This is the second blog providing a “friendly overview” of a systematic review I co-authored on biological predictors of clozapine response, which has recently been published in Frontiers Psychiatry. In the first, I explained why predicting response to clozapine is important and gave a brief explainer of what systematic reviews are. In this one, I’ll explain how we conducted our review, and the key messages from our findings.

Obviously you can find all of the detail in the paper itself, but this is primarily aimed at non-scientists or people outside of the field. If anyone is interested in more detail of a particular aspect, I’ll happily write a follow up on certain aspects or answer questions in the comments/on Twitter!

Recap of Part 1:

Why is this topic important? Because a significant chunk of people with schizophrenia could benefit from clozapine (an anti-psychotic medication) but wait years to receive it. Having physical tests that we could run to determine that clozapine is likely to reduce someone’s symptoms might encourage clinicians and patients to start clozapine earlier. Equally, if we can determine that clozapine is unlikely to benefit someone, we can avoid unnecessarily exposing them to the potential side effects. 

What is a systematic review? An approach to doing reviews of research areas which tries to ensure that every relevant study is identified and reported on fairly. You have to follow a structured set of rules when searching for studies and when writing up what you find.

Summary of Part 2

How did we conduct our systematic review? We decided on the key aspects of what we were looking for – peer-reviewed studies which measured the relationship between biological features of their participants before starting clozapine and later response to clozapine. In January 2018, we searched a large biomedical research database for studies mentioning clozapine, response/outcome and schizophrenia. We then sifted through the results to find only those studies which met our criteria, and we also looked through relevant articles to see if they mentioned any other studies which we had missed. We ended up with 98 studies to include in our review.

What are the key messages?

• Greater blood flow and metabolism in the front of the brain, as well as the front of the brain having a larger volume, may predict good response to clozapine in patients.
• Having a greater turnover of serotonin than of dopamine in the brain may also predict good response to clozapine. 
• The vast majority of studies have looked at whether genes can predict clozapine response, but without much success. 
  • We need to stop doing the same candidate gene studies of clozapine response! 
• Very few studies reported important information on clozapine plasma levels. 
• Studies also looked at electrical activity in the brain, levels of various chemical transmitters in the blood, heart function, and the immune system but none of these provided consistent results.
• Final thought – hardly any of these studies tested whether this information could be used to predict an outcome for an individual.

In Detail…

How did we conduct our systematic review?

Our aim was to find studies on biological predictors of clozapine response, so we came up with the following key characteristics a study needed to have to be a relevant study (for a full list, read the Study Selection section of the paper):

• The studies obviously needed to measure the relationship between a biological feature of their participants and their participants’ clinical response to clozapine
• The biological features needed to either have been measured before the participant started taking clozapine, or be genetic.
  • To explain why, let’s imagine we wanted to use a brain scan to predict clozapine response. Our end goal is to be able to give someone a brain scan before they start clozapine, identify that the front of the brain is small (as a simple example) and say that this makes them a great candidate for clozapine. Now the front of their brain might change in size while someone takes clozapine and become a lot bigger. If we did research on people who have taken clozapine for a year, the results might show that people who respond well to clozapine have normal sized brains and we might completely miss the fact that the front of their brain was much smaller before starting clozapine. Basically, if we want to be able to develop a test to use before someone starts clozapine, we need to base this on research which gives us information about people before they start clozapine. So, we only included studies which measured biological features before the participants took their first doses of clozapine.
  • Genetic studies are a little different because the sequence of our genes stays the same throughout our life. This means the sequence will be exactly the same whether we measure it before or after taking clozapine, so any information on relevant genes after starting clozapine would also have been relevant before starting clozapine. This meant we could make an exception and also include studies which measured their gene sequences after taking their first dose of clozapine.
• The studies needed to have been peer-reviewed, to check they met a basic level of scientific quality

We searched PubMed – one of the largest online databases for biomedical research – on 20th January 2018, to find papers written in English, conducted in humans, which mentioned “clozapine” AND (“response” OR “outcome”) AND “schizophrenia” in their titles or abstracts. This search found 753 studies meeting these criteria.

