Autism was formally described for the first time 71 years ago. The medical notes for "Case one", a 10-year-old from Mississippi, US, referred to as Donald T, describe a perplexing condition that was different from "anything reported so far". In 1943, when Donald Triplett was diagnosed, autism was considered extremely rare and treatment consisted of institutionalisation and – all too often – isolation.

Today we know "autism disorder" as one of a number of autism spectrum disorders alongside Asperger's syndrome, pervasive developmental disorder and single gene disorders such as Rett syndrome. But of all neuropsychiatric conditions, autism remains one of the least understood.

We now know that genetics almost certainly plays a key role, with researchers finding that if a family has one child with autism, then the likelihood of a future child having the condition is as high as 25%. But to what extent autism is defined by genes remains a mystery.

"Everyone recognises that genes are part of the story but autism isn't 100% genetic," says Professor Simon Baron-Cohen of the Autism Research Centre at the University of Cambridge. "Even if you have identical twins who share all their genes, you can find that one has autism and one doesn't. That means that there must be some non-genetic factors."

One of the most controversial theories about how autism develops is neuroinflammation. MRI scans of autistic patients have revealed abnormalities in the white matter – the wiring tissue responsible for connecting brains areas. Some scientists have drawn comparisons with multiple sclerosis, in which inflammatory processes attack the myelin sheath around the axons of brain cells, slowing down signalling and making it less efficient.

If neuroinflammation is involved in autism, this could potentially yield some fairly straightforward drug treatments involving anti-inflammatories, but the theory has yet to be proven and with a multitude of other possible explanations for these white matter abnormalities, not everyone is convinced.

The lack of a concrete theory for autism can hamper the process of diagnosis, because the condition shares a number of overlapping symptoms with other autism spectrum disorders. However, over the past decade, the entire field of neuropsychiatric disorders has undergone something of a revolution with the growing realisation that they are not only conditions of the brain but of the entire body, raising the possibility of detecting them in the blood.

One approach is to compare blood samples from autism patients and healthy individuals and search for what is known as a protein fingerprint – a set of protein levels that is consistently and markedly different in people with autism. So far this has been done relatively successfully in Asperger's syndrome, forming the basis of a blood test that can diagnose the disorder with 80% accuracy, and there are hopes this feat can soon be replicated for autism disorder.

While this research shows promise, however, there's still a long way to go before it becomes clinically available. "I think it could possibly happen within five years but it's premature to be thinking these tests are just around the corner," Baron-Cohen says.

"The whole ethos behind medicine is to do no harm and if the test is only 80% accurate, it means a proportion of people will be told they have the condition when they don't, so you've raised anxieties unnecessarily. Equally if the test is missing people, then they'll be going away thinking I'm fine when they could be getting support."

Whether measuring protein levels alone should ever be sufficient for a diagnosis is also open to question. Like all neuropsychiatric conditions, autism has varying degrees of severity, meaning some patients require constant care while those with "high-functioning autism" are capable of living independently, adapting to society around them and holding down a job. Right now, such a test would merely pool everyone with autism into the same category. Should we be intervening at all in some cases?

"It doesn't just rest on biology, it also rests on how well you're coping," Baron-Cohen says. "One of the criteria for a diagnosis in psychiatry is that the symptoms are interfering with your everyday life. If you have high-functioning autism, you may well have a lot of autistic traits but if you've got a particular lifestyle where it's possibly an advantage to be leading a solitary lifestyle and be quite obsessive, you're clearly able to function and maybe even make valuable contributions in your work, so arguably you don't need a diagnosis."

The most likely future scenario is that clinical assessment would be combined with a range of biological examinations including blood tests and possibly brain scans. But if a blood test for autism were to become available, it would be a major step towards one of the ultimate goals in the field – prenatal screening.

Once a conclusive biological trace has been identified, be it at the gene or protein level, this could be used at any point in development from before birth right through to adulthood. But unless the current diagnostic accuracy improves, there would be profound ethical concerns.

A large percentage of parents would almost certainly use a prenatal autism test to make a decision on whether to terminate the pregnancy – if the statistics for Down's syndrome since the introduction of prenatal screening are anything to go by. It is believed that around 90% of pregnancies in England and Wales that receive a diagnosis of Down's syndrome are aborted.

The existence of a prenatal screening test would also have implications for potential treatments. There are currently no drugs for treating autism disorder but in the near future, various hormonal treatments may become available. If clinicians were tempted to start medical intervention very early, there would be concerns about side effects in the unborn child. If the diagnosis proved to be faulty, the consequences of these decisions could have lasting effects.

"The best case use of a prenatal test at the moment would be if you could say to a parent, your child has got an 80% likelihood of autism and so once the baby's born, we would like to keep a close eye on that child in case they need extra support like speech therapy or social skills training or some sort of behavioural approach," Baron-Cohen says.

"That would mean that there were no potential side effects and you might be able to intervene at a much younger age. So from an ethical point of view, if there was a screening test, using it for early intervention via a psychological approach would be quite risk-free and could carry a lot of benefit."

David Cox researches neuropsychiatric disorders at Cambridge University, focusing on drug discovery and diagnosis