From: Multivariable prediction models of caries increment: a systematic review and critical appraisal
First author, Year of publication [reference] | Predictors in final model Number (N) Predictor levels: • Societal structural • Life-style situational • Physiological • Oral biological • Tooth • Caries experience • Other predictors | Model performance (according to included study) Classification measures: - Sensitivity (Sens) - Specificity (Spec) - Positive predictive value (PPV) - Negative predictive value (NPV) - Positive likelihood ratio (LR +) - Negative likelihood ratio (LR −) Discrimination: - Area under curve (AUC) | Positive likelihood ratio (LR +) Negative likelihood ratio (LR −) calculated by review authors based on sensitivity and specificity presented in included study (95% confidence intervals) | Model interpretation by study authors | Comments |
---|---|---|---|---|---|
Studies of model development | |||||
Coronal caries | |||||
Angulo, 1995 [13] | N = 3 • Oral biological (˃ 104 ms/ml saliva, ˃ 104 lbc/ml saliva) • Caries experience (DS ˃ 3 cavity) | Sens 0.19, Spec 0.86, PPV 0.36, NPV 0.71 | LR + 1.36 (0.43–4.31) LR − 0.94 (0.74–1.20) | High ms counts + high lbc counts + Caries experience presented higher specificity but lower sensitivity than Caries experience alone | Sample with low socioeconomic status. Caries prevalence at baseline NR Caries increment (1.5 years) 45% |
Demers, 1992 [14] | Model 1 (full model) N = 5 • Societal structural (parents’ education) • Oral biological (≥ 105 ms/ml saliva, ≥ 105 lbc/ml saliva debris index-screening level ≥ 1.9) • Caries experience (dmfs ˃ 0 cavity) Model 2 (Socioeconomic model) N = 3 • Societal structural (parents’ education, family structure) • Life-style situational (fluoride supplements) | Model 1 (full model) Sens 0.94, Spec 0.41 Model 2 (socioeconomic model) Sens 0.86, Spec 0.35 | Model 1 (full model) LR + 1.60 (1.40–1.83) LR − 0.14 (0.07–0.28) Model 2 (socioeconomic model) LR + 1.31 (1.15–1.50) LR − 0.40 (0.26–0.64) | Caries experience reached sensitivity and specificity values close to the best model over 1 year. More predictors add to costs and complexity of screening | Caries prevalence at baseline NR Caries increment (1 years) 47% |
Disney, 1992 [15] | Aiken N = 23; Portland N = 22 • Societal structural (referral caries score) • Life-style situational (between-meal snacks, Tooth-brushing–2 predictors, F-rinse) • Physiological (age, race) • Oral biological (ms in saliva, lbc in saliva, plaque score) • Tooth (tooth morphology, fluorosis, sound permanent surfaces, sealants)• Caries experience (dmfs, DMFS, white spot lesions) • Other predictors (examiner agreement–4 predictors, predicted caries score, fluorosis x white spot lesions) | Aiken age 6.6 years Sens 0.59, Spec 0.83 Portland age 6.9 years Sens 0.59, Spec 0.84 Aiken age 10.7 years Sens 0.62, Spec 0.81 Portland age 10.8 years Sens 0.62, Spec 0.84 | Aiken age 6.6 years LR + 3.47 (2.89–4.16) LR − 0.49 (0.42–0.58) Portland age 6.9 years LR + 3.69 (3.05–4.45) LR − 0.49 (0.42–0.57) Aiken age 10.7 years LR + 3.26 (2.73–3.91) LR − 0.47 (0.40–0.55) Portland age 10.8 years LR + 3.88 (3.19–4.71) LR − 0.45 (0.38–0.54) | Models will be useful for predicting those at low and high risk of caries increment. Clinical predictors were the most important group ms, lbc, sociodemographic, and dental behavior data contributed little over a 3-year follow-up | Caries prevalence at baseline NR Caries increment (3 years): Aiken 6.6 years 58% 10.7 years 66% Portland 6.9 years 31% 10.8 years 46% Thresholds for ms, lbc, and some other predictors NR |