We then went through all of the abstracts of these 753 paper, to determine whether we thought they could potentially meet our criteria for being included or whether they definitely didn’t. We determined that 627 definitely weren’t relevant, leaving 126 studies to make a final decision on. After checking the full articles, which provide more details, we excluded 57 articles for being irrelevant. Ruta Samanaite and I did both of these stages separately, to check we came to the same conclusions. Of the 69 articles left, we then went through all the studies they mentioned to identify any other relevant articles which we might have missed – this found 29 extra studies. Ultimately, we ended up with 98 studies which had relevant information on biological predictors of clozapine response.

What are the key messages?

Greater blood flow and metabolism in the front of the brain, as well as the front of the brain having a larger volume, may predict good response to clozapine in patients with treatment-resistant schizophrenia. Three studies of blood flow and metabolism in the brain found that participants with greater blood flow and metabolism in the dorsolateral prefrontal cortex had better improvements after taking clozapine than participants with reduced blood flow and metabolism. Similarly, three studies of brain structure in patients with treatment-resistant schizophrenia all found that participants with a larger prefrontal area of the brain had better improvements after taking clozapine than participants with a smaller prefrontal area of the brain. One study found different results, but wasn’t specifically in patients with treatment-resistant schizophrenia.

Having a greater turnover of serotonin than of dopamine in the brain may predict good response to clozapine. Cerebrospinal fluid supplies nutrients to the brain and removes waste, and can provide a good indicator of how the brain is using different chemical messengers such as serotonin and dopamine. Three studies of cerebrospinal fluid all found that dopamine and serotonin metabolites (evidence of the brain processing dopamine and serotonin) weren’t linked to clozapine response when looked at separately, but that lower ratio of dopamine metabolites to serotonin metabolites (i.e. greater turnover of serotonin) predicted good response to clozapine.

The vast majority of studies have looked at whether genes can predict clozapine response, without much success. 70 out of the 98 studies in the review looked at whether genes could predict clozapine response, collectively looking at over 350 individual differences in just one “letter” of the genetic code (for some basics of genetics, read my previous post). However, if you look at table 9 of the paper you’ll see that the vast majority did not find a link to clozapine response. In fact, only four of these differences were found by two independent research groups to be good predictors of clozapine response, and three of these have also been reported to not be good predictors.

We need to stop doing the same candidate gene studies of clozapine response! The vast majority of these genetic studies have taken an approach called the “candidate-gene” approach, where you pick one or two genes (your candidate genes) to see if they are linked to the thing you’re interested. There are quite a few technical and statissues with candidate-gene studies (that’s a story for another blog post…) but the main point is that this review demonstrates that despite years of research and many studies using this approach, it has failed to produce reliable predictors of clozapine response. It’s likely that newer approaches – including those that look at thousands of genes and those which use a combination of multiple genes- will be more fruitful, but time will tell.

Very few studies reported important information on clozapine plasma levels. People respond to the same dose of medication differently depending on body size, metabolism, whether they smoke etc. Taking a blood test to check the level of a drug in someone’s plasma is a way of checking that the dose they’ve been given is enough for them to benefit from the drug. The vast majority of studies (91%) didn’t report on clozapine plasma levels, meaning we don’t know whether someone’s symptoms remaining unaffected by clozapine is simply because they didn’t have a high enough dose. This is something future research should aim to measure to help us be more certain about how to understand their results.

Studies also looked at electrical activity in the brain, levels of various chemical transmitters in the blood, heart function, and the immune system but none of these provided consistent results.

Final thought – hardly any of these studies tested whether this information could be used to predict an outcome for an individual. One of the problems when we talk about predictors is that often we’re actually talking about research which looks at groups – research which finds that, across a group of people, having a certain feature is generally linked to a certain outcome later. But this doesn’t necessarily mean that it’s helpful information when applied to an individual person to try and predict their outcome. For example, across the whole of England, we’d probably find a general link between whether you live in the north or south and whether you call your evening meal “tea”, “dinner”, “supper” etc – but if we picked out one person currently living somewhere in the south of England, I wouldn’t want to bet much money on what they say because there’s so many regional differences within the south, I don’t know where they grew up or where their parents grew up etc. Even if there is a strong link when we split the population into two groups, that doesn’t mean the link is there for each individual. So, if we want to use the results of biological research in medical practice, we need to start testing whether these predictors are helpful for individual patients.

(Note: This blog post, while heavily based on the co-authored paper, has been written solely by myself and reflects my views and interpretations).