From oxygen electrode to sensor on the arm: How FreeStyle Libre learned to read glucose

A person passes a reader over a small disc on the back of the arm. A glucose value appears, followed by an arrow. The gesture takes a second and looks almost effortless.

Nothing behind that number is effortless.

The sensor is not in a blood vessel. It does not measure milligrams per decilitre directly. It sits in subcutaneous tissue, where glucose arrives after leaving the circulation. An enzyme reacts with that glucose. A polymer carries electrons from the enzyme to an electrode. Electronics measure a minute current. Software turns the signal into a glucose estimate and a rate of change. A person may then use that information to decide whether to eat, exercise or administer insulin.

That last step changes the engineering question completely. A laboratory instrument can be judged by analytical performance under controlled conditions. A wearable glucose system must remain dependable while attached to moving, sweating, sleeping human beings, across production lots, temperatures and days of wear. A falsely high value can lead to excessive insulin and severe hypoglycaemia. A falsely low value can lead to missed insulin, hyperglycaemia or diabetic ketoacidosis. The US Food and Drug Administration, or FDA, explicitly considered those consequences when it reviewed the first American FreeStyle Libre.¹

Approval therefore did not mean that every Libre value was correct. It meant that the complete evidence supported reasonable assurance of safety and effectiveness for the intended use, with known limitations, warnings and risk controls.¹

How did a reaction first proposed in 1962 become a product trusted for treatment decisions in 2017? The answer is a fifty five year sequence in which chemistry, physiology, manufacturing, software, clinical evidence and regulation repeatedly exposed one another's weaknesses.

First understand the measurement chain

Libre measures glucose in interstitial fluid, the water based fluid surrounding cells in subcutaneous tissue. It does not continuously sample venous blood, serum or plasma. Its electrochemical sensor oxidises glucose from the interstitial fluid and produces a current. The reader then applies signal processing to the sensor data and displays a glucose result.¹

The chain can be stated without pretending that the intermediate steps are identical:

Plasma glucose → interstitial glucose → glucose flux into the sensor → enzyme reaction → electron current → digital signal processing → displayed glucose value → treatment decision

Glucose reaches the interstitial compartment from the circulation. In a human tracer study, the mean physiological delay from the intravascular to the interstitial compartment was approximately 5.3 to 6.2 minutes.² That figure is not a universal correction factor. Local blood flow, tissue site, the direction and speed of glucose change, sensor dynamics and signal processing all influence the difference visible at a particular moment.³

This produces the familiar reversal around a meal. During a rapid rise, plasma glucose can be higher than interstitial glucose because the circulation changes first. During a rapid fall, interstitial glucose can temporarily remain higher. Abbott's instructions therefore warn that sensor and blood glucose may differ during rapid change and that symptoms which do not match the sensor require a blood glucose check.⁴

The word blood also hides an analytical detail. A finger stick meter measures a whole blood specimen, but modern meters generally report a plasma equivalent glucose concentration. Plasma contains more water per unit volume than whole blood, and glucose is dissolved in that water. The International Federation of Clinical Chemistry recommended plasma equivalent reporting so that results from different specimen types could be compared more consistently.⁵

The red curve below therefore represents modeled plasma glucose, not an unconverted molar concentration per litre of whole blood.

Three curves: Plasma, interstitium and the displayed value

The following chart is an illustrative mathematical model, not measured patient data and not a reconstruction of Abbott's proprietary algorithm. It shows a meal shaped plasma glucose excursion, a first order interstitial response and an illustrative device response. The controls expose the assumptions rather than hiding them.

At the default setting, the modeled plasma curve peaks first. The interstitial curve follows with a six minute response time. The third curve adds a two minute modeled display response. Setting that additional response to zero makes the displayed and interstitial curves overlap exactly.

The separation is clearest around the peak. Plasma glucose may already be falling while the interstitial concentration is still rising. That is not a contradiction; the tissue is responding to what happened in the circulation a few minutes earlier. A real device result can differ for additional reasons, including sensor sensitivity, local tissue conditions, temperature, interference, calibration and filtering.

The chart also demonstrates why correlation is not enough. Two curves can rise and fall together and therefore correlate strongly, while individual paired values still differ at clinically important moments. Device evaluation must examine agreement across glucose ranges and rates of change, not merely whether the overall shapes resemble one another.

1962: The first enzyme electrode borrows oxygen

The scientific story begins before wearable electronics and before modern medical device software existed.

