Dental Implants and Diabetes: What the Studies Show About Survival, Healing and Long-Term Risk

A patient with diabetes who asks whether they can have implants is asking a question that the literature answers in three separate pieces, and the reason the internet gives them contradictory answers is that most sources collapse the three into one. The first question is whether the implant will integrate and survive. The second is whether it will integrate as quickly and as reliably as it would in a patient without diabetes. The third is whether, years later, it will be more prone to the infection of the surrounding bone called peri-implantitis. The evidence on those three questions is respectively reassuring, cautionary, and concerning, and a good consultation says all three.

Naujokat's 2016 systematic review is a fair starting summary. It identified 22 clinical studies and 20 reviews, heterogeneous in method, and concluded that patients with poorly controlled diabetes suffer impaired osseointegration, an elevated risk of peri-implantitis and a higher level of implant failure, while patients whose diabetes is well controlled can have implants with a complication rate similar to healthy patients. The influence of how long someone has had the disease was, in their assessment, not fully clear. Javed and Romanos reached the same shape of conclusion in 2009 from eighteen studies including animal models: under good metabolic control, osseointegration can be accomplished; under poor control, bone-to-implant contact appears to fall over time.

Those are review-level conclusions. The meta-analyses that count actual failures are more surprising, and it is worth walking through them, because they are the reason I no longer treat a diabetes diagnosis as a reason to say no.

Chrcanovic, Albrektsson and Wennerberg pooled fourteen publications in 2014. Implant failure rates did not differ significantly between diabetic and non-diabetic patients — a risk ratio of 1.07, with a confidence interval of 0.80 to 1.44 and a p-value of 0.65. What did differ was marginal bone loss, by a mean of 0.20 mm in favor of non-diabetic patients, a difference that was statistically significant and clinically small. The authors were careful to say that the studies were confounded and that larger trials reporting both groups separately were lacking.

Moraschini's 2016 meta-analysis of fourteen studies published between 2000 and 2015, all rated high quality on their risk-of-bias tool, reached the same conclusion: the number of implant failures did not differ between diabetic and non-diabetic subjects, nor between type 1 and type 2 diabetes, while marginal bone loss again favored non-diabetics. Shi's 2016 meta-analysis in the Journal of the American Dental Association asked the narrower question of whether poorly controlled diabetes fails more often than well-controlled diabetes, pooling seven studies, 252 patients and 587 implants, and could not show a difference either — a pooled relative risk of 0.62 with a confidence interval so wide, 0.23 to 1.71, that it demonstrates mainly how few patients have been studied.

The one review that found a trend is Annibali's 2016 survival analysis of seven studies and 1,142 implants in diabetic patients. Cumulative survival was 96% before loading, 93% at one year and 91% at the end of follow-up, and the hazard of failure — 4% during the osseointegration period and 3% during the first year of loading — was concentrated early and then held constant for six years. That is consistent with the biology: if diabetes affects anything, it affects the healing phase. Once an implant has integrated in a diabetic patient, the survival curve looks like everyone else's.

The work that turned a vague concern into a measurement came from Oates' group in San Antonio. Their 2009 study followed 42 implants in ten non-diabetic volunteers and twenty patients with type 2 diabetes whose glycated hemoglobin ranged from 4.7% to 12.6%, using resonance frequency analysis to track stability over four months. Every implant showed its minimum stability between two and six weeks after placement — the dip I describe in the osseointegration review. Patients with an HbA1c of 8.1% or higher had a greater maximum fall in stability from baseline and took longer for stability to return to its starting level. The impairment was in direct relation to the level of hyperglycemia.

The 2014 follow-up, in the Journal of the American Dental Association, is the largest prospective study on the question. One hundred and seventeen edentulous patients each received two lower-jaw implants, 234 in total, loaded after four months and followed for a further year, with baseline HbA1c up to 11.1% and readings as high as 13.3% during the study. Among the 110 patients followed for the full year, implant survival was 99.0% in patients without diabetes, 98.9% in those with well-controlled diabetes and 100% in those with poorly controlled diabetes. Counting the seven patients lost to follow-up as failures gave conservative estimates of 93.0%, 92.6% and 95.0%, with no significant difference. But delays in the implants becoming stable were, again, directly related to poor glycemic control.