Fontana, 2011 [16] | N = 3–7 in 11 models Predictor common for all models except for model 11 • Societal structural (caregiver (CG) does not consider child’s oral health to be very good) 1 year follow-up ICDAS ≥ 3 Models 1, 2, 3 Model 1 N = 4 • Life-style situational (time elapsed since last dental visit) • Caries experience (tooth extracted, tooth restored) Model 2 N = 5 Model 1 + dmfs/DMFS ICDAS ≥ 3 at end Model 3 N = 7 Model 1 + dmfs/DMFS ≥ 3 at start and • Societal structural: (CG received a referral note for child) • Life-style situational (soda drinks, soda between meals) ICDAS ≥ 1 Models 4, 5 Model 4 N = 3 dmfs/DMFS ICDAS ≥ 3 at end + • Caries experience (tooth extracted) Model 5 N = 3 dmfs/DMFS ICDAS ≥ 3 at start + • Caries experience (tooth extracted) 2 years follow-up ICDAS ≥ 3 Models 6, 7, 8 Model 6 N = 3 • Life-style situational (time elapsed since last dental visit) • Caries experience (tooth restored) Model 7 N = 4 Model 6 + dmfs/DMFS ICDAS ≥ 3 at end Model 8 N = 3 dmfs/DMFS ICDAS ≥ 3 at start + • Caries experience (child had tooth restored) ICDAS ≥ 1 Models 9, 10, 11 Model 9 N = 4 • Societal structural (CG has current caries, CG received a referral note for child) • Caries experience (current caries, tooth restored) Model 10 N = 5 Model 9 + dmfs/DMFS ICDAS ≥ 3 at end Model 11 N = 3 dmfs/DMFS ICDAS ≥ 3 at start + • Societal structural (CG has current caries, CG received a referral note for child) | 1 year follow-up ICDAS ≥ 3 Model 1 Sens 0.80, Spec 0.58, AUC 0.75 Model 2 Sens 0.81, Spec 0.58, AUC 0.77 Model 3 Sens 0.81, Spec 0.57, AUC 0.79 ICDAS ≥ 1 Model 4 Sens 0.79, Spec 0.58, AUC 0.77 Model 5 Sens 0.79, Spec 0.58, AUC 0.77 2 years follow-up ICDAS ≥ 3 Model 6 Sens 0.73, Spec 0.61, AUC 0.70 Model 7 Sens 0.73, Spec 0.61, AUC 0.73 Model 8 Sens 0.73, Spec 0.61, AUC 0.70 ICDAS ≥ 1 Model 9 Sens 0.82, Spec 0.59, AUC 0.75 Model 10 Sens 0.84, Spec 0.59, AUC 0.76 Model 11 Sens 0.75, Spec 0.61, AUC 0.77 | 1 year follow-up ICDAS ≥ 3 Model 1 LR + 1.90 (1.57–0.31) LR − 0.34 (0.26–0.46) Model 2 LR + 1.93 (1.59–2.34) LR − 0.33 (0.24–0.44) Model 3 LR + 1.88 (1.65–0.28) LR − 0.33 (0.25–0.45) ICDAS ≥ 1 Model 4 LR + 1.88 (1.34–2.64) LR − 0.36 (0.26–0.50) Model 5 LR + 1.88 (1.34–2.64) LR − 0.36 (0.26–0.50) 2 years follow-up ICDAS ≥ 3 Model 6 LR + 1.87 (1.49–2.35) LR − 0.44 (0.35–0.56) Model 7 LR + 1.87 (1.49–2.35) LR − 0.44 (0.35–0.56) Model 8 LR + 1.87 (1.49–2.35) LR − 0.44 (0.35–0.56) ICDAS ≥ 1 Model 9 LR + 2.00 (1.35–2.95) LR − 0.31 (0.22–0.43) Model 10 LR + 2.05 (1.39–3.02) LR − 0.27 (0.19–0.39) Model 11 LR + 1.92 (1.28–2.89) LR − 0.41 (0.30–0.56) | Items related to caries experience in child or caregiver and caregiver’s rating of child’s oral health could be used to screen at-risk children (aged 5–13) in this rural population. The models were fair in their ability to predict caries | Caries prevalence at baseline NR Caries increment: -1 year 89% had ≥ 1 Surface with any Progression and 61% With progression Towards cavitation -2 years 91% had ≥ 1 Surface with any Progression and 68% With progression Towards cavitation ICDAS Ismail et al,. [34] Code 1: when seen wet no evidence of any change in colour attributable to carious activity, but after prolonged air drying a carious opacity (white or brown lesion) is visible that is not consistent with the clinical appearance of sound enamel Code 3: localized enamel breakdown because of caries with no visible dentin or underlying shadow |
