5-ALA fluorescence

Three hours before surgery, the patient drinks a glass of water with a dissolved powder. The compound is 5-aminolevulinic acid, a molecule the body normally uses to make the haem group of haemoglobin. Glioma cells metabolise it poorly: they accumulate an intermediate, protoporphyrin IX, which under violet light emits an intense red. In the operating room the surgeon switches the microscope filter and the tumour lights up pink against a brain that turns blue.

It is the most spectacular image in oncological neurosurgery, and one of the few operating-room technologies with a phase 3 randomised trial behind it. That trial is worth looking at closely, because what it demonstrates and what it does not are separated by a thin line that is almost never drawn.

The trial, and the question it does not answer

In 2006 the study that established the technique was published: 322 patients with malignant glioma randomised to conventional white light or fluorescence¹. The result was emphatic on its primary endpoint. Complete resection of contrast-enhancing tumour was achieved in 65 per cent of the 5-ALA group and 36 per cent of the white light group. Six-month progression-free survival nearly doubled: 41.0 against 21.1 per cent.

Now the part that is rarely quoted. The trial did not measure overall survival as its primary endpoint, and when it is analysed the difference does not reach statistical significance: a hazard ratio of 0.82, with a confidence interval running from 0.62 to 1.07 and therefore including the possibility of no effect at all². The most recent Cochrane review says it without hedging: there is no evidence of improvement in overall survival.

Resecting more and living longer are not the same sentence. In malignant glioma the relationship between the two is likely, is consistent with everything we know about the disease, and remains undemonstrated by a randomised trial. Twenty years on, that trial is still the only one, never independently replicated.

There is also a documented cost, reported by the authors themselves. A later analysis of the same study, across 349 patients, found that those in the fluorescence arm more frequently deteriorated neurologically at 48 hours³. The effect was transient — there was no difference in the Karnofsky performance scale — and concentrated in patients who already had deficits unresponsive to steroids. But it is there: resecting more aggressively has a price, and that price is not zero.

What it means when it glows

Fluorescence works extraordinarily well in one direction and very poorly in the other. That is the central asymmetry of this technique and the source of almost every misunderstanding.

On the good side: when tissue glows strongly, it is tumour. A study correlating visible fluorescence with histology found a positive predictive value of 100 per cent in strongly fluorescing tissue and 95 per cent in weakly fluorescing tissue⁴. Strong fluorescence corresponds to solidly proliferating tumour and high cell density; weak fluorescence, to infiltrating tumour. The meta-analysis of prospective studies gives an overall sensitivity of 87 per cent and specificity of 89⁵.

On the uncomfortable side: absence of glow means nothing. A meta-analysis of 32 studies and 947 patients in the setting of stereotactic biopsy calculated the negative predictive value of 5-ALA at between 8 and 11 per cent⁶. That is: if the specimen does not fluoresce, the probability that there is genuinely no tumour is around one in ten. In open resection the published values are better but still poor, ranging from 12.5 to 69 per cent.

Remarkably, the regulator built that limitation into the text of the authorisation. When the US agency approved the compound in June 2017 — a decade after Europe — it did so as an optical imaging agent indicated as an adjunct for the visualisation of malignant tissue during surgery, in patients with glioma suspected to be grade III or IV on preoperative imaging⁷. Not as a treatment, not as a determinant of margins. And the label carries an explicit warning: non-fluorescing tissue in the surgical field does not rule out the presence of tumour.

It is an administrative sentence, but it describes the limit of the technique exactly.

Where it simply does not work

In low-grade gliomas fluorescence simply does not appear. The foundational study on tumours without contrast enhancement found focal fluorescence in eight of nine grade III gliomas, and in zero of eight grade II gliomas⁸. Later series found somewhat higher proportions — between 8 and 35 per cent depending on the study and the optics — but always a minority.

One paper quantified the diagnostic yield of looking down the microscope in a low-grade glioma, and the result is devastating: 38 per cent accuracy and an area under the curve of 0.514⁹. An area under the curve of 0.5 is exactly chance. Looking adds no information. Spectroscopic measurement, which detects protoporphyrin below the visible threshold, rises to 67 per cent, and found that 45 per cent of specimens without visible fluorescence contained diagnostically significant levels.

This has a practical consequence that is genuinely useful, and worth understanding properly: in a tumour that appears low-grade, a focus that fluoresces marks, with very high probability, an area of greater aggressiveness¹⁰. The technique does not guide resection, but it guides where to take the specimen so as not to undergrade the tumour. That is a different indication from the one usually attributed to it.

The other dyes

Sodium fluorescein is increasingly used because it is cheap, requires no advance dosing and needs simpler equipment. The best available study, a prospective multicentre phase 2 trial with 46 evaluable patients, reported complete resection in 82.6 per cent and sensitivity and specificity of 80.8 and 79.1 per cent¹¹. There is no phase 3 randomised trial of fluorescein in glioma. The asymmetry of evidence between the two dyes is real and worth naming.

But there is a deeper difference, and it is mechanistic. Fluorescein does not enter tumour cells: it accumulates where the blood-brain barrier is broken. A study of 347 biopsy specimens measured it precisely: in contrast-enhancing lesions sensitivity was 84 per cent; in non-enhancing lesions, 12¹². Put another way, fluorescein is gadolinium made visible in the operating room. It marks exactly what was already visible on the preoperative MRI. That is useful — it helps resect the enhancing part — and it is limited, because the infiltrating margin, which is where glioma recurs, does not enhance.

5-ALA, by contrast, does depend on tumour cell metabolism. That is why it can mark tissue beyond the enhancement. And that is why its limit is a different one: the minimum cell density needed for the human eye to see the glow.

Who signs

It is worth knowing who writes the literature in this field. The principal investigator of the pivotal trial declares consultancy and speaker fees from Carl Zeiss, Leica, Photonamic, Medac and NX Development Corp⁷. The co-author of the paper narrating the US approval declares being a consultant to NX Development Corp — the holder of the authorisation — and receiving royalties from it, in addition to fees from Carl Zeiss and Leica¹³.

None of that invalidates the results. The 2006 trial is a good trial and its figures stand. But a reader deserves to know that the principal advocates of each technique have financial ties to those who manufacture it, because that explains part of the enthusiasm with which it is told and part of the silence about what was not demonstrated.

What to take away

5-ALA fluorescence is one of the few operating-room technologies that did what it promised: it reproducibly increased the proportion of complete resections, and it did so in a large randomised trial. That is more than almost any other surgical adjunct can say.

At the same time its virtue is strictly one-directional. If it glows, it is tumour: that is near-certain. If it does not glow, nothing is known: that is near-certain too. A surgeon who resects until the pink goes out will have resected well what the marker could show, and will have left exactly what the marker cannot show.

Hence fluorescence is never used alone. It is one layer of information among others — imaging, navigation, stimulation, judgement — and its value depends entirely on whoever is looking knowing what they are not seeing.