In 1962, Leland Clark and Champ Lyons described an electrode system for continuous chemical monitoring. Their glucose concept placed glucose oxidase close to an oxygen electrode. Glucose oxidase is an enzyme, a protein catalyst that selectively reacts with glucose. As the enzyme consumed glucose and oxygen, the electrode detected the change in oxygen.⁶

In 1967, Stuart Updike and George Hicks immobilised glucose oxidase in a gel over an electrode. Immobilisation means holding the enzyme in place while allowing small molecules to reach it. Their enzyme electrode could measure glucose in biological solutions and tissue in vitro.⁷

The idea was elegant: give an electrode chemical selectivity by placing a biological recognition element in front of it. But the first generation contained a hidden dependency. Oxygen was not merely part of the environment; it was part of the measurement route. In subcutaneous tissue, local oxygen availability varies. A sensor signal could therefore reflect both glucose and the oxygen available to support the reaction.

The enzyme could recognise glucose. The unresolved problem was how to move its electrons to an electrode without letting tissue oxygen control the measurement.

1984: A molecular courier replaces oxygen

Researchers in the 1980s introduced redox mediators. A redox reaction transfers electrons. A mediator is a molecule that accepts electrons from an enzyme and passes them to an electrode.

In 1984, Anthony Cass and colleagues reported a ferrocene mediated glucose enzyme electrode. Ferrocene derivatives provided a deliberately chosen electron transfer route, reducing the need to infer glucose from oxygen consumption or hydrogen peroxide production.⁸

This solved one problem and revealed another. A soluble mediator could diffuse away. A long lived implanted sensor needed the mediator, enzyme and electrode to remain in reliable electrical contact while surrounded by fluid and moving tissue.

The courier had to become part of the sensor architecture.

1987 to 1992: Adam Heller wires the enzyme

The catalytic centre of glucose oxidase is buried inside the folded protein. An electrode cannot simply touch that centre and collect its electrons. Adam Heller and his collaborators developed chemical relays and then redox polymers that connected, or wired, enzyme reaction centres to electrodes.⁹

The decisive structure was a hydrated, three dimensional polymer network carrying many redox sites. In the glucose sensor lineage, osmium complexes acted as electron relays. The enzyme transferred electrons to nearby osmium sites; electrons then moved through the network until they reached the working electrode. The mediator no longer needed to float freely through the tissue. It was retained inside the crosslinked sensing layer.¹⁰

In simplified language, glucose oxidase provides molecular recognition, the osmium polymer provides the electron pathway, and the electrode converts electron flow into a measurable current.

This arrangement offered two important engineering advantages. First, it reduced dependence on local oxygen. Second, the Navigator lineage could operate at approximately 40 millivolts relative to an Ag/AgCl reference electrode. A low operating potential reduces the oxidation of other electroactive substances that could otherwise imitate part of the glucose signal.¹¹

The word amperometric simply means that the system measures electrical current. It does not mean the electrode directly reads a concentration in mg/dL. Under controlled diffusion conditions, more glucose reaching the enzyme produces more electron transfer and therefore more current. Calibration and signal processing are required to map that current to a clinically meaningful glucose estimate.

This was the electrochemical foundation of the later Abbott systems, but it was not yet a product. A useful continuous monitor also needed a reproducible membrane, stable enzyme activity, miniature electronics, insertion hardware, sterilisation, packaging, software, calibration and clinical evidence.

1994 to 2000: TheraSense chooses the easier first product

Heller later wrote that he and his son Ephraim founded TheraSense in 1994 and obtained venture financing in 1996.¹² The distinction matters because company formation, financing and product launch were separate steps, not one event.

By the mid 1990s, the team had prototypes for both a blood glucose meter using an extremely small sample and a wearable continuous monitor with a miniature subcutaneous sensor. The continuous product was the more ambitious goal. The blood glucose strip was the more practical first route to market.

That choice was strategically important. A disposable strip and hand held meter forced the company to industrialise electrode printing, enzyme chemistry, calibration, electronics and quality control without first solving every problem of multi day implantation.

The FreeStyle blood glucose monitoring system received FDA clearance in January 2000 and entered the market that year. It required approximately 0.3 microlitres of blood, allowing sampling from sites that many users found less painful than conventional finger sticks.¹³

The product did more than generate revenue. It converted wired enzyme chemistry into high volume manufacturing experience. A reaction that worked once in a laboratory had to work across large numbers of strips, environmental conditions and ordinary users.

