Skip to main content

Table 4 Risk of bias for included cohort studies*

From: Association of tooth loss with morbidity and mortality by diabetes status in older adults: a systematic review

First Author, Year, Country

Participant Selection,

Sample Size,

Response Rate

Measurement

Confounding

Statistical Significance Criterion,

Confidence Interval

External Validity/Applicability

Suzuki [42]

2021

Japan

• National Database of Health Insurance Claims and Specific Health Checkups.

• Very large sample size.

Predictor:

Data extracted for those with periodontitis dental claims which may overestimate TL.

Dental linkage performed by hash value from insurer’s ID (ID1), which tends to underestimate TL.

Number of teeth included 3rd molars. Not included edentulous. Claims data from multiple dental providers may introduce non-differential misclassification bias for number of teeth.

Outcome:

DM status extracted from medical and pharmacy insurance claims for outpatient services using ICD-10 codes. Data linkage performed by ID1 tends to overestimate patients with DM.

Medical expenditure determined by the sum of medical and pharmacy expenditure stored in the national database, recorded in Japanese yen.

Included:

age and sex.

Not included:

daily oral hygiene, diet, dental care utilization, SES comorbidities, duration of DM, smoking, alcohol use, prescription use of DM medication, cholesterol reducing drugs and antihypertensives, blood pressure, height, weight, waist and hip circumference, cholesterol panel, non-fasting glucose, hs-CRP.

Descriptive statistics.

Public Health Expenditure presented as mean and median values with 25th and 75th percentiles.

• National database of Japanese older adults.

Ruokonen, [43] 2017, Finland

• Convenience sample from urban university hospital of patients with CKD at pre-dialysis stage.

• No sample size calculation or response rate included

Predictor:

WHO criteria for oral health status, no information for examiner training or calibration.

Outcome: national death registry information for 62 of 65 deaths.

Included: demographics, smoking, number of medications.

Not included:

daily oral hygiene, diet, dental care utilization, alcohol use, duration of DM, comorbidities.

P = 0.10 for entry in multivariable regression model and P = 0.05 for removal.

HR (95%CI)

• Convenience sample from urban university hospital

• Small sample size.

• 5% loss-to-follow-up over 13 years.

• Finnish study participants may be less diverse than U.S. population.

Demmer, [44] 2008,

U.S.A.

• National probability sample of non-institutionalized adults.

• Large sample size.

Predictor: examiners trained for Periodontal Index and DMFT, no kappa scores included.

Outcome: incident DM identified by death certificate, self-report, pharmacological treatment, hospital stay with discharge diagnosis of DM.

Included: demographics, SES, BMI, skin-fold thickness, physical activity, cholesterol, blood pressure, smoking, diet.

Not included:

daily oral hygiene, dental care utilization, alcohol use, duration of DM, prescription use of DM medication, comorbidities.

p ≤ 0.05

OR (95% CI)

Multivariable adjustment for potential confounders.

• National probability sample of U.S. non-institutionalized older adults.

• 6% loss-to-follow-up over 17 ± 4 years.

HÃ¥heim, [45]

2017,

Norway

• National population-based sample of men only.

• Large sample size.

Predictor:

oral health measures extracted from database. No information on specific indices, examiner training, or calibration.

Number teeth included 3rd molars. DM status extracted from database, no criteria for diagnosis included.

Outcome: national death registry.

Included: demographics, smoking, alcohol use, prescription use of DM medication, cholesterol reducing drugs and antihypertensives, blood pressure, height, weight, waist and hip circumference, cholesterol panel, non-fasting glucose, hs-CRP.

Not included:

daily oral hygiene, diet, dental care utilization, comorbidities, duration of DM.

p < 0.05

HR (95%CI)

Multivariable adjustment for potential confounders.

• Large, national population-based study limited to men only.

• 35% loss to follow-up over 12.5 years.

• Norwegian population of men only may be less diverse than U.S. population.

Liljestrand, [24]

2015,

Finland

• National population-based sample.

• Large sample size.

Predictor:

nurses trained to count teeth but not to distinguish between natural teeth, implants, and pontics therefore bias towards null. No kappa scores. Number teeth included 3rd molars. Survey used WHO’s MONICA protocol.

Outcome: incident DM determined via national registries of drug reimbursement and hospital discharge.

Included: demographics, geographic location, BMI, blood pressure, antihypertensives, physical inactivity, parent with DM, smoking, hs-CRP, diet.

Not included:

daily oral hygiene, dental care utilization, prescription use of DM medication, alcohol use, periodontal disease status at baseline, duration of DM, comorbidities.

p ≤ 0.05

HR (95%CI)

Multivariable adjustment for potential confounders.

• Large, national population-based sample.

• Loss to follow-up not described.

• Finnish population may be less diverse than U.S. population.

  1. Abbreviations: CKD (Chronic kidney disease), DMFT (Decayed, Missing, Filled Teeth), DM (Diabetes mellitus), OR (Odds ratio), HR (Hazard ratio), CI (Confidence interval), BMI (Body mass index)
  2. *The criteria of the table extracted from Critical Appraisal Skills Programme (CASP) Checklist for Cohort Studies and Center for Evidence-Based Medicine (CEBM) Critical Appraisal of a Cross-Sectional Study (Survey)