Microfluidics is a potentially transformative technology in healthcare. Dozens of academic papers have detailed new approaches for using microfluidic chips to diagnose a wide range of conditions. Take cancer: the use of a tiny sample of blood or saliva to diagnose the presence of cancer cells, rather than the traditional biopsy, would be much less invasive for the patient, and could help diagnose the disease at a much earlier stage. In developing countries, the technology could make it possible to slow down or halt an epidemic of a lethal disease such as Ebola.

Researchers have even suggested that microfluidic chips could be used to identify whether someone has suffered a traumatic brain injury. Yet, so far, the commercial potential of microfluidics for diagnosing diseases has not been realised.

The technology of microfluidics operates at the micron scale, a micron (μm) being a millionth of a metre (a human hair is about 100μm thick). A tiny chip, a few square millimetres in size and made from acrylic, glass, silicon or polydimethylsiloxane – a polymer – acts as an entire pathology lab. The tiny, hair-width channels on each chip can contain minute amounts of blood or saliva, and the interaction of these droplets with a similarly small amount of reagent to detect a biomarker can produce a diagnosis very quickly. The effort, cost and time usually involved in collecting a sample, packaging and sending it to a lab, not to mention the tense wait for results, becomes almost non-existent.

Perhaps the most obvious use for microfluidicbased diagnosis is in countries where infectious disease is widespread, such as those in sub-Saharan Africa. In many of those countries, there are limited laboratory facilities for testing and patients may find it hard to travel to a clinic to have a sample taken. A lab-on-a-chip could enable a patient to be diagnosed instantly in their own home by a health worker, enabling them to be treated at an earlier stage of a disease and, if necessary, isolated from others.

Microfluidics make it easy

In the era of Covid-19, when diagnosis can still take several days, the technology could also be transformative in the developed world. Hsueh- Chia Chang, Bayer professor of chemical and biomolecular engineering at the University of Notre Dame, says: “In the US, when we do Covid-19 screening, we do the swabs and we put the swabs into a vial containing virus transport medium, and then we have to send the whole vial to a centralised laboratory. The average turnaround time is four or five days. So, [for] four or five days, the carrier could be infecting other people.

“It almost makes the screening test irrelevant because the turnaround is so slow. So, it would be nice if we could do the detection within half an hour or an hour after the sample is taken. That really makes a difference in the very rapidly spreading pandemic that we have now.” It is an area his own lab has been working on.

There is a big difference, however, between making a prototype microfluidic device that works and creating a commercial product. “We can make anything larger than, say, 100μm, very well, but anything below 100μm and it becomes challenging,” says Chang. “It’s difficult to prototype, it’s difficult to design, it’s not very robust, it can be clogged by debris and the products are not very reproducible.”

There are two main barriers to making a mass-producible product, says Chang. A chip is manufactured by making thin grooves on the surface and then enclosing each groove by making a second layer, forming a microchannel. The layers must be bonded together in such a way to make the channels leakproof. The first barrier, says Chang, is that the manufacturing of microchannels and microchips is not easy. “You have to use microinjection moulding, so even prototyping is very difficult.” Although 3D printing has been suggested as a way of manufacturing microchips, in practice it has many limitations. “You cannot do 3D printing as a way of prototyping.”

The second barrier is that microfluidic channels “will be clogged by the debris in the sample”. Any heterogeneous sample is, therefore, likely to cause clogging. “That means that you have to do quite a bit of pre-treatment,” says Chang. “You have to start with a larger volume or sample than you need and then remove all the debris.”

The advantages and limitations of the lateral flow assay

Some medical device companies have avoided the problems of manufacturing chips by employing similar lateral flow assay (LFA) technology, which is now used widely in the developing world. As with a chip-based microfluidic device, an LFA is portable and capable of producing a quick result. The challenges of manufacturing are avoided by using paper devices instead of chips. Because the paper is porous, says Chang, “it can filter out the debris”.

The downside, however, is that the sensitivity is low. “When it’s moving the debris by filtration through the paper, you quite often lose the target molecules that you want to measure.”

“We may not recognise it as a microfl uidics device but it will be based on some of the principles we work on.”

There is already a substantial market for LFAs because of their many advantages: they have a long shelf life, they are cheap to produce, easy to use and highly portable. In some cases – for example, in pregnancy tests, where the amyloid concentration is very high – the low sensitivity doesn’t make a difference. The problem arises with diseases such as malaria, where high sensitivity is essential to be sure of an accurate diagnosis.

