The doctor will see your phone now: Sequencing and speedy diagnosis give chance of timely care

Positive results for genetic disease…will probably be the first thing primary care physicians experience from our work

A patient walks into a consulting room and begins rattling off some complex symptoms.

“OK,” says the GP, “let’s see what’s going on. Can I take a look at your phone?”

That’s the vision, says Stephen Kingsmore, president and chief executive of Rady Children’s Institute of Genomic Medicine in San Diego.

“We are working toward elevating a genome sequence from a medical test, a transaction, to a healthcare delivery system based on an ongoing, lifelong interaction with your genome,” Dr Kingsmore says.

He spoke to New Zealand Doctor Rata Aotearoa between presentations at last month’s Pathology Update 2023 in Melbourne.

Dr Kingsmore set a Guinness World Record for the fastest genetic diagnosis (26 hours) in 2015 and the team he leads at Rady followed up three years later with a 19.5-hour record.

Now he is running a multinational pilot, BeginNGS, which will use genome sequencing to screen newborns for about 750 genetic diseases before they get sick.

The sequences will be analysed and aggregated by an AI, “and the more sequences it sees, the more powerful it will become”, he says.

“They will become the data hogs of the hospital system.”

The longer-term aim is to make an individual’s genome accessible via a smartphone app: “Initially, mum and dad will control that app on behalf of their child, after initially consenting to their baby getting a genome sequence at birth. Then, when the baby is of age, they can opt in or out to continue.

“It’ll be a while before that becomes standard, but that’s happening now and that’s because the cost of genome sequencing has dropped to about US$100 ($161.40), so it’s now realistic to think about screening an entire population.”

The genome is derived from a dry blood spot via the heel-prick test, with positive results for genetic disease being returned to the GP along with follow-up information.

“That will probably be the first thing primary care physicians around the world are going to experience from our work,” Dr Kingsmore says.

The other area will be with children admitted to intensive or urgent-care facilities: “We are already seeing broad adoption of genome sequencing for diagnosis in the US, and that is offering benefits, with children being discharged much more quickly. Because we make a rapid diagnosis, they are getting on to effective therapy.”

In his plenary presentation, Dr Kingsmore used a case study from his hospital in San Diego to illustrate the possibilities.

A baby aged five weeks arrived at the emergency department at 10.49pm. The mother said he had been agitated, restless, would not take the bottle and had a downward eye deviation. A CT scan showed lesions in the brain. The parents later told staff his older sister had been admitted at the same age with the same symptoms, 10 years earlier, and had died.

A blood sample was taken at 3.25pm the following day, and by 7.24am next morning, the completed sequencing was uploaded to the artificial intelligence. The provisional diagnosis of autosomal recessive biotin-thiamine-responsive basal ganglia disease followed at 7.34am.

The baby started having seizures at 8.11am and treatment, based on the AI result, started at 12.13pm, with the baby becoming alert and feeding normally by 6pm.

“Ten years ago, his sibling died of an irreversible condition,” says Dr Kingsmore, “now this is a treatable condition, and that baby has an excellent prognosis...this is the promise of genome sequencing.”

GPs will also be involved should the AI deliver bad news, he says.

“If there isn’t an effective therapy and it is a hopeless prognosis, we can at least give the family the power of an answer and allow them to drive the decision-making in terms of palliative care and hopefully getting the child home, and GPs will be engaged in this process.

“Such events will be uncommon until testing becomes industrialised, but they will soon occasionally see these children who have a genome sequence as part of their medical record.”

Without the aid of genome sequencing, Dr Kingsmore says the average time for the diagnosis of a genetic disease is 4.8 years and involves on average 7.3 specialists.

However, he estimates there are about 30 million patients with rare diseases in the US alone.

Seventy per cent of known genetic diseases start in childhood, and 2.5 per cent of children are affected. “One of the myths regarding rare disease is that they are rare and very difficult to treat and we have to change that,” he says.

To succeed will require changes in diagnostic approach, and Dr Kingsmore illustrates current attitudes with a medical school mantra from the 1940s: “When you hear hoofbeats, you think horses not zebras.”