Gao. 2010 [17] | N = 6–12 predictors in 5 models Predictor common for all models • • Physiological (age) Prediction models Model 1 (screening) N = 8 • Societal structural (father’s education, • Life-style situational (between-meal-sweets) • Physiological (race, months of breastfeeding, no health problems) • Oral biological (plaque index) • Caries experience (dmft ˃ 2 cavities) Model 2 (full-blown) N = 12 • Societal structural (father’s education, no annual dental checks, age regarded by parents as appropriate for dental check • Life-style situational (months of breastfeeding, using fluorides) • Physiological (age, no health problems) • Oral biological (plaque index, ms in saliva, lbc in saliva, average pH) • Caries experience (dmft ˃ 2 cavities) Risk models Model 3 (screening) N = 7 • Societal structural (never lived in non-fluoridated community • Life-style situational (bedtime feeding, between-meal-sweets, bedtime sweets, months of breastfeeding) • Physiological (age) • Oral biological (plaque index) Model 4 (full-blown) N = 6 • Life-style situational (months of breastfeeding) • Physiological (age) • Oral biological (plaque index, ms in saliva, lbc in saliva, average pH) Model 5 Community-screening model N = 6 • Societal structural (parent’s belief about “Tooth worm”, parents do not know about bedtime milk, child’s dental status according to parents) • Life-style situational (using fluorides) • Physiological (race) | Prediction models Model 1 (screening) Sens 0.82, Spec 0.73, AUC 0.85 Model 2 (full-blown) Sens 0.90, Spec 0.90, AUC 0.96 Risk models Model 3 (screening) Sens 0.81, Spec 0.62, AUC 0.77 Model 4 (full-blown) Sens 0.83, Spec 0.92, AUC 0.95 Model 5 Community-screening model Sens 0.82, Spec 0.81, AUC 0.89 | Prediction models Model 1 (screening) LR + 3.06 (2.73–3.43) LR − 0.24 (0.21–0.29) Model 2 (full-blown) LR + 9.04 (7.41–11.03) LR − 0.11 (0.08–0.13) Risk models Model 3 (screening) LR + 2.15 (1.96–2.35) LR − 0.30 (0.26–0.36) Model 4 (full-blown) LR + 10.26 (8.20–12.84) LR − 0.18 (0.16–0.22) Model 5 Community-screening model LR + 4.37 (3.80–5.04) LR − 0.22 (0.19–0.22) | Risk models were established for a range of uses in community and clinical setting | Caries prevalence at baseline 40% Caries increment (1 year) 44% (dmft) |
Hänsel Petersson, 2002 [18] | N = 12 Cariogram • Societal structural (school, dental clinic) • Life-style situational (diet content, diet frequency, fluoride program) • Physiological (sex, related diseases) • oral biological (ms in saliva, saliva secretion, saliva buffer, plaque amount) • Caries experience (DMFT with cavities) | Sens 0.41, Spec 0.80 | LR + 2.03 (1.48–2.79) LR − 0.74 (0.63–0.87) | Cariogram predicted caries increment in this population (aged 10–11) more accurately than any included single-factor model | Data collected as lbc counts used as a measure of cariogenic diet Predictors in oral biological and caries experience levels given a score according to a predetermined scale with 3–4 scores Caries prevalence at baseline 40% Caries increment (2 years) 31% |