Eskow and Oates then took the hardest group on its own. Twenty-four adults with an HbA1c between 8.0% and 12.0% received two or more implants; survival was 98.6% at one year and 96.6% at two years, complications — mostly mucositis — occurred in 29% of participants, and the number of complications did not correlate with the HbA1c level. The authors concluded that implant therapy can be applied more broadly in type 2 diabetes with poor control than the older contraindication implied. I read those three studies together as a single finding: high blood sugar does not stop an implant integrating, it slows it, and a protocol that gives it more time is the correct response rather than a refusal.

The concerning piece is the third. Monje, Catena and Borgnakke's 2017 systematic review, which used a strict case definition of peri-implantitis to reduce bias, found the risk of peri-implantitis about 50% higher in patients with diabetes than without — a relative risk of 1.46 with a confidence interval of 1.21 to 1.77, and an odds ratio of 1.89 — and found that association to be independent of smoking. Diabetes was not associated with the milder condition, mucositis. The authors asked for great caution in interpreting the finding, and the World Workshop review by Schwarz a year later graded the diabetes evidence as inconclusive, because the studies remain confounded. Both statements are true at once: the best available estimate says the risk is higher, and the certainty behind that estimate is moderate.

That is also the finding that best fits what the biology predicts. Hyperglycemia impairs neutrophil function, alters collagen turnover and amplifies the inflammatory response to bacteria; those are the mechanisms behind the well-established link between diabetes and periodontitis, and there is no reason the tissue around an implant would be exempt. The slightly greater marginal bone loss that both Chrcanovic and Moraschini measured is probably the same process seen in a different unit. Survival tells you whether the implant is still there; bone loss and peri-implantitis tell you how well it is doing, and it is on the second measure that diabetes shows.

The practical consequence is that a diabetic patient's implant treatment is weighted toward the years after surgery rather than the day of it. The peri-implantitis review sets out what maintenance actually achieves, and for a patient with diabetes the recall interval sits at the short end of the range it describes.

The first thing I ask for is not a scan but a number: the most recent HbA1c, and the name of the physician who follows it. There is no threshold in the literature above which implants are forbidden — Oates loaded implants successfully in patients with readings above 11% — but the Oates data are unambiguous that the higher the reading, the slower the integration, and I plan the timeline accordingly. A patient with an HbA1c comfortably under 7% is treated on the same schedule as anyone else. Above 8%, I lengthen the healing period before loading, I am more conservative about placing a provisional on the day of surgery, and I talk with the patient about whether improving control before surgery is realistic; sometimes it is, and it helps everything else too. A reading that is very high and not being managed is a reason to involve the physician before the mouth, not a reason to refuse — the medical-clearance page explains how that conversation works.

Peri-operatively, Naujokat's review notes that supportive antibiotics and chlorhexidine appear to improve implant success in diabetic patients, and Javed and Romanos make the same point about antiseptic rinses and hygiene. This is one of the situations in which I do prescribe prophylactic antibiotics despite the general trend away from them in healthy patients, because the host response is the variable that has changed. Smoking on top of diabetes is a combination I ask patients to break before surgery rather than after, since each independently raises the risk of the same complication.

Afterwards, the plan is simply the peri-implantitis plan applied with more discipline: a baseline radiograph and probing depths at delivery, screw-retained restorations that leave no cement, a recall interval of three to four months rather than six, and an agreement that a rising HbA1c is something the dental team wants to know about. None of that is exceptional medicine. It is the ordinary maintenance every implant patient should have, done consistently because the margin for skipping it is smaller. Related reading: what your medical history changes and when implants are genuinely not the answer.

The largest weakness is size. Shi's meta-analysis pooled 587 implants; Oates' prospective study, the biggest of its kind, had 234. Confidence intervals that wide cannot rule out a real difference in failure between good and poor control, and the field needs a prospective cohort in the thousands, stratified by HbA1c, with survival, bone loss and peri-implantitis all recorded to the same definitions. Almost every study has been in type 2 diabetes; type 1, which typically begins younger and runs longer, is poorly represented, and Naujokat's observation that the effect of disease duration is unclear remains true a decade later.

The newer questions are pharmacological. The glucagon-like peptide-1 agonists and sodium-glucose cotransporter-2 inhibitors that now dominate diabetes treatment change body weight, bone metabolism and inflammation, and their effect on osseointegration has not been studied in humans. And the mechanism behind the peri-implantitis association — whether it is the hyperglycemia itself, the altered immune response, or the periodontitis history that so often accompanies both — would tell us which patients to watch most closely. Until it is settled, I watch all of them, and I tell them why.