PF V9: Garbage In, Garbage Out

I could hardly avoid this paper. I have previously researched pulse oximetry of the dental pulp, and designed a probe for dental use. My own work is amongst the studies referenced here. More importantly, 30 years ago I discussed the subject at length with Joseph M. Schmitt from the National Institutes of Health, Division of Research Services, Biomedical Engineering and Instrumentation Branch, Bethesda, Maryland, one of the biomedical scientists who had actually worked on the underlying physics of pulse oximetry in teeth.

Schmitt persuaded me of something rather important: measuring absolute oxygen saturation inside a tooth with a conventional pulse oximeter was going to be extraordinarily difficult, perhaps close to impossible, because the physics of light transmission through teeth does not behave like the soft tissues for which these instruments were calibrated. He demonstrated this with modelling as well as experiment. The problem was not whether light could pass through a tooth, or whether a pulsatile signal could be detected. It could. The problem was whether the displayed SpO₂ value could be trusted as a true measurement of pulpal oxygen saturation.

 That is the distinction this systematic review seems largely to have stepped around.

The authors set out to establish reference values for oxygen saturation in healthy mature permanent teeth. Forty studies were included, twenty entered the meta-analysis, and pooled values varied by tooth type from roughly 86% to 92%. The review is methodologically busy: multiple databases, prospective registration, risk-of-bias assessment, sensitivity analyses, random-effects modelling. It looks reassuringly scientific.

But none of that solves the more basic question: are these studies measuring what they think they are measuring?

A pulse oximeter does not directly measure oxygen saturation. It infers it from the behaviour of red and infrared light passing through tissue, using calibration algorithms developed for soft tissue. Teeth are a completely different optical environment. Enamel, dentine, pulp volume, tooth thickness, curvature, mineralisation, sensor geometry and probe position all alter the optical path. The pulp space itself is highly heterogeneous between teeth and between individuals. Trying to generate a universal tooth algorithm from that is not merely difficult; it may be fundamentally unstable.

Schmitt’s work was directed precisely at this problem. His conclusion was that conventional dual-wavelength pulse oximetry could not provide absolute oxygen saturation independently of tooth geometry and sensor position. Yet his paper is not referenced here.

That omission is extraordinary, because it is not a peripheral paper. It attacks the validity of the entire measurement concept.

Instead, the dental literature appears to have taken a rather circular route. A number of investigators placed pulse oximeter probes on teeth, obtained numbers, published those numbers, and subsequent investigators then treated the growing pile of numbers as evidence that the numbers must represent pulpal oxygen saturation.

They may not.

This is the scientific version of garbage in, garbage out. If the primary measurement is not validated, collecting more of it does not improve its validity. Pooling it does not improve its validity. Meta-analysing it certainly does not improve its validity. It simply sorts the garbage outbox more efficiently.

The heterogeneity in this review is therefore not a statistical nuisance to be managed with a random-effects model. It may be the signal telling us that the measurement itself is unstable. I² values are extremely high, often approaching 100%. Different studies used different oximeters, probes, adaptations and sensor holders. The authors themselves acknowledge that enamel and dentine thickness, tooth dimensions, pulp anatomy and sensor position can affect readings. 

At that point, one has to ask what exactly is being averaged.

If one system gives 82% and another 94%, and those values depend partly on tooth geometry, optical scattering and proprietary algorithms rather than actual haemoglobin saturation, the pooled estimate is not a physiological reference value. It is an average of instrument outputs.

That may sound pedantic, but it is the central issue.

The reported differences between tooth types illustrate the problem nicely. The review suggests that different teeth may have different normal oxygen saturation values. Possibly. But incisors, canines and molars also have very different dimensions, pulp spaces and optical path lengths. If the geometry affects the reading, then differences between tooth types may simply be optical artefact masquerading as physiology.

This is where dentistry has, in my view, become seduced by the number.

A digital display confers an extraordinary sense of authority. Ninety-one per cent looks objective. Ninety-one point seven three per cent looks positively scientific. But precision is not accuracy, and decimals do not rescue an invalid measurement model.

There is still something potentially valuable here. Pulse oximetry may be useful as a means of detecting a pulsatile vascular signal. That would be genuinely important. A demonstration of blood flow is a far more direct assessment of pulp vitality than asking whether a patient feels cold or electricity. In traumatised or immature teeth, that could have real clinical value.

But detecting circulation is not the same as measuring absolute oxygen saturation.

The authors call for standardisation, but standardising probes and holders is not enough. The real challenge is to establish, from first principles, whether an optical model can reliably account for the enormous anatomical variability of teeth. That requires biomedical optics, calibration modelling and validation against a defensible reference standard. Until that exists, a “normal pulpal SpO₂” remains an attractive number in search of physiological meaning.

What troubles me most is that the most important objection was raised decades ago by a scientist who understood the measurement physics far better than most of us in dentistry ever will. I spoke with Joe Schmitt at length and found his argument compelling. The dental literature seems instead to have proceeded on the assumption that if enough people measure something, it must eventually become real.

Science does not work that way.

Before we establish normal oxygen saturation values for the dental pulp, we should first establish that a pulse oximeter can actually measure oxygen saturation in a tooth.

Otherwise, we are not refining the evidence.

We are merely putting the garbage into smaller, neater bins.