Pang, 2021 [19] | N = 7 • Societal structural (one-child family) • Physiological (sex) • Oral biological (plaque index, cariostat score) • Tooth (genetic markers rs3790506-enamel formation gene, rs1996315-water channel protein gene AQP5 • Caries experience (DMFT ICDAS codes 3–6 = > 0 decayed teeth) | Total AUC 0.70 (0.66–0.74) Low caries risk (DMFT ≤ 1 caries lesion) Sens 0.41, Spec 0.74, PPV 0.42, NPV 0.73 Moderate caries risk Sens 0.46, Spec 0.69, PPV 0.49, NPV 0.66 High caries risk Sens 0.54, Spec 0.69, PPV 0.74, NPV 0.48 Very high caries risk Sens 0.68, Spec 0.75, PPV 0.95, NPV 0.25 | Low caries risk LR + 1.57 (1.24–1.99) LR − 0.80 (0.71–0.89) Moderate caries risk LR + 1.48 (1.20–1.82) LR − 0.79 (0.69–0.89) High caries risk LR + 1.73 (1.42–2.12) LR − 0.67 (0.58–0.77) Very high caries risk LR + 2.71 (2.18–3.38) LR − 0.43 (0.3–0.51) | Model based on both environmental and genetic factors using an algorithm based on machine learning | Caries prevalence at baseline 34% Caries increment (1.7 years) 58% Presents calibration. Study of model validation described below ICDAS Pitts and Ekstrand, [35] Code 3 = clinically detectable “cavities” limited to enamel, Codes 4,5 = clinically detectable lesions in dentine, Code 6 = lesions into pulp |
Sánchez-Pérez, 2009 [20] | N = 4 • Oral biological (acid production by bacteria in saliva) • Tooth (total teeth present, fissure morphology) • Caries experience (dmfs + DMFS ≥ 1 cavity) | Sens 0.79, Spec 0.80, AUC 0.88 | LR + 3.85 (2.04–7.27) LR − 0.27 (0.16–0.46) | In a developing country, caries experience was most powerful to predict caries in 6-year-old children over 3 years, but prediction was improved by fissure morphology and bacterial acid production | Caries prevalence at baseline 58% Caries increment (3 years) 59% |
Coronal and root caries | |||||
Powell, 1991 [21] | N = 3 • Physiological (sex) • Oral biological (salivary secretion) • Caries experience (root caries index) | Threshold ≥ l new caries Sens 0.88, Spec 0.60 | Threshold ≥ l new caries LR + 2.20 (0.74–6.53) LR − 0.20 (0.04–0.90) | Model was sensible and variables easy to collect | Caries prevalence at baseline 53% Caries increment coronal and/or root caries (1 year) 76% |
Root caries | |||||
Ritter, 2016 [22] | N = 8–13 predictors in 5 models Predictors common for model 1 and all other models Model 1 N = 8 • Physiological (age, sex, race) • Caries experience (at-risk years, at-risk root surfaces, at-risk root surfaces squared, root caries index, root caries index squared) Model 2 N = 9 Model 1 + • Life-style situational (tobacco use) Model 3 N = 13 Model 1 + • Life-style situational (cereals + sugar, tea/coffee + sugar, drinks and juices, juices, sweets) Model 4 N = 11 Model 1 + • Life-style situational (toothbrush use, how often toothbrush is replaced) • Physiological (wears removable prosthesis–appliances) Model 5 N = 11 Model 1 + • Societal structure (income, education, insurance) | Primary data analysis (sample n = 155) Model 1 Sens 0.74, Spec 0.61, AUC 0.83 Model 2 Sens 0.76, Spec 0.66, AUC 0.84 Model 3 Sens 0.71, Spec 0.61, AUC 0.84 Model 4 Sens 0.74, Spec 0.62, AUC 0.83 Model 5 Sens 0.75, Spec 0.63, AUC 0.83 | Model 1 LR + 1.88 (1.38–2.55) LR − 0.43 (0.29–0.66) Model 2 LR + 2.23 (1.60–3.10) LR − 0.36 (0.23–0.56) Model 3 LR + 1.81 (1.33–2.47) LR − 0.48 (0.32–0.71) Model 4 LR + 1.94 (1.42–2.65) LR − 0.42 (0.28–0.64) Model 5 LR + 2.04 (1.49–2.81) LR − 0.39 (0.26–0.60) | Model 2 presented best performance. The results can inform identification of high-risk root caries individuals only in a caries-active population | Caries (coronal and root) prevalence at baseline 100% Caries increment (3 years) 49% |
