The clinical dilemma

Every surgeon who works with atrophic maxilla rehabilitation knows the point where the options run out. Cawood-Howell V or VI, a pneumatized sinus, a knife-edge ridge, a 68-year-old patient who has already lost implants twice and will not accept another iliac crest graft. The classic alternatives — extensive grafting, zygomatic implants, pterygoids — exist and work, but each has its price: treatment time, donor-site morbidity, learning curve, or an anatomy that simply does not cooperate.

It is in this space that the subperiosteal implant has returned to the discussion. Not because someone decided to resurrect a technique from the 1940s, but because cone-beam tomography, computer-aided design and direct titanium sintering solved exactly the problem that killed the original technique: the adaptation of the framework to the bone.

The subperiosteal implant of the 1950s was cast from a bone impression obtained in a first surgery. The fit was approximate, the framework was rigid over remodeling bone, and progressive exposure was almost inevitable. Today's device is planned on the patient's DICOM, printed in Ti6Al4V, and placed in a single intervention. It is the same idea; it is not the same device.

The question, then, is what the literature published to date allows us to state — and what it does not yet allow.

What the short-term data show

There is remarkable convergence among the systematic reviews for follow-ups of up to three years. Anitua and colleagues, in a review of 13 articles with 227 implants, recorded 97.8% in function after a mean of 21.4 months [1]. Al-Nawas and colleagues, analyzing 24 articles and 246 implants, found 97.6% functional at a mean of 17 months [2]. The most recent meta-analysis, by Cosola and colleagues, arrived at the same 97.8% for studies with follow-up of up to three years [3].

On the bone side, the findings are better than intuition would suggest. Van den Borre and colleagues superimposed pre- and postoperative CT scans of 15 patients and measured mean remodeling of 0.26 mm at the crest and 0.088 mm in the supporting bone over twelve months [4]. The authors' reading is biomechanical: the reduction of stress shielding in a calibrated framework does not induce radiographically significant crestal atrophy. Vaira and colleagues, in sectional rehabilitations followed for up to five years, measured 0.18 mm of resorption under the abutments in the first year [5].

On the patient side, the change is striking. The OHIP-14 score dropped from 17.20 to 5.80 in twelve months in a prospective multicenter cohort [6], and the 40-patient series followed for about 2.5 years maintained high satisfaction [7].

Compared directly with the zygomatic implant in a cohort of 150 patients over five years, the subperiosteal implant showed statistically equivalent survival (97.1% versus 96.3%; p = 0.278), with less peri-implantitis and without the sinus complications that occurred in 12.4% of the zygomatic group [8]. There is also a relevant practical difference: the subperiosteal implant allows reimplantation after failure; the zygomatic does not.

Where the curve changes

Here the literature becomes uncomfortable, and this is the part any surgeon needs to read before indicating the technique.

The same meta-analysis that records 97.8% up to three years points to an overall pooled survival of 92.4%, with substantial heterogeneity — and notes that the study with the longest published follow-up, at six years, records a drop to 54.1% [3]. That study is by Onică and colleagues: 36 patients, 61 implants, and only nine cases considered successful at the end of the period [9]. The remaining 27 accumulated exposure of the metal framework, early mobility in the first four to six months, or late mobility due to recurrent infection and progressive exposure.

It is a single study, and a single study does not define a technique. But nothing published in that time range contradicts it. To claim long-term safety today is to claim beyond the available data.

This distinction between survival and success is not semantic. Vatteroni and colleagues applied explicit criteria of stability, inflammation and deviation from planning to a cohort of 18 patients followed for two years: cumulative survival was 100%, and the success rate, 73.1% [10]. The same group of patients, two numbers that differ by 27 percentage points. When a marketing piece cites "97% survival," it is answering the wrong question.

The complication that dominates — and it is not the titanium

In none of the systematic reviews does structural failure of the device appear as the main cause of failure. The problem is always the same, and it is soft tissue.