2003 to 2004: Human data turns a platform into an acquisition

The continuous sensor programme advanced in parallel. In a 2003 study, 48 wired enzyme sensors were implanted for three days in people with type 1 diabetes. The sensors were calibrated using capillary blood glucose and compared with venous plasma glucose sampled every 15 minutes. The investigators reported that 98 percent of readings fell in the clinically accurate or clinically acceptable zones of the Clarke error grid.¹⁴

An error grid does not ask only how far two numbers differ. It asks whether the difference would be likely to lead to a clinically dangerous decision. This was an early sign that continuous glucose monitoring could not be evaluated as chemistry alone.

In January 2004, Abbott announced that it would acquire TheraSense. The transaction closed in April for approximately $1.2 billion in cash.¹⁵ Abbott acquired a blood glucose business, the wired enzyme platform, manufacturing knowledge and a continuous sensor programme already supported by human data.

The pieces were now held by an organisation with global diabetes operations, regulated manufacturing and the resources to carry a complex system through large clinical programmes.

2007 and 2008: Navigator proves the platform, but not the experience

The direct technical predecessor of Libre was the FreeStyle Navigator.

Navigator used a narrow plastic sensor approximately 0.6 millimetres wide. Carbon working and counter electrodes and an Ag/AgCl reference electrode were printed in a stacked geometry. The active wired enzyme layer covered only about 0.1 square millimetres of the working electrode and was covered by a membrane that limited glucose flux.¹¹ The sensor sat approximately 5 millimetres under the skin and transmitted glucose information to a receiver every minute.¹⁶

The membrane was crucial. Without limiting glucose transport, the enzyme layer could consume oxygen or substrate in ways that made the response nonlinear and unstable. A flux limiting membrane turned uncontrolled exposure into a more predictable relationship between external glucose and sensor current.

Navigator received a European CE mark in June 2007 and FDA premarket approval on 12 March 2008.¹⁷¹⁶

Its regulatory position, however, exposed how far the product still had to travel. The original US Navigator was adjunctive. Users were required to confirm a reading with a conventional blood glucose test before changing therapy. It also required finger stick calibration, used a separate transmitter and receiver, and was intended for up to five days of sensor wear.¹⁶

Technologically, Navigator demonstrated the complete wired enzyme chain in a commercial continuous monitor. As an everyday product, it still asked users to insert, calibrate, carry, confirm and maintain several components.

The next challenge was no longer to prove that continuous wired enzyme sensing could work. It was to remove the rituals surrounding it.

The product problem: Factory calibration

A sensor calibrated by its user can be adjusted against that person's finger stick result. A factory calibrated sensor cannot rely on that rescue step. The relationship between sensor current and displayed glucose must be predictable before the package is opened.

That requirement moves the centre of innovation from one elegant reaction to an industrial system. The manufacturer must control, among other variables:

• enzyme loading and activity;
• redox polymer composition;
• membrane thickness and permeability;
• electrode geometry and printed dimensions;
• baseline current and sensitivity;
• temperature response;
• sterilisation effects;
• ageing during storage;
• drift after insertion;
• lot to lot variation; and
• the relationship between in vitro characterisation and in vivo performance.

Small variation in any one layer can propagate through the final result. A membrane that is slightly thicker changes diffusion. A change in enzyme activity changes current. A change in current changes the input to the calibration model. The displayed value may then influence an insulin decision.

Factory calibration was therefore not simply the removal of a user step. It was a claim that manufacturing and characterisation had become predictable enough to replace millions of individual calibration events.

2010: A prototype lot tests calibration without finger sticks

A 2010 pilot study gives a glimpse of the transition. Twelve volunteers wore two Navigator systems in parallel during two consecutive five day sensor periods, providing four sensors per person and 48 sensors in total. The researchers used a prototype sensor lot selected for low variation in glucose sensitivity and investigated whether in vitro batch characterisation could support continuous sensing without in vivo calibration.¹⁸

The study did not by itself create Libre. It showed the nature of the problem. Factory calibration required predictable sensor sensitivity, stable performance over wear and limited variation between sensors and between people.

The chemistry had to become a population of manufactured objects whose error distribution could be measured and controlled.

2014: Europe gets the defining Libre proposition

On 3 September 2014, Abbott announced European CE marking for FreeStyle Libre. The product combined the features that defined the new category: a small sensor worn on the upper arm for up to 14 days, factory calibration, a current glucose result and trend information obtained by scanning the sensor.¹⁹

The first Libre did not merely miniaturise Navigator. It changed the user contract.