Sánchez-García, 2011 [23] | N = 6 • Life- style situational (dental mouthwash–oral hygiene routines, smoking) • Physiological (limitations in basic daily activities–general health) • Oral biological (ms ≥ 105 CFU/ml saliva) • Caries experience (healthy root surfaces, root caries index) | Sens 0.16, Spec 0.98, PPV 0.67, NPV 0.81 AUC 0.75 | LR + 7.09 (3.29–15.30) LR + 0.86 (0.80–0.93) | A good prediction model for 21-month root caries incidence in elderly (mean age 71.8; range 60–74) | Caries prevalence at baseline coronal caries 100% root caries 44% Caries increment (1 year) 22% |
Studies of model validation | |||||
Coronal caries | |||||
Beck, 1992 [24] | Modified models of high-risk model Disney et al. [15] N = 20 Model 1 Any risk prediction model • Societal structural (education-head of household, referral caries score) • Life-style situational (tooth-brushing, F-tablets, dental visit last year–3 predictors) • Physiological (race) • Oral biological (ms in saliva, lbc in saliva) • Tooth (tooth morphology, sound permanent surfaces) • Caries experience (baseline dmfs, baseline DMFS, white spot lesions) • Other predictors (examiner agreement–4 predictors, predicted caries score) N = 13 Model 2 Any risk etiologic model • life-style situational (dental visit last year–2 predictors) • Tooth (tooth morphology, fluorosis, sound permanent surfaces, sealants) • Oral biological (ms in saliva, lbc in saliva, mean plaque score) • Other predictors (examiner agreement–4 predictors, predicted caries score) | Model 1 Any risk prediction model Aiken age 6.6 years Sens 0.80, Spec 0.61, PPV 0.73, NPV 0.69 Portland age 6.9 years Sens 0.66, Spec 0.78, PPV 0.57, NPV 0.84 Aiken age 10.7 years Sens 0.84, Spec 0.54, PPV 0.78, NPV 0.64 Portland age 10.8 years Sens 0.76, Spec 0.71, PPV 0.68, NPV 0.78 Model 2 Any risk etiologic model Aiken age 6.6 years Sens 0.74, Spec 0.55, PPV 0.68, NPV 0.62 Portland age 6.9 years Sens 0.59, Spec 0.74, PPV 0.51, NPV 0.80 Aiken age 10.7 years Sens 0.81, Spec 0.50, PPV 0.75, NPV 0.58 Portland age 10.8 years Sens 0.69, Spec 0.65, PPV 0.62, NPV 0.71 | Model 1 Any risk prediction model Aiken age 6.6 years LR + 2.05 (1.82–2.31) LR − 0.33 (0.28–0.39) Portland age 6.9 years LR + 3.00 (2.57–3.50) LR − 0.44 (0.37–0.51) Aiken age 10.7 years LR + 1.83 (1.62–2.06) LR − 0.30 (0.24–0.36) Portland age 10.8 years LR + 2.62 (2.27–3.03) LR − 0.34 (0.28–0.40) Model 2 Any risk etiologic model Aiken age 6.6 years LR + 1.64 (1.47–1.84) LR − 0.47 (0.40–0.55) Portland age 6.9 years LR + 2.27 (1.95–2.64) LR − 0.55 (0.48–0.63) Aiken age 10.7 years LR + 1.62 (1.44–1.82) LR − 0.38 (0.31–0.46) Portland age 10.8 years LR + 1. 97 (1.73–2.25) LR − 0.48 (0.41–0.56) | Any risk etiologic models with fewer significant variables, appear to have similar, but slightly lower utility (sensitivity and specificity) compared to Any risk prediction models and appear to be more broadly applicable across populations | Caries prevalence at baseline NR Caries increment (3 years) Aiken 6.6 years 57% 10.7 years 66% Portland 6.9 years 31% 10.8 years 45% High-risk model identical with that presented by Disney et al. [15] |
Birpou, 2019 [25] | N = 7–9 predictors in 8 modified Cariogram models with 1- and 2 years follow-up Standard set Model 1 Cariogram 1 N = 9 • Life-style situational (diet content, diet frequency, fluoride programme) • Physiological (related diseases) • Oral biological (ms in saliva with four thresholds, saliva buffer, plaque amount) • Caries experience (dmft with two thresholds) • Other predictors (clinical judgement) Model 2 Cariogram 2 N = 8 Model 1 excluding saliva buffer Model 3 Cariogram 3 N = 8 Model 1 excluding ms in saliva Model 4 Cariogram 4 N = 7 Model 1 excluding saliva buffer and ms in saliva High risk