Peri-implantitis: if the foundational model is incomplete, how long can we rely on it?

A model can be useful and still be structurally wrong.

The dominant initiation story for peri-implantitis is a plaque story. It has evidence. It has teaching slides. It has products. It also has a quiet omission: the implant is not a tooth, the interface is not a PDL, and the crest is not reading the same bending moment.

I am a periodontist. I will not be the person who pretends biofilm is a conspiracy. I will be the person who asks what the biofilm colonised, and whether we opened the niche ourselves with a stiffness mismatch and a vector we delivered in ceramic.

Consider the substitution we make without noticing. Periodontitis has a ligament. The ligament is a damper, a proprioceptive organ, a vascular bed, and a fibre network that inserts into cementum. An implant has a bone–implant interface we specified in a catalogue. When the crest around that interface saucerizes in the first years, the plaque-only model has one move: hygiene. Sometimes that move is correct. Sometimes it is a category error.

Mechanics does not need to 'win' this argument. It needs to be allowed into the first paragraph. Micromotion, contact, abutment geometry, an oblique load from a cantilevered restoration, a thin buccal plate — these are initiation candidates. Inflammation then changes the material we assigned to 'bone' last week. A complete model is coupled. An incomplete model picks a favourite field and calls the rest 'contributing factors'.

How long can a field rely on an incomplete foundational model?

Longer than it should. Dentistry is practical. Incomplete models that produce a protocol will outlive complete models that produce a differential equation. The cost is paid in patients who clean meticulously around an implant that is being bent, and in research programmes that keep adding mouthwash arms to a question that was never only chemical.

If we are going to build virtual patients that include implants — and we are; the same CBCT that segments a defect will segment a fixture — then the constitutive story cannot be 'tooth, but grey'. No PDL. Different damping. Different perfusion architecture in the mucosa. A crest that concentrates bending. Initiation research that wants to be more than a brand war has to start there, and then invite the microbiome back as a resident of a mechanical niche.

I do not have a trial percentage to sell you. I have a habit I would like the field to pick up: when the foundational model cannot explain a clean failure, stop adding adjuncts. Repair the model.