Attrition and implant research

Appraisal of “Clinical and Radiographic Outcomes of Dental Implant Therapy”*  by Mark-Steven Howe

*(J Derks et al. 2015; Jan Derks et al. 2015; Derks, J. Schaller, D. Håkansson, J. Wennström, J.L. Tomasi, C. and Berglundh 2015)

implant astra


The initial study protocol(Jan Derks et al. 2015) was created to evaluate patient-reported outcomes following implant-supported restorative therapy in a randomly selected patient sample taken from the Swedish Social Insurance Agency. Having selected the data set the research evolved into a further study on peri-implantitis and implant survival. This was registered with (NCT01825772) in 2013.

The research was broken into two papers using the same data set:

  1. The primary research question was to report on the prevalence, extent, and severity of peri-implantitis in a large and randomly selected patient sample identified from the data register of the Swedish Social Insurance Agency.(Derks, J. Schaller, D. Håkansson, J. Wennström, J.L. Tomasi, C. and Berglundh 2015)
  2. The secondary research question was to report on the loss of dental implants assessed in a large and randomly selected patient sample. (J Derks et al. 2015)


A sample of 4716 patients aged between 45 and 75 who had received implants in 2003 were randomly selected from the data registered with the Swedish Social Insurance Agency. From the 3827 respondents, patient files were collected on 2765 for analysis. From this group 900 randomly selected patient were invited to a clinical examination. Nine years following implant placement 596 of the 900 were examined by two periodontists, the clinical examination also included radiographs.

The patients were divided into two groups by age (45-54 and 65-74). Two examiners extracted from the files basic patient data that included information on diabetes, cardiovascular diseases, periodontitis at the time of implant therapy, smoking and recall frequency. Clinical evaluation included data on private or public dental clinical setting, general practitioner or registered specialist for the Swedish National Board of Health and Welfare. Additional data was collected on whether the operators were specialists or general practitioners. Implants were categorized according to brand, length (

A multiple logistic regression model was used to analyse the data.



  “In 98 (23.0%) of the 427 patients with baseline radiographs, no signs of peri-implant disease were detected. In addition,137 (32.0%) patients exhibited only peri-implant mucositis, while 192 (45.0%) presented with peri-implantitis. Moderate/severe peri-implantitis was observed in 62 (14.5%) patients.”

Peri-implantitis effecting implants by brand

Implant brand Odds Ratio 95% Confidence interval
Straumann Dental Implant System 1
Brånemark System Replace Select 3.77 1.60 to 8.87 0.002
Astra Tech 3.55 1.29 to 9.77 0.014
Remaining implants 5.56 1.70 to 18.24 0.005

Implant Survival:

Early Loss Late Loss Total Loss @ 8.9 years
Patients affected 4.4% 4.2% 7.6%
Implants Lost 1.4% 2.0% 3.0%
Implant brand Odds Ratio Odds Ratio
Straumann Dental Implant System 1 1
Brånemark System Replace Select 1.94 6.13
Astra Tech 2.10 5.23
Remaining implants 7.79 58.15

 Authors conclusions

Patients with periodontitis and with ≥4 implants, as well as implants of certain brands and prosthetic therapy delivered by general practitioners, exhibited higher odds ratios for moderate/severe peri-implantitis. Similarly, higher odds ratios were identified for implants installed in the mandible and with crown restoration margins positioned ≤1.5 mm from the crestal bone at baseline. It is suggested that peri-implantitis is a common condition and that several patient and implant-related factors influence the risk for moderate/severe peri-implantitis

In terms of implant loss, the present study reported on outcomes in implant dentistry assessed in a large and randomly selected patient sample representing effectiveness of the treatment procedures. Almost 8% of patients had lost ≥1 implants, and several patient and implant-related factors influencing early and late occurring loss were detected.


This was a well-conducted long-term retrospective cross-sectional study of a large randomly selected group of Swedish patients treated by both specialists and general practitioners. It is interesting to note it was registered on when it is not technically a clinical trial. The registered protocol inclusion criteria was for 55 – 85 year olds but appears to have changed later to 45-54 and 65-74.

The only major problem with the results was the effect of attrition bias (Schulz & Grimes 2002). After initial randomisation 41% of the data was not available/excluded and from this reduced data set a further 900 samples were randomly selected of which 34% of the data was not collected. Since this data had been randomised this loss of data will most likely bias the results and affect the internal validity of the study, as it is unlikely to be all due to random causes. For example, patients who have had poorer outcomes or are less motivated may exclude themselves from the consent process or clinical examination phase. Though considerably more time consuming a clearer picture of the prevalence of peri-implantitis and survival rates could have been achieved if an ‘intention-to-treat principal’ had been followed. That is taking the whole randomised populations data into account when producing the results either via sensitivity analysis or imputation of the missing data.By taking out the data relating to patients who died, are sick or moved house (assume increased failures among the non-attenders) one can impute the total loss figure to be somewhere between 2.o% and 11.o%. Research has been carried out on analysing attrition bias and drop-out rate (Heneghan et al. 2007) to identify different group characteristics from those who responded to the authors questionnaires and clinical examination.

A couple of additional observations where the tight criteria for moderate/severe peri-implantitis (Bleeding on probing/suppuration and bone loss greater than 2mm) considering the 9-year time frame and pooling the moderate and severe results together. This may have been done to increase sample size. Since the majority of the implants were 10mm long and the mean bone loss on the implants with peri-implantitis was 29%, >3mm of bone loss would have given better resolution for clinically significant peri-implantitis. It is also a pity the results were not collected at 10-years to fall into place with other long-term survival/success studies.

In conclusion, the results of this well conducted study should be interpreted with caution due to the high attrition rates and therefore, even though the differences in odds ratios for infection/implant survival rates between the various brands are most likely accurate the results may be over-optimistic when compared to some systematic reviews (Srinivasan et al. 2016)


Derks, J. Schaller, D. Håkansson, J. Wennström, J.L. Tomasi, C. and Berglundh, J., 2015. Effectiveness of Implant Therapy Analyzed in a Swedish Population:Prevalence of Peri-implantitis. Journal of Dental Research, 94(3), pp.44–51.

Derks, J. et al., 2015. Effectiveness of Implant Therapy Analyzed in a Swedish Population : Early and Late Implant Loss. Journal of Dental Research, 94(3), pp.44–51.

Derks, J. et al., 2015. Patient-reported outcomes of dental implant therapy in a large randomly selected sample. Clinical Oral Implants Research, 26(5), pp.586–591.

Heneghan, C. et al., 2007. Assessing differential attrition in clinical trials : self-monitoring of oral anticoagulation and type II diabetes. , 12.

Schulz, K.F. & Grimes, D.A., 2002. Epidemiology series Sample size slippages in randomised trials : exclusions and the lost and wayward. , 359, pp.781–785.

Srinivasan, M. et al., 2016. Dental implants in the elderly population: a systematic review and meta-analysis. Clinical oral implants research. Available at:


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s