set Model 5 Cariogram 5 N = 9 Same predictors as Model 1 Model 6 Cariogram 6 N = 8 Model 1 excluding saliva buffer Model 7 Cariogram 7 N = 8 Model 1 excluding ms in saliva Model 8 Cariogram 8 N = 7 Model 1 excluding saliva buffer and ms in saliva | Standard set Model 1 Cariogram 1 1 year follow-up Sens 0.66, Spec 0.57, AUC 0.62 2 years follow-up Sens 0.68, Spec 0.59, AUC 0.65 Model 2 Cariogram 2 1 year follow-up Sens 0.65, Spec 0.57, AUC 0.62 2 years follow-up Sens 0.66, Spec 0.59, AUC 0.62 Model 3 Cariogram 3 1 year follow-up Sens 0.60, Spec 0.65, AUC 0.61 2 years follow-up Sens 0.64, Spec 0.68, AUC 0.65 Model 4 Cariogram 4 1 year follow-up Sens 0.56, Spec 0.65, AUC 0.61 2 years follow-up Sens 0.60, Spec 0.68, AUC 0.65 High risk set Model 5 Cariogram 5 1 year follow-up Sens 0.66, Spec 0.57, AUC 0.62 2 year follow-up Sens 0.68, Spec 0.58, AUC .65 Model 6 Cariogram 6 1 year follow-up Sens 0.65, Spec 0.65, AUC 0.62 2 years follow-up Sens 0.66, Spec 0.59, AUC 0.66 Model 7 Cariogram 7 1 year follow-up Sens 0.57, Spec 0.65, AUC 0.61 2 years follow-up Sens 0.61, Spec 0.68, AUC 0.65 Model 8 Cariogram 8 1 year follow-up Sens 0.56, Spec 0.65, AUC 0.62 2 years follow-up Sens 0.60, Spec 0.68, AUC .66 | Standard set Model 1 Cariogram 1 1 year follow-up LR + 1.55 (1.13–2.13) LR − 0.59 (0.41–0.85) 2 years follow-up LR + 1.63 (1.17–2.26) LR − 0.55 (0.38–0.82) Model 2 Cariogram 2 1 year follow-up LR + 1.52 (1.11–2.09) LR − 0.61 (0.42–0.88) 2 years follow-up LR + 1.60 (1.15–2.22) LR − 0.58 (0.40–0.84) Model 3 Cariogram 3 1 year follow-up LR + 1.69 (1.17–2.44) LR − 0.62 (0.45–0.86) 2 years follow-up LR + 1.97 (1.33–2.90) LR- 0.54 (0.38–0.76) Model 4 Cariogram 4 1 year follow-up LR + 1.58 (1.09–2.30) LR − 0.68 (0.50–0.93) 2 years follow-up LR + 1.88 (1.24–2.74) LR − 0.60 (0.43–0.83) High risk set Model 5 Cariogram 5 1 year follow-up LR + 1.55 (1.13–2.13) LR − 0.59 (0.41–0.85) 2 years follow-up LR + 1.63 (1.17–2.26) LR − 0.55 (0.38–0.82) Model 6 Cariogram 6 1 year follow-up LR + 1.85 (1.29–2.65) LR − 0.54 (0.38–0.77) 2 years follow-up LR + 1.60 (1.15–2.22) LR − 0.58 (0.40–0.84) Model 7 Cariogram 7 1 year follow-up LR + 1.62 (1.12–2.35) LR − 0.66 (0.49–0.90) 2 years follow-up LR + 1.88 (1.27–2.79) LR − 0.58 (0.42–0.80) Model 8 Cariogram 8 1 year follow-up LR + 1.58 (1.09–2.30) LR − 0.68 (0.50–0.93) 2 years follow-up LR + 1.84 (1.24–2.74) LR − 0.60 (0.43–0.83) | Cariogram with various factors and settings displayed suboptimal ability to predict caries in this population (aged 2–5) | Caries prevalence at baseline 37% Caries increment 1 years 52% 2 years 53% Considering the high prevalence of caries increment it may have been expected that the high set would result in higher sensitivity and specificity than standard set |
Campus, 2012 [26] | N = 7 Modified Cariogram • Life-style situational (diet content, diet frequency, fluoride programme) • Physiological (related diseases) • Oral biological (≥ 105 ms in saliva, plaque amount) • Caries experience (cavity from 0–caries-free to 3 as sum of dmft and DMFS) | Sens 0.83, Spec 0.85, AUC 0.93 | LR + 5.53 (4.36–7.03) LR − 0.2 (0.16–0.25) | Results showed high validity of Cariogram in schoolchildren | Data collected as lbc counts used as a measure of cariogenic diet Caries prevalence at baseline 29% Caries increment (2 years) 54% |