In a nutshell, the problem, says Chang, is that droplet microfluidic devices have a high sensitivity but are very expensive, and, in their present form, are not portable. LFAs, on the other hand, are portable and inexpensive, but not very sensitive. The holy grail is to develop a microfluidic device that is “both sensitive and also very cheap to manufacture”.

Nonetheless, there are a number of companies marketing microfluidic technology – though principally for research rather than for diagnostics. Chang cites 10X Genomics as one that is having success with droplet microfluidics, in the form of a single cell assay that sequences RNA and DNA inside an individual tissue cell. The product, however, is extremely expensive. Other companies operating in the space include Dolomite – which is focusing on vaccine development – Ion Torrent Systems and Bio-Rad, which, in 2017, acquired RainDance Technologies for its microfluidic capabilities. While all are successfully producing microfluidic applications, the ability to mass produce microfluidic chips for point-of-care testing is proving elusive. Chang also cites Oxford Nanopore, which develops products based on nanofluidic technology, dealing with materials smaller than 1_m to carry out DNA/RNA sequencing, and is now working on high-speed Covid-19 tests.

Too great a challenge

Chang himself has licensed his technology to three start-up companies using microfluidic technologies. The first one, F Cubed, aimed to use microfluidics to detect pathogens in food. If successful, it would have made it quick and simple to detect in situ whether food was safe, simply by taking a swab of, for example, a piece of meat and running the sample through the biochip. Unfortunately, the company – in which Chang was the chief scientific officer – has now folded. “We just could not make the chips economically so we could sell them to the customers,” says Chang.

The second, in which Chang also owns equity, is AgenDx, which is looking to find ways of detecting pancreatic cancer through microfluidics. The third is ImpeDx, which is using microfluidic technology to develop tools for antimicrobial susceptibility testing (AST), so that patients with bacterial infections can be treated with the most suitable antibiotics. Both companies are “trying to come out with technologies that will not cost as much to manufacture”, says Chang. Much of the work involves “trying to integrate membrane technologies into the larger channels – the same kind of membranes we use for ultrafiltration, for dialysis or even waste water treatment. These membranes are cheaper to make, he says. “The industry has already figured out how to make these and so it’s a matter of assembling them into chips – they will be more advanced than the paper membranes on neutrophil assays, but hopefully more precise.”

Chang is still optimistic about the potential of the technology. He believes that the use of microfluidic point-of-care devices to diagnose cancer is still some way off, but expects the technology to diagnose infectious disease will be available more quickly, particularly as Covid-19 has created a greater sense of urgency. Instead of making chips at a very small scale, it’s more likely that they will employ the filtration technologies to clean up the sample and use larger channels and larger devices that are easier to manufacture. “We may not recognise it as a microfluidics device but it will be based on some of the principles we work on,” says Chang. “It may not be a manufactured chip – it may just be a tube with some integrated membranes in there.”

IBM Research’s approach to high-volume microfluidics

“In a typical lateral flow assay, you need maybe 100μl of a sample,” says Yuksel Temiz of IBM Research. “In our case, maybe we can do the same using 10μl. So, we need much less of the sample, much less reagent and, at the end, the results are much more precise.”

More than that, the company’s reconfigurable chip, which is discussed in more detail on page 99, has been designed from the ground for high-volume manufacturing.“Many research groups in microfluidics show one prototype that takes two days to fabricate,” Temiz continues, “but, because we work with industrial partners, we always had in mind that our technology should be manufacturing-compatible.”

From the first, that meant they had to steer clear of polydimethylsiloxane (PDMS), the most common material in microfluidics research. Despite its suitability for droplet-based microfluidic applications in particular, PDMS is both unstable over time and extremely difficult to manufacture at scale.

As IBM Research’s chip works through capillary action, PDMS would also have been too hydrophobic to generate the necessary liquid flow. Instead, the team uses a silicon substrate layered over with a photoresistant SU-8 resin – which is often used with PDMS – to form hydrophilic microfluidic channels and closes the chip with a similarly hydrophilic dry-film resist tape.

“The fabrication is based on silicon micromachining,” explains Temiz. “It’s very similar to how integrated circuits or CMOS chips are fabricated, but way easier. We use the tools developed for electronic circuit fabrication, but it’s more like microelectromechanical systems (MEMS) fabrication. We take a silicon or glass substrate for our base, and place and photopattern an SU-8 layer on top. When we want to put electrodes, we also put metal contacts, and sometimes we add chemical reagents using inkjet spotting. Then, at the end of the process, we simply laminate the chips for a closed channel at 45oC. Everything is done in a high-throughput compatible manner.”