HIFU body contouring: there is a sham-controlled trial — but the answer changes with which analysis you read

A 2011 randomised sham-controlled trial in Plast Reconstr Surg. Its title contains the words 'sham-controlled'.

First, who entered. Adults aged 18 to 65 with subcutaneous abdominal fat 2.5 cm thick or more who met screening criteria. The paper presents the device as an approach for non-obese patients.

They were randomised into three groups, with treatment of the anterior abdomen and flanks at three passes each.

• 47 J/cm (141 J/cm total)
• 59 J/cm (177 J/cm total)
• 0 J/cm — no energy applied, sham control

The primary endpoint was the change from baseline in waist circumference at the iliac crest level at posttreatment week 12. Subjective assessments included the Global Aesthetic Improvement Scale and a patient satisfaction questionnaire; safety assessments were adverse events, laboratory values and physical examinations.

The results. Here are the two analyses side by side.

In the intent-to-treat population:

• 59 J/cm — significant versus sham (−2.44 cm, p = 0.01)
• 47 J/cm — not significant (−2.06 cm, p = 0.13)

In the per-protocol population:

• 59 J/cm — significant (−2.52 cm, p = 0.002)
• 47 J/cm — significant (−2.10 cm, p = 0.04)

And investigator subjective measures of global aesthetic improvement and patient satisfaction also favoured each active treatment versus sham. Adverse events were mild to moderate discomfort, bruising and edema, and laboratory values and physical examinations were unremarkable.

Conclusion: treatment with this high-intensity focused ultrasound device reduced waist circumference and was generally well tolerated. Reduction in waist circumference was statistically significant with both active treatments (per protocol).

Now how to read it. This is the most important part of the entry.

First, look at the parenthesis in the conclusion. It says '(per protocol)'. That is, when the paper says both active treatments were significant, that rests on the per-protocol analysis. The paper did not hide it — it put it in brackets — and this site does not delete it either.

Second, why the two analyses differ.

• intent-to-treat counts everyone who was randomised. People who dropped out or could not follow the plan are still counted in the group they were assigned to. It is more conservative.
• per protocol counts only those who finished as planned. But the reason someone could not follow the plan may be related to the treatment — if they dropped out because it hurt, removing them tilts the result favourably.

So this site reads intent-to-treat first. On that basis, the lower energy (47 J/cm) was not significantly different from sham (p = 0.13).

Third, so here is the sentence this entry can write. At the higher energy (59 J/cm) it was significant on either analysis. At the lower energy, the answer changes with the analysis. This site does not put those two in one sentence.

Fourth, look at the size of the number. −2.44 cm. That is the difference in waist circumference against sham. Whether that changes how clothes fit will be felt differently by different people, and this site does not make that judgement. But that the measure is in centimetres is good — it is not a score assigned by eye. So the endpoint is marked 'instrument'.

Fifth, the abstract does not give the total number of participants. The group sizes are absent. We do not invent them. That is why the participant field on the badge above is empty. Without the numbers, the weight of −2.44 cm is hard to read.

Sixth, the funding. The author list includes Medicis Technologies Corporation. This site marks that as industry funding. With one trial this entry could not be 'strong' anyway, but the funding is recorded alongside.

Seventh, 12 weeks and a single treatment. Whether it lasts longer, or changes with repeated sessions, is not something this trial looked at.

Eighth, it is 2011. And devices and settings differ between manufacturers.

Ninth, bruising is among the adverse events. That is why this site links this procedure from the bruising page.

So it sits at 'moderate'. Good conditions — a sham control and an endpoint in centimetres — but one trial, no participant count in the abstract, industry funding, and at the lower energy an answer that changes with the analysis.