In Africa, he says, that paradigm is reversed: when you hear hoofbeats, you think zebras rather than horses. “A lot depends on context, and we need to relearn our thinking in terms of differential diagnosis.”

The healthcare system remains “broken” for children with genetic diseases, with only an estimated 20 per cent of cases being diagnosed.

With the US having a relatively high infant mortality rate of 0.4 per cent, he was part of a study into how many deaths may be attributable to genetic disease and whether death could have been avoided if practitioners had been warned.

“The answer, shockingly, was, yes, in about a half of cases, although we still had a proportion who were given the appropriate drugs who still died.”

He says underdiagnosing remains a leading cause of infant mortality despite the relatively high volumes of genome sequencing. “We found 62 per cent of death certificates do not mention a genetic disease diagnosis, which was also a shock.”

Dr Kingsmore says there are now 750 genetic diseases with “somewhat effective treatments”, out of 75,000 known genetic diseases. But, he says, an era of effective new cell therapies, gene therapies, gene editing and protein therapies has arrived.

“In some cases, yeah, we can now say they are curative, but in other cases the issue is the timing, getting that effective therapy to the individual before symptom onset.”

A gene therapy is available for spinal muscular atrophy: “Hitherto, babies with that disease died at age two, but the anterior horn cells you lose before getting the gene therapy don’t regrow, so if you delay giving that therapy even by a couple of months, you will have defects you are not recovering from.”

As well as enabling more timely treatment, Dr Kingsmore is hopeful BeginNGS will support further research into cures by highlighting previously unrecognised need. The use of AI should also improve standardisation of diagnosis, he says, and make it accessible to countries with limited pathology resources.

The programme also intends to over-sample minority ethnicities as the current state of genomic knowledge leans overwhelmingly towards white, northern Europeans.

But he accepts there are concerns about patient privacy and security – health records are already a target for hackers and insurance companies would love to know individual predispositions for disease. He hopes more countries introduce legislation similar to the US’ Genetic Information Nondiscrimination Act of 2008.

READ MORE:

• AI assistance is heading your way: ChatGPT not half bad at describing aspects of primary care
• The first time using ChatGPT: Have fun and don’t get left behind
• Watch out, the MCEDs are coming

Further concerns were raised by New South Wales health pathology clinical director Elizabeth Salisbury during a Pathology Update 2023 plenary session. Dr Salisbury says application of AI is expanding rapidly.

“There won’t be an area in our lab that will be left untouched...and it is going to create an ability to personalise medicine in ways we haven’t previously thought about; everything is now up for grabs,” she says.

Her concern is the new technology will be embedded before the necessary safeguards are in place and debates on its use have been thrashed out. She pointed to the role of AI in the global financial crisis of 2007 as an example of unforeseen consequences.

Dr Salisbury says more needs to be known on how legal disputes involving AI and algorithms will be resolved. “Who will have liability? Who owns the diagnostic information derived from an AI’s processing capabilities, is it the provider or the developer? Who will be the ultimate source of truth (in diagnosis), the pathologist or the AI?”

She suspects practitioners such as pathologists may be more likely to take shortcuts because of the assumed accuracy and infallibility of an AI-derived diagnosis. “[We must introduce AI] in a way that is transparent and matches the needs and expectations of society,” says Dr Salisbury.

“We all know the US is a big driver in the medical environment and practices on a global basis and that it will drive the uptake of AI...

“What we are seeing now is just a taste of all the different things [medical practitioners] do on a daily basis that have the potential to be modified or totally performed by AI.”

While predicting entirely new categories of patients and care will soon be created using data from AI systems, Dr Salisbury says the patient must remain at the focus.

“The use of AI algorithms can disconnect decision-making from the patient by not taking the whole person and patient-specific conditions into account, which may not be easily incorporated in the system.

“I don’t know the answers to my questions yet, but I do think they are questions we need to start asking.”

*Alan’s flights and accommodation were paid for by the RCPA