Machine learning is a type of AI. It is a developing and exciting tool which can be useful for making predictions.
But can it be used to help guide the correct dose of medicines? For example, the immunosuppressant tacrolimus, which is used after liver transplants.
A study by S.B. Yoon and colleagues, working at a hospital in South Korea, looked to see whether this was a possibility.
What is Tacrolimus and why is it given?
After an organ transplant, rejection can happen. This is because the immune system thinks that the new organ is foreign. The immune system then begins to attack it which can lead to the organ transplant failing.
To stop this from happening an immunosuppressant medicine such as Tacrolimus is given. This reduces the immune systems activity, helping the body to accept a new organ.
For tacrolimus to be an effective immunosuppressant, the correct dose must be carefully calculated. This dose will vary from person to person.
An initial dose is based upon body weight and age. It is then adjusted up or down to find the best dose for an individual.
Why do we need to find the perfect dose?
Tacrolimus has a very limited therapeutic range. This means that there is a small range for the medicine to be effective.
Too low a dose and there is the risk that the body’s immune system will attack the new liver causing it to be rejected. But too high and there is the possibility of serious side effects.
The dose has to be just right. Doctors monitor people taking tacrolimus very closely. The dose is adjusted up or down to ensure no issues arise
This study by Soo Bin Yoon and colleagues looked at a new way of predicting the correct dose. They hoped that this would reduce the chance of the dose needing to be changed. To do this they developed a new machine learning model to look at tacrolimus doses.
What is machine learning?
In machine learning, information is given to a computer model to ‘teach’ it. As the model “learns” from this information its performance improves, and it can begin to find patterns.
It can then use this pattern recognition to begin to make predictions.
How has this study used machine learning?
In this research, a machine learning model was produced and ‘trained’ to predict the correct post-transplant dose of tacrolimus.
To ‘train’ the machine learning model, it was given data from 443 people. They had all had a liver transplant and received tacrolimus post-transplant. The model then ‘learnt’ from this information and identified patterns.
The model was tested by entering information about other people who were given tacrolimus. It was not told the dose that these people had ended up taking.
Using patterns and predictions, the model suggested appropriate tacrolimus doses for these other individuals. The researchers then compared these suggested doses to the actual doses each person had ended up taking.
Findings
In this study, they found that the machine learning model showed a ‘clinically acceptable performance’. This means it was able to make accurate and reliable enough predictions. And that these predictions can be used to support medical decisions.
The researchers also looked at patients whose actual dose of tacrolimus was different to the dose suggested by the model. They found that on average these patients spent longer in intensive care after their transplant.
The researchers here suggest that the model has the potential to be used as a guide. Which could find the correct dose of tacrolimus after transplant.
But there are limitations, and it is not able to be used on its own. At the moment it should only be used alongside laboratory testing. And individuals taking tacrolimus would still need close monitoring by medical experts.
Studies such as this one, suggest that in the future machine learning could improve medical treatment. It could increase the likelihood that the correct dose of medicines such as tacrolimus will be given without the need for lots of trial and error. And reduce the risks associated with an incorrect dose.
You can read the full research paper here.
This blog was written for us by Amanda Gilbert.
Amanda is a PhD student at the University of Southampton. She is working to create a better model of the liver to be used in research labs, hoping that this can improve future studies and knowledge about the liver. She has recently completed a placement with the British Liver Trust where she has spent time learning about liver disease and the amazing support provided by the trust.
