8], depression [9], and cancer [10].
In 2014, IBM launched ‘Watson for Oncology,’ which aimed to use ML to recommend treatment plans for cancer, based on combined inputs from research, a patient’s clinical notes, and the clinician [11]. However, this project has so far failed to deliver the kinds of commercial products which had been envisioned [12]. Other reports have used ML to predict treatment outcomes for cancer. One study used ML to predict patient survival based on microscopy images of cancer biopsy tissue and genomic markers [13]. Another study used ML to predict response to treatment in patients with cervical and head cancers based on PET images [14]. Despite such advances, there are currently no ML-based tools approved by regulators.
A significant limitation of these kinds of exploratory approaches to ML-based tools for clinical practice is that they tend to rely on types of data that are expensive to collect and may require specialist skills to analyze (e.g., genomics or MRI/PET imaging). This limitation can be a barrier to translating systems into tools for routine clinical practice. One possible way to address this is to base such systems on data from Phase III clinical trials: studies that provide the pivotal data for regulators to assess whether a new drug should be commercially approved. Phase III clinical trials are typically large enough for ML (usually 1000+ subjects) and include a wide range of data (e.g., demographic, clinical, and biochemical), which can be easily stored in tabular format. Moreover, the current trend is towards making clinical trial datasets publicly available. Thus, Phase III clinical trial data may be a good source for developing practical ML-based tools for precision medicine. This approach is relatively untried: a literature search revealed only one prior study that used clinical trial data to predict treatment responses [15]. This study used ML to predict responses to an anti-depressant after 12 weeks. This was an important first step, however the accuracy level was modest (65%). It is possible that, with different data, and with a different approach to modeling, a level of accuracy could be achieved which could lead to the development of tools for real-world clinical practice. […]
Read more: www.mdpi.com
Share this: