A department replaces one laser console and keeps the protocol. Eighty joules per centimetre at 10 W, pull back at the same speed, same tumescent recipe, same everything. The first few cases come back with more bruising and a longer recovery than the old machine ever produced, and somebody starts asking whether the new wavelength was a mistake.
It almost never is. What happened is that a number was moved from one machine to another without moving the physics behind it. At 1940 nm the water in the vein wall absorbs about 4.8 times more of the incident light than it does at 1470 nm, and the useful penetration depth drops to roughly a quarter. Deliver the same LEED and you are not reproducing the treatment — you are over-treating a thinner layer of wall.
That single fact reorganises this entire comparison: the wavelength difference is real and measurable, it changes the protocol you should be running, and it barely changes the occlusion rate once the protocol is right. The rest of this piece works through the physics, the trials, and the part almost nobody writes about — what 1940 nm demands from the fiber itself. The fibers we build for both wavelengths are on the radial fiber and bare fiber pages; the clinical picture sits on the vascular surgery hub.
What 4.8× actually means
Water absorption is the only thing that matters in truncal ablation, because at both wavelengths the target is the water in the vein wall rather than haemoglobin in the blood. The published absorption coefficients in water, in the near-infrared:
| Wavelength | Emitting source | μₐ water (cm⁻¹) | Relative absorption vs 1470 nm | Penetration depth in water |
| 980 nm | InGaAs diode | 0.45 | 0.018× | ~3.0 mm |
| 1470 nm | InGaAsP diode | 24.8 | 1.0× | ~220 µm |
| 1910 nm | AlGaIn diode | 90.3 | 3.6× | — |
| 1940 nm | AlGaIn diode / Tm fiber | 119.8 | 4.83× | ~48 µm |
| 2100 nm | Ho:YAG | 26.9 | 1.1× | — |
Those numbers, compiled in the clinical-technical review published in Jornal Vascular Brasileiro and corroborated by independent absorption measurements (27.6 cm⁻¹ at 1470 nm and 128 cm⁻¹ at 1940 nm in a separate ex-vivo study), are the whole physical argument. Two consequences follow, and they point in opposite directions for the surgeon.
The good consequence: energy that is absorbed in a shorter distance produces a higher temperature rise in a smaller volume. Collagen denaturation and necrosis happen between roughly 70 °C and 100 °C; anything above that carbonizes tissue without adding anything useful. At 1940 nm you reach the therapeutic window with less power and less total energy, and you reach the carbonization threshold later for a given energy input. An ex-vivo multi-wavelength comparison on porcine tissue makes the practical gap visible: at 1940 nm a whitish coagulation surface appeared at 5 W, whereas at 1470 nm the same appearance required 10–20 W, with carbonization setting in around 40 W for both.
The bad consequence: 48 µm of penetration depth is shallower than the wall you are trying to kill. A vein wall is hundreds of microns thick, and the treatment has to drive a transmural injury. At 1940 nm the heat has to conduct outward from an extremely thin absorptive layer, and conduction takes time. That is why 1940 nm protocols use slower pull-back speeds and lower energies per centimetre rather than simply “less energy overall” — the geometry of heat deposition changed, so the timing had to change with it.
There is a third number worth keeping in view, because it shows how completely water takes over at the longer wavelength. Attenuation measured in porcine kidney tissue at 1470 nm was 8.0 ± 1.0 mm⁻¹, of which water accounted for 2.8 mm⁻¹ and scattering plus other chromophores for 5.2 mm⁻¹. At 1940 nm, total attenuation was 13.8 ± 1.2 mm⁻¹, of which water accounted for 12.6 and everything else just 1.2 mm⁻¹. At 1470 nm you are still treating a tissue; at 1940 nm you are treating its water content. That is precisely why outcome is so insensitive to the patient’s blood content at 1940 nm — and why the same insensitivity makes the protocol numbers less forgiving.
The protocol does not transfer
Here is where most of the clinical confusion in this comparison originates. LEED — joules per centimetre of treated vein — is a wavelength-specific quantity. Treating it as portable is the single most common mistake made when a department moves from a 1470 nm console to a 1940 nm one.
Published 1940 nm protocols cluster around 40–55 J/cm. Published 1470 nm protocols cluster around 70–90 J/cm. A 1940 nm series with individualised pull-back at 1 mm/s and up to 10 W achieved 100% occlusion at six months and 96.5% at 24 months at a mean of 53 J/cm — against a literature comparison group at 77–82 J/cm on 1470 nm. The fEVLA retrospective of 277 patients found the 1940 nm group needed 42 J/cm versus 62 J/cm for the same segments (P < 0.001). The randomised trial of 206 patients reported lower power and lower LEED in the 1940 nm arm by design.
Now read the failure case in the randomised trial, because it is the most instructive line in the whole literature. One anatomical failure occurred, in the 1470 nm arm. Power was 10.67 W — right on the group mean of 10 W. Vein diameter was 5.5 mm — essentially the group mean of 5.8 mm. The one parameter that stood out was LEED: 46.76 J/cm against a group mean of 89.14 J/cm. The authors’ conclusion is the correct one: at that energy level, at 1470 nm, the delivered dose was simply not enough to raise the wall past collagen necrosis.
That is not a wavelength failure. It is an under-dosing failure that the wavelength made possible, and it is the mirror image of the over-dosing failure you get by running an 85 J/cm protocol on a 1940 nm console. Both errors come from the same source — assuming LEED is transferable.
The rule of thumb: expect the 1940 nm LEED target to land at roughly half the 1470 nm number for the same vein, and confirm it against your own duplex follow-up rather than adopting anyone’s figure, including the one in this paragraph.
What 1940 nm demands from the fiber
This is the part that gets left out of clinical comparisons, and it is where a buyer should spend their diligence.
The fiber has to be low-OH silica. The 1900–2100 nm region sits inside the water absorption band of the glass itself. A high-OH (UV-VIS grade) fiber is effectively unusable above 2 µm, and its loss climbs steeply through the 1900s. Ultra-low-OH silica performs the way you need: published attenuation curves for medical-grade low-OH fiber run around 1 dB/km at 1500–1700 nm, roughly 5–6 dB/km at 1900 nm, 15 dB/km at 2000 nm and 25–50 dB/km at 2100 nm.
Do the arithmetic before you panic about those numbers: 6 dB/km over a 3 m delivery fiber is 0.018 dB — under half a percent of the delivered power. The fiber attenuation is not the practical problem. The practical problem is what the same physics does to everything else in the light path. Any water-based contamination on the end face, in the connector ferrule, or in a poorly dried sterile pack absorbs energy where you cannot see it. At 1470 nm a careless fingerprint is a transmission nuisance; at 1940 nm it is a heating element sitting a few centimetres from your hand. This is the reason a low-OH specification and a clean, dry interface standard belong in the purchase order, not just in the datasheet — and why the optical transmission data for a probe should always be quoted at the wavelength you actually run.
Source technology drives the cost. A 1470 nm system is an InGaAsP semiconductor diode — a mature, cheap, high-power multimode device. A 1940 nm system is an AlGaIn diode or a thulium-doped fiber laser. The thulium route buys you a genuinely better beam: single-mode output, diffraction-limited, coupleable into far smaller cores than a multimode diode stack. That advantage is real for applications using 50 µm probes; for EVLA with a 600 µm radial fiber it changes little, and you pay for it in console cost. This is why the 1940 nm decision is partly a capital-budget decision, and why a department with four 1470 nm consoles in service has a legitimate reason to standardise.
Tip behaviour differs in a way that favours 1940 nm. The observation that 1940 nm cases show less fiber sticking is consistent with the absorption physics: when the light is absorbed within tens of microns of the surface it passes through, less of it is available to be absorbed by a carbonized layer sitting on the fiber face. That is the “hot needle” failure mode discussed in the radial versus bare-tip comparison, and it is at least partly a wavelength story rather than purely a tip-design one. Confirm it with the fiber supplier’s bench data rather than accepting it as a marketing line — the claim sits at the intersection of two variables, exactly where claims are easiest to make and hardest to audit.
“1940 nm compatible” is not a specification. Any fiber transmits 1940 nm for a few milliseconds. What matters is the silica grade, the fused-tip geometry and its integrity at the higher peak irradiance of a tightly absorbed beam, the connector’s thermal margin, and the tested transmission after sterilisation. Ask for the attenuation curve at the operating wavelength, not a claim of compatibility.
Choosing between them
Stay on 1470 nm when: you have a fleet in service and the marginal clinical gain will not repay a console replacement; your case mix is dominated by large truncal diameters where you want the deeper penetration and the greater energy headroom; your protocols and your team’s muscle memory are established; or your follow-up data at 1470 nm is already at 96%+ occlusion and the binding constraint on your practice is something other than recovery time.
Move to 1940 nm when: your patients care most about the week after the procedure — return to work, analgesia, bruising — and you can see that difference in your own audit; you treat superficial segments close to skin or segments running near sensory nerves, where a shallower absorptive zone with lower required energy is an advantage; or you are buying a console anyway and the incremental cost is small next to the recovery difference you can measure.
Whichever way you go, do not run the same protocol on both. The wavelength sets the absorption depth, and the absorption depth sets the energy that makes sense. A number that worked on one console is a hypothesis on the other, and the published failure cases in this literature are under-dosing errors at 1470 nm and over-dosing errors at 1940 nm — the two sides of the same mistake.