There was no routine finger stick calibration. There was no separate bulky transmitter to maintain. The sensor stored glucose data, and a scan returned the current value, recent history and trend direction. The interaction became quick enough to repeat many times without the physical and psychological cost of repeated finger puncture.

Europe's conformity assessment under the Medical Device Directive left less public technical detail than the later FDA file. We can verify the CE marking and the product claims announced at launch, but we cannot reconstruct every document reviewed by the notified body from public sources. The American review would later make much more of the evidence chain visible.

2015: Performance becomes publishable

A prospective performance study published in 2015 enrolled 72 adults with type 1 or type 2 diabetes. Participants wore Libre sensors for up to 14 days. Sensor results were compared with capillary blood glucose measurements and, during clinic visits, with venous glucose measured using a YSI laboratory analyser.²⁰

The reported mean absolute relative difference, or MARD, was 11.4 percent against capillary reference measurements and 12.0 percent against the YSI reference. MARD is the average absolute percentage difference between sensor and reference measurements. It is useful as a summary, but it does not describe the entire error distribution or tell us what happens in every low glucose episode.

The study also reported 86.7 percent of results in zone A of the consensus error grid and a mean time lag of 4.5 minutes, with a standard deviation of 4.8 minutes, relative to YSI. Accuracy remained relatively stable over the 14 day wear period.²⁰

Those figures were not evidence of perfection. They were evidence that a factory calibrated sensor could produce clinically useful agreement over two weeks without being adjusted by the user.

2014 to 2016: Accuracy is not clinical benefit

Analytical performance asks whether the sensor agrees with a reference. Clinical performance asks what happens when people use the information.

That second question was addressed in randomised trials. In the IMPACT study, adults with well controlled type 1 diabetes who used flash glucose monitoring spent 1.24 fewer hours per day in hypoglycaemia after six months than participants using conventional self monitoring, a relative reduction of 38 percent.²¹ The study was funded by Abbott.

The REPLACE trial studied insulin treated adults with type 2 diabetes. Its primary HbA1c outcome did not differ overall between groups, but the intervention reduced time in hypoglycaemia and improved treatment satisfaction.²² This study was also funded by Abbott.

These results matter because a device can be analytically respectable and still fail to improve care. It may be too difficult to use, may provide information at the wrong time or may cause users to react inappropriately. Libre's low friction interaction changed how often people could access glucose information, and the trials examined consequences that paired accuracy measurements alone could not show.

2016: The FDA receives a riskier proposition

Abbott submitted the original US Libre premarket approval application on 1 August 2016. The requested device classification described an invasive, non adjunctive, factory calibrated and user initiated glucose sensor.¹

Each word carried regulatory weight.

Invasive meant that a sensor tail penetrated the skin and remained in subcutaneous tissue. Factory calibrated meant that the user could not correct the sensor with personal finger stick entries. User initiated meant that the system delivered its information when scanned rather than continuously alarming in the background. Most importantly, non adjunctive meant that the displayed result could be used for diabetes treatment decisions without routine confirmation by a separate blood glucose meter.

That made errors more consequential. A falsely high result could prompt excessive insulin and severe hypoglycaemia, including seizure or loss of consciousness. A falsely low result could prompt unnecessary carbohydrate intake or insufficient insulin, increasing the risk of hyperglycaemia and diabetic ketoacidosis. An incorrect rate of change could also alter treatment even when the point estimate looked plausible.¹

The product could no longer be defended as additional information beside a trusted blood glucose measurement. Its own output had to become actionable.

What the FDA reviewed: The whole chain

The public FDA summary shows how misleading it would be to think of approval as a test of the enzyme electrode alone.

Bench testing addressed the system and its components. The review included electrical safety, electromagnetic compatibility, environmental operation, usability, reader mechanics, insertion device function, sensor sensitivity, variability, response time, linearity, temperature sensitivity and interfering substances.²³

The sensor component was tested for dimensional requirements, maximum current, current measurement range and resolution. Packaging, shipping and shelf life were evaluated. The sensor pack and introducer sharp underwent electron beam sterilisation validation. Biocompatibility testing covered the casing, adhesive and implanted sensor tail.²³

The FDA summary also states that software verification and validation confirmed that the system software performed according to established specifications and met its requirements.²³

In other words, the regulated product was the entire path from chemistry to behaviour:

glucose transport → enzyme reaction → sensor current → analogue electronics → digitisation → signal processing → calibration model → glucose result and trend → user interface → treatment decision

The proprietary algorithms did not need to be published in full. They did need to be specified, verified and validated within the manufacturer's quality and regulatory evidence.

