Diagnosis of Childhood Growth Hormone Deficiency and Prediction of Therapeutic Response Using Gene Expression Data

Growth Hormone Deficiency is a condition in which children grow very poorly without treatment with injections of growth hormone. It is very important to make this diagnosis as children with growth hormone deficiency will grow to a normal height if treated. Currently to make a diagnosis of growth hormone deficiency doctors need to undertake a combination of blood tests and brain scans. It is very difficult to measure growth hormone during the day as children produce this hormone only overnight. Since blood tests overnight are not easy doctors use special type of test called a stimulation test to diagnose growth hormone deficiency using blood taken during the day. The stimulation tests involve children having a drip placed in an arm, administration of medication via the drip (this medication causes the body to produce growth hormone) and many blood samples over the next 3 hours. The medication often have side effects such as nausea and vomiting. These tests require the child to be admitted to a hospital ward for the morning and a dedicated nurse to do the test thus they are relatively expensive for the NHS.

We have started to develop a new method for testing if a child has low growth hormone levels using the level of gene activity in the blood. This test uses a single blood sample, does not require the use of mediation (thus avoiding any unpleasant side effects) and would not require the child to be admitted to hospital. The test works by measuring the activity levels of all the genes in the blood and the using a mathematical technique called Random Forest Analysis to tell whether children have low growth hormone levels or not.

Our initial studies have indicated that this test works extremely well but before it is rolled out across the NHS we need to test it on larger number of children to be completely sure that it works. We will therefore recruit another 60 children with low growth hormone levels and 60 children with normal growth hormone levels from the Royal Manchester Children's Hospital. Each child will have a blood sample taken to measure their gene activity levels and we will then compare how accurate our test is compared to the results of the current growth hormone stimulation tests.

Once children are diagnosed with growth hormone deficiency they are treated with injections of growth hormone once each day. Although generally this treatment works well sometimes children don't respond to the treatment. At the moment the only way of telling who will respond is to give them injections for 6-12 months and measure how well they grow. We have been able to show that the gene expression levels measured on the same test we use to diagnose Growth Hormone Deficiency can also tell us who will not respond to treatment. It would be great to identify those children who won't respond to treatment so we could either give them a larger dose of growth hormone or in some cases we may choose not to treat them (thus the child avoids up to 1 year of ineffective daily injections). Once again we need to confirm our initial findings in a second group of children. For those children with growth hormone deficiency recruited to the study we will look 1 year after treatment to see who has responded well and who has responded badly. We will then use the gene expression levels and the Random Forest Analysis technique to validate out technique for predicting who does not respond well to treatment.

The result of this study is that with one blood sample we will be able to
1. Tell if a child has low growth hormone levels - saving the child side effects from medication, the need for multiple blood tests and saving the NHS significant bed and nursing time
2. Tell whether a child will grow well when treated with growth hormone. This may save the child a year of unnecessary injections or allow us to use a higher dose to ensure they do benefit from the injections.

Over 2,000 children are assessed for growth hormone deficiency (GHD) using pharmacological stimulation tests in the UK each year. These tests involve the child fasting, having multiple blood samples taken via an intravenous cannula and require administration of medications which cause side effects e.g. nausea and vomiting. Of the 2000 children evaluated around 500 of the children will be diagnosed with GHD and started on treatment. Unfortunately around 25% will show a poor response to GH treatment. Currently response can only be assessed after around 1 year of therapy which requires daily subcutaneous injections at an average cost of £7500 per year.

We have shown in a preliminary study that it is possible to predict both the diagnosis of GHD and response to GH therapy using peripheral blood gene expression data measured using Affymetrix arrays. In 98 GHD children and 20 controls using rank regression we identified 347 genes whose expression correlated with peak GH concentrations on the stimulation test. A Random Forest Algorithm using these genes gave an AUC of 0.98 for predicting GHD status (GHD versus controls). Using the same gene expression data, this time correlating gene expression with height velocity over the first year of therapy we were able to predict a poor response (defined as lower quartile) with an AUC of 0.89.

This study aims to recruit 120 patients (60 with GHD and 60 non-GHD short stature patients). Peripheral blood gene expression samples will be taken at the same time as the GH stimulation test will be assessed with RNASeq. The random forest algorithms for predicting GHD status and response to therapy will be developed using the technique above in a test set of 80 patients and validation set of 40 patients. The algoirhtms will be developed separate code and made available to diagnostic services.

The net result of this study is to produce a test based on a single blood sample to both diagnose GHD and predict response to treatment.

Impact on the NHS:
The Royal Manchester Children's Hospital serves an area with a population of ~6 million (around 9% of the UK population) and undertakes approximately 200 growth hormone stimulation tests per year with ~50 children testing positive for growth hormone deficiency. Assuming similar rates of testing across the UK this equates to 2,200 paediatric stimulation tests in the UK each year. With each test requiring occupation of a bed for 1/2 day there is a potential saving of 1,100 bed days per year in paediatric endocrine investigations units. The cost per test listed on the NHS National Schedule of reference costs is between £476 and £539 for a paediatric endocrine day case admission. Thus current total yearly costs are up to £592,000 per year. With commercial RNA sequencing costs currently as low as £150 per sample and the potential to automate the analysis process the replacement of GH stimulation tests with out gene expression based test has the potential to substantially reduce costs of investigations as well as save bed days. Importantly current tests require experienced nurses while the new test will require only a phlebotomist - potentially freeing experienced nurses for other tasks. In some centres GH stimulation tests require the presence of a doctor, with the small number of doctors working within paediatric endocrinology services the potential saving of their time will have a substantial impact on patient care.

Around 500 children start on GH treatment or Growth Hormone Deficiency in the UK each year. Depending upon definition of poor response used up to 25% of patients treated will show a poor response to treatment. Annual cost of treatment is around £7500 per year. Currently to assess response requires 1 year of therapy and with 25% of children displaying a poor response in the UK we spend £937,500 treating 125 children with limited benefit. Developing a gene expression based prediction test used prior to start of treatment will allow us to reduce costs by not treating where therapy is likely to be ineffective.


Impact on Patients and their Families:
~2,200 patients currently attend UK paediatric endocrine centres for GH stimulation tests requiring them to arrive in hospital fasting, have multiple blood tests and are given medications such as arginine, clonidine and glucagon which cause side effects such as nausea and vomiting. Clearly replacing the current test with a single blood sample will be significantly less burdensome for patients. In addition each patient will be able to have the new test in clinic when they see their physician reducing the need for an extra day off school (for children) and off work (for parents).

The test to predict response in children with GHD would avoid 1 years ineffective injections in 125 children -avoiding over 46,000 ineffective subcutaneous injections per year. The burden of these daily injections should not be underestimated in children.

Worldwide impact: The issues surrounding the use of GH stimulation tests and poor response to treatment are not confined to the UK - both of these issues are present worldwide and the potential impact of our new test within paediatric endocrinology is substantial.