Georgia Burgess was four years old and seemingly winning her leukaemia battle when her right eye started turning inwards. Her doctors were stumped. For weeks, tests and scans revealed nothing.
Finally, a lumbar puncture showed that while her blood was clear, the fluid in her spinal cord and brain was laced with leukaemia cells and the extra pressure was causing her eye to turn.
"She appeared healthy and active and there were no other symptoms, but it showed she had a central nervous system relapse of her leukaemia and the prognosis was very, very poor," said her father, Jake Burgess.
"What's so hard is the powerlessnesss. You're charged with looking after your kids, the most important things in your life, and suddenly you have no control and you can't take the burden from them, an unfair burden."
What is noteworthy about Georgia's case is that when she was first diagnosed, she was placed in the "standard" risk group, meaning she was given less intensive treatment than high-risk patients.
While this tailored treatment plan has been highly praised for reducing the overall impact of the side effects of chemotherapy, which can include secondary cancers in later life, medical experts faced a conundrum: One in six children in the "medium" and "standard" risk categories were relapsing.
In a world-first, researchers at the Children's Cancer Institute have developed a way to identify children at high risk of relapse but currently "hiding" in a lower risk group and therefore not receiving adequate treatment.
They found the absence of two DNA fragments in leukaemia cells were important predictors of relapse.
To more accurately determine a child's risk of relapse, they have developed a new risk scoring system that combines information about these two specific gene "microdeletions" with minimal residual disease (MRD) test results and a National Cancer Institute risk score.
"There's always a play-off between getting the treatment right and the side effects, and right now we're dividing patients into risk groups so we can increase their chances of survival," said study leader Associate Professor Rosemary Sutton.
"But in years to come, treatment will be custom designed for that child, and we're doing our little bit to get that cure rate higher and higher."
Mr Burgess, managing director of the new Sydney Zoo being built in Blacktown, said had the findings been made earlier, Georgia, now nine, may have been placed in the high risk group, receiving more aggressive treatment and avoiding a relapse.
But he said it was exciting that incremental but substantial findings were being made and he was hopeful acute lymphoblastic leukaemia, the most common childhood cancer, would one day become a curable and preventable disease.
"There was an unusual MRD test and I distinctly remember the conversation with the doctor where he said, 'I'm contemplating changing the protocol to higher risk, it's another four months and it's increasingly toxic', and I remember the complete sense of dread and how much I wanted to avoid that," Mr Burgess said.
"But I can tell you now, the relapse was unbelievable, worse than the first time and significantly harder to treat, so anything to avoid it is a massive, positive step."
It's been four and a half years since Georgia had a bone marrow transplant as part of her post-relapse treatment. In six months' time, if all goes well, she will be declared cancer-free.
"She's going very well and physically strong, loved by the friends and teachers," he said.
Dr Toby Trahair???, study co-author and oncologist at Kids' Cancer Centre at Sydney Children's Hospital, Randwick, said a better stratification of patients at diagnosis improved outcomes for everybody.
"It will mean more kids can conquer this horrible disease, which only 50 years ago had survival rates of close to zero."
The study, published in the latest British Journal of Haematology, involved 475 patients from six children's hospitals in Australia and New Zealand.
The results have been validated by Dutch researchers and the system will be trialled in Australia next year.