2017: A design change forces new clinical evidence

The most revealing part of the FDA history occurred after the PMA had already been submitted.

Abbott had conducted one US study with 125 subjects and two studies outside the United States with a combined 126 subjects. After submission, the company modified the device to improve accuracy over the wear period. It amended the application with an updated design and new clinical data.²³

Fifty additional participants were enrolled in a new US pivotal study; 48 were evaluable. Each wore two sensors, one on each upper arm, for the claimed ten day US wear period after a twelve hour warm up. The study was specifically intended to validate the modified device and establish accuracy, precision, message performance and reading availability across wear.²³

This is medical device development in its least cinematic and most important form:

design → verification and validation → clinical evidence → identified performance problem → design modification → new verification → new clinical validation

The chemistry did not change history by being elegant once. The product became approvable because performance problems triggered controlled iteration and new evidence.

The final pivotal data — better, but not perfect

The design modification produced a visible improvement. MARD in the original US study was 12.5 percent. In the pivotal study of the modified device it was 9.7 percent against the YSI laboratory reference, based on 5,772 paired measurements.²⁴

Two simultaneously worn sensors generated 7,319 paired readings and agreed with a coefficient of variation of approximately 6 percent. That result measured sensor to sensor precision on the same person, not agreement with the laboratory reference.²⁴

Performance also changed over wear. The proportions of results within either 20 mg/dL or 20 percent of the laboratory reference, depending on glucose range, were:²⁴

Day of wear	Results within ±20 mg/dL or ±20%
Day 1	87.4%
Day 4	91.4%
Day 7	93.5%
Day 10	92.3%

The corresponding daily MARD values were 10.7 percent on day 1, 9.6 percent on day 4, 9.1 percent on day 7 and 9.3 percent on day 10.²⁴ The data show why one aggregate number cannot describe the entire device. Accuracy varies with concentration, rate of change and time since insertion.

The FDA also evaluated trend arrows separately. When Libre displayed the horizontal arrow, indicating change between minus 1 and plus 1 mg/dL per minute, the laboratory reference was in the corresponding slowly changing range 83.0 percent of the time. In a small proportion of cases, the reference was rising or falling faster than the arrow suggested.²⁴

This distinction matters. A patient may alter insulin not only because the displayed glucose is 150 mg/dL, but because the arrow says it is rising quickly. Rate estimation is therefore a clinical output in its own right.

The uncomfortable data — residual risk remains

The same FDA file documents errors that make the benefit and risk reasoning more credible, not less.

The reader displayed “LO” when it estimated glucose below its reportable range of 40 mg/dL. During the pivotal study, “LO” appeared on 21 occasions. None of the paired laboratory reference values was below 40 mg/dL. Four of the reference values, 19 percent, were above 80 mg/dL and fell between 81 and 120 mg/dL.²⁴

At a low glucose threshold of 70 mg/dL, the system displayed an appropriate low glucose message within a plus or minus 15 minute window for 85.4 percent of reference measurements below the threshold. It missed 14.6 percent. Among displayed low messages, 39.9 percent did not have a reference value below 70 mg/dL within that window.²⁴

These numbers should not be used to judge later Libre generations. They describe the first US system reviewed in 2017. They do reveal the regulatory logic:

Approval does not require zero error. It requires that remaining risk be understood, mitigated and acceptable in relation to the expected clinical benefit.

For that first US label, the reader displayed a symbol directing users to perform a blood glucose test in specified situations, including low readings, predicted low glucose, rapid change, absence of a trend arrow and a result above the reportable range.²⁵

Warnings were therefore part of the device's risk control, not an admission that the evidence had failed.

September 2017: The output becomes actionable

On 27 September 2017, the FDA approved FreeStyle Libre for adults with diabetes in the United States. The indication stated that the system was designed to replace blood glucose testing for diabetes treatment decisions.²⁶

That represented a larger regulatory step than Navigator's approval in 2008.

Navigator provided continuous glucose information, but treatment changes still required confirmation. Libre made the output of the sensor, calibration model, trend analysis and user interface sufficiently trustworthy to support decisions directly, within the stated limitations.

The FDA concluded that the pivotal evidence provided reasonable assurance of safety and effectiveness. It also required a post approval study to confirm performance in the intended population.²⁵

Approval was therefore neither a certificate of infallibility nor a single finish line. It was a decision that the evidence, controls and continuing obligations were proportionate to the intended use.

The deeper innovation: A reproducible error distribution

The glucose electrode was already decades old when Libre reached Europe. The deeper achievement was turning a biologically variable, drifting electrochemical sensor into a mass produced system whose errors were sufficiently predictable to manage clinically.

That required agreement between disciplines that are often discussed separately:

• physiology defined how plasma and interstitial glucose diverge during change;
• electrochemistry converted local glucose into electron current;
• membranes controlled diffusion and extended the useful range;
• manufacturing constrained variation between sensors and lots;
• electronics measured very small currents under daily life conditions;
• software converted signals into results and trends;
• clinical studies characterised accuracy, use and outcomes;
• human factors engineering made scanning and interpretation workable;
• regulation connected each failure mode to its possible clinical consequence.

No single discipline could substitute for the others. Better chemistry without controlled manufacturing would remain a promising experiment. A stable sensor without usable insertion and wear would remain a prototype. A convenient product without sufficient accuracy would be dangerous. An accurate point estimate without reliable trend behaviour would provide an incomplete basis for treatment.

The small disc on the arm is therefore not merely an electrode. It is the visible end of a validated measurement model.

What the public evidence cannot tell us

The public record establishes the electrochemical principle, factory calibration, device components, broad signal processing function, verification categories and clinical performance.

It does not disclose Abbott's current source code, coefficients, lot characterisation methods or proprietary compensation algorithms. It would be wrong to infer them from the illustrative chart in this article.

It would also be wrong to assume that one fixed delay converts interstitial glucose into plasma glucose. The relationship changes with physiology, local tissue conditions, rate of change, reference method and sensor behaviour. The chart is a teaching model whose assumptions are visible and adjustable. It is not hidden product knowledge.

Frequently asked questions

Does FreeStyle Libre measure plasma glucose directly?

No. The sensor measures an electrochemical response to glucose in subcutaneous interstitial fluid. The system converts that signal into a displayed result intended to be clinically comparable with plasma equivalent glucose measurements.¹

Why can Libre and a finger stick differ?

They sample different physiological compartments and use different measurement systems. During rapid change, glucose in the circulation changes before glucose in interstitial fluid. Both the sensor and the blood glucose meter also have measurement uncertainty.²⁴

What does factory calibration mean?

It means the user does not establish calibration by entering repeated finger stick results. The manufacturer characterises the relationship between sensor response and glucose before use. It does not mean that every sensor value is exact.⁴

Why did Navigator still need confirmation?

The original US Navigator indication was adjunctive. Its continuous readings supplied information and alarms, but users were instructed to confirm glucose with the built in blood glucose meter before adjusting therapy.¹⁶

Did FDA approval prove every Libre result correct?

No. The FDA evaluated the complete benefit and risk balance, including known inaccuracies, false and missed messages, warnings and conditions in which a blood glucose test was required.²⁵²⁴

The life behind the invention: Adam Heller

Adam Heller's biography gives this technical history a human depth that should not be turned into a tidy invention myth.

He was born into a Jewish family in Cluj, Romania, on 25 June 1933. In 1944, when he was eleven, he and his family were caught in the destruction of Hungarian and Transylvanian Jewry. A University of Texas profile records that he was imprisoned in the Kolozsvár Ghetto and then in the Bergen-Belsen concentration camp. Heller and his immediate family survived through the Kasztner transport, which carried a group of Jews out of Nazi controlled Europe. The family later emigrated to what was then Mandatory Palestine.²⁷

Heller studied chemistry and physics at the Hebrew University of Jerusalem and completed his doctorate in 1961. His career included postdoctoral work at the University of California, Berkeley, research leadership at Bell Laboratories and, from 1988, the University of Texas at Austin. His work ranged from liquid lasers and solar energy conversion to lithium batteries, bioelectrochemistry and medical sensors.²⁸

He has connected scientific work to survival and responsibility. The University of Texas account describes both his family's Holocaust experience and the later death of a daughter after a misdiagnosis as forces behind the family's commitment to technologies that reduce suffering. In Heller's words, “I owe it as long as I can work.”²⁷

In 2008, President George W. Bush presented him with the 2007 US National Medal of Technology and Innovation for contributions to electrochemistry and bioelectrochemistry that led to products improving millions of lives.²⁸

The medal recognised work far broader than glucose monitoring. Yet wired enzyme sensing may be its most intimate expression: a difficult route for electrons became, through decades of engineering and evidence, a routine act of care.

Sources and notes