• Partial exposure in 25.6% of implants in the Anitua review [1]
• Partial exposure in 37% in the Al-Nawas review [2]
• Soft tissue dehiscence in 28.6% in the meta-analysis dedicated to the topic, with 368 implants [11]
• Framework exposure in 65% of patients in the cohort that examined soft tissues in depth [12]

That last number deserves attention because it comes from the same group that published the favorable satisfaction and bone remodeling results. It is not a study by detractors — it is the same protocol, viewed through the lens of the periodontium. And it is precisely this work that identifies the risk factors with statistical significance: thin biotype and presence of mucositis (p < 0.05), and smoking with an approximately seven-fold higher risk of recession [12].

The dehiscence meta-analysis adds a data point that changes planning: the maxilla carries a 3.8-times higher risk of failure than the mandible (HR 3.843; 95% CI 1.157–12.723; p = 0.028) [11]. The authors themselves caution that part of this may reflect the severity of the cases referred to each site, and not the site itself. Even so, the difference is significant and consistent with the rest of the literature: mandibular series are systematically cleaner — Vaira and colleagues report 30 implants in the posterior mandible without a single exposure, infection or loss over 22.5 months [13].

The variables under the surgeon's control

If failure is a matter of soft tissue and not of the device, then the outcome depends on decisions made before and during surgery. The literature allows four to be identified with a measured effect.

Number of surgical interventions

This is the strongest finding of the entire review. Sánchez-Labrador and colleagues analyzed 14 studies with 958 patients and 973 implants: 100% survival when placement occurred in a single intervention, versus 85% when there were two surgical stages [14]. The digital workflow that eliminates the prior bone impression is not a convenience — it is the variable that separates the two outcomes.

Extent of the rehabilitation

Fixed partial rehabilitations show a substantially lower complication profile: 4.13% biological complications across 121 implants, versus the 25% to 37% rates observed in predominantly full-arch series [15]. The larger the framework, the larger the surface at risk of exposure.

Patient selection

Thin biotype, active mucositis and smoking are the factors identified with significance [12]. None is surprising to anyone who works with the periodontium — what changes is that here there is no bone interposed between the framework and the soft tissue covering it.

Accuracy of fit

Where fit fails, the technique fails. In the ten-year follow-up published by Onică and colleagues, the two implants removed over the period were removed for inadequate fit and recurrent infection [16]. Ayhan and colleagues recorded implant-to-bone fit problems in 11 of 31 patients [17]. Vatteroni measured a mean horizontal deviation of 1.46 mm between planned and achieved — three times the vertical deviation [10].

What the evidence does not yet offer

There is no randomized clinical trial on custom subperiosteal implants. A search on ClinicalTrials.gov returns three records: two case series on digital workflow accuracy and one numerical study with a single patient. No prospective comparative study is underway.

The entire available base is retrospective — cohorts and case series, with methodological heterogeneity that prevented quantitative meta-analysis in at least two of the reviews [1][2]. The consensus report published in 2024, following an expert meeting the previous year, found only six articles meeting minimum inclusion criteria and recorded a range of biological complications between 5.7% and 43.8% [18]. The authors' conclusion is the most honest formulation of the current state:

Where the indication holds

Taking the body of evidence together, the literature outlines a narrow, defensible indication:

• Severe atrophy classified as Cawood-Howell IV to VI
• A patient in whom extensive grafting and zygomatic implants are contraindicated, have been refused, or have already failed
• A digital protocol with placement in a single surgical intervention
• Adequate soft tissue — thick biotype, absence of active mucositis
• A non-smoking patient, or with documented cessation
• Explicit awareness, recorded in the informed consent, that long-term data beyond three years are limited to a single published study

Cosola and colleagues call this a "salvage option for carefully selected patients" [3]. The expression may sound restrictive, but it accurately describes the patient who actually arrives at the office with this need — the one for whom the conventional routes have already been exhausted. The available evidence does not invalidate the technique; it defines its boundary, and demands that the indication be made within it.

Summary of studies

FEA: finite element analysis. Biomechanical studies describe behavior under simulated load and do not replace clinical evidence. Where the original design is a case series or report, this is indicated — the hierarchy of evidence should be considered when reading each row.