Christian, 2020 [27] | N = 13 Modified CAMBRA • Societal structure (parent/caregiver: low socioeconomic status and/or low health literacy, mother/caregiver: decay-free last three years, mother/caregiver: xylitol chewing-gum/lozenges 2–4 × daily) • Life-style situational (developmental problems, no dental home/episodic dental care, dental home and regular dental care, frequent between-meal snacks of sugars/cooked starch/sugared beverages, continually uses bottle—contains fluids other than water, sleeps with a bottle or nurses on demand, lives in a fluoridated community, or takes fluoride supplements, fluoridated toothpaste daily) • Oral biological (obvious plaque on teeth and/or gums bleed easily) | Follow-up 1.5 years Sens 0.74, Spec 0.35, PPV 0.20, NPV 0.86, AUC close to 0.5 Follow-up 2.5 years Sens 0.70, Spec 0.36¸ PPV 0.37, NPV 0.69, AUC close to 0.5 | Follow-up 1.5 years LR + 1.14 (0.92–1.41), LR − 0.74 (0.42–1.31) Follow-up 2.5 years LR + 1.09 (0.90–1.33), LR − 0.83 (0.55–1.26) | CAMBRA in its current form may not be ideal for use in risk-based disease management among young Victorian children. Due to its low specificity, it is highly likely that the use of this risk assessment tool could be driving over-treatment | Caries prevalence at baseline 0% Caries increment 1.5 years 18% 2.5 years 35% |
Dolic, 2020 [28] | N = 9 Cariogram • Life-style situational (diet content, diet frequency, fluoride programme) • Physiological (related diseases) • Oral biological (ms in saliva, saliva secretion, saliva buffer, plaque amount) • Caries experience (cavity) | Threshold 1 (moderate, low, and very low risk) Sens 0.54, Spec 0.69, PPV 0.78, NPV 0.42 LR + 1.47, LR − 0.66 Threshold 2 (moderate, high, and very high risk) Sens 0.81, Spec 0.58, PPV 0.80, NPV 0.60 LR + 1.92, LR − 0.32 | Threshold 1 (moderate, low, and very low risk) LR + 1.74 (0.93–3.25) LR − 0.67 (0.45–0.98) Threshold 2 (moderate, high, and very high risk) LR + 1.93 (1.21–3.09) LR − 0.33 (0.17–0.62) | Cariogram can be a useful tool for caries prediction. It is valid and highly predictive | Data collected as lbc counts (Dentocult) used as a measure of cariogenic diet Caries prevalence at baseline NR Caries increment NR No report of thresholds for ms, lbc, or caries experience |
Gao, 2013 [29] | N = 5–14 in 10 different models Model 1 NUS-CRA Comprehensive N = 11 • Societal structure (family socioeconomic status) • Life-style situational (infant feeding history, diet, fluoride. plaque index) • Physiological (age, ethnicity, systemic health) • Oral biological (ms in saliva, lbc in saliva) • Caries experience (cavity) Model 2 NUS-CRA screening N = 9 Model 1 excluding ms and lbc in saliva Model 3 Cariogram Comprehensive N = 9 predictors see study below by Hänsel Petersson et al. [30] Model 4 Cariogram screening N = 5 Model 3 excluding saliva secretion, saliva buffer, ms in saliva, lbc in saliva Model 5 CAT Comprehensive I N = 12 • Societal structural (family socioeconomic status, dental attendance) • Life-style situational (diet, fluoride, oral hygiene, dental appliance) • Physiological (systemic health) • Oral biological (ms in saliva, saliva secretion) • Caries experience (past caries, white spot lesion, enamel defects) Model 6 CAT comprehensive II N = 11 Model 5 without family socioeconomic status Model 7 CAT screening I N = 10 Model 5 excluding saliva secretion and ms in saliva Model 8 CAT screening II N = 9 Model 7 without socioeconomic status Model 9 CAMBRA Comprehensive N = 14 • Societal structural (family socioeconomic status) • Life-style situational (infant feeding history, diet, fluoride, oral hygiene, dental appliance, dental attendance) • Physiological (systemic health, medication) • Oral biological (ms in saliva, lbc in saliva, saliva secretion) • Caries experience (past caries, white spot lesion) Model 10 CAMBRA Screening N = 11 Model 9 excluding ms in saliva, lbc in saliva, saliva secretion | Model 1 NUS-CRA comprehensive Sens 0.78, Spec 0.85, AUC 0.88 Model 2 NUS-CRA screening Sens 0.74, Spec 0.88, AUC 0.85 Model 3 Cariogram comprehensive Sens 0.65, Spec 0.79, AUC 0.78 Model 4 Cariogram screening Sens 0.63, Spec 0.78, AUC 0.76 Model 5 CAT comprehensive I Sens 1, Spec 0.04 Model 6 CAT comprehensive II Sens 0.99, Spec 0.04 Model 7 CAT creening I Sens 0.98, Spec 0.05 Model 8 CAT screening II Sens 0.98, Spec 0.05 Model 9 CAMBRA comprehensive Threshold ≥ Moderate Sens 0.92, Spec 0.40 Threshold ≥ High Sens 0.84, Spec 0.63 Model 10 CAMBRA Screening Threshold ≥ Moderate Sens 0.98, Spec 0.20 Threshold ≥ High Sens 0.94, Spec 0.44 | Model 1 NUS-CRA comprehensive LR + 5.31 (4.01–7.03) LR − 0.26 (0.19–0.34) Model 2 NUS-CRA screening LR + 4.81 (3.64–6.35) LR − 0.31 (0.24–0.40) Model 3 Cariogram comprehensive LR + 3.0 (2.36–3.82) LR − 0.45 (0.37–0.55) Model 4 Cariogram screening LR + 2.85 (2.24–3.61) LR − 0.48 (0.39–0.58) Model 5 CAT comprehensive I LR + 1.04 (1.02–1.06) LR − 0 Model 6 CAT comprehensive II LR + 1.03, (1.01–1.06) LR − 0.15 (0.02–1.10) Model 7 CAT screening I LR + 1.04 (1.01–1.08) LR − 0.21 (0.05–0.92) Model 8 CAT Screening II LR + 1.04 (1.0–1.08) LR − 0.31 (0.09–1.04) Model 9 CAMBRA Comprehensive Threshold ≥ Moderate LR + 1.52 (1.37–1.68) LR − 0.21 (0.13–0.35) Threshold ≥ High LR + 2.26 (1.92–2.65) LR − 0.26 ( 0.18–0.37) Model 10 CAMBRA Screening Threshold ≥ Moderate LR + 1.22 (1.15–1.30) LR- 0.14 (0.06–0.33) Threshold ≥ High LR + 1.66 (1.50–1.86) LR − 0.14 (0.08–0.26) | Our findings supported algorithm modelling. NUS-CRA appeared to be a baseline useful program with sufficient sensitivity and specificity in Hong Kong children | Caries prevalence at baseline 35% Caries increment (1 year) 37% |
Hänsel Petersson, 2015 [30] | N = 9 Cariogram • Life-style situational (diet content, diet frequency, fluoride programme) • Physiological (related diseases) • Oral biological (ms in saliva with four thresholds, saliva secretion, saliva buffer, plaque amount) • Caries experience (DMFT cavity with four thresholds) | Threshold % DFS new lesions 80% Sens 0.89, Spec 0.34 PPV 0.42, NPV 0.85 60% Sens 0.61, Spec 0.71 PPV 0.53, NPV 0.77 40% Sens 0.26, Spec 0.91 PPV 0.60, NPV 0.69 20% Sens 0.12, Spec 0.95 PPV 0.55, NPV 0.67 | Threshold % DFS new lesions 80% LR + 1.34 (1.26–1.43) LR − 0.32 (0.23–0.45) 60% LR + 2.11 (1.82–2.45) LR − 0.55 (0.47–0.63) 40% LR + 2.72 (2,02–3.68) LR − 0.82 (0.77–0.88) 20% LR + 2.29 (1.48–3.55) LR − 0.93 (0.89–0.97) | Cariogram did not perform better than a risk assessment scheme based on past Caries experience and caries progression, over a 3-year period in young adults | Data collected as lbc counts used as a measure of cariogenic diet Caries prevalence at baseline 77% Caries increment (3 years) 35% |
Hänsel Petersson, 2010 [31] | Model 1 Cariogram N = 9 • Life-style situational (diet content, diet frequency, fluoride program) • Physiological (related diseases) • Oral biological (ms in saliva with four thresholds, saliva secretion, saliva buffer, plaque amount) • Caries experience (dentin caries DMFS with four thresholds) Model 2 N = 8 Model 1 excluding ms in saliva Model 3 N = 8 Model 1 excluding saliva buffer Model 4 N = 8 Model 1 excluding saliva secretion Model 5 N = 6 Model 1 excluding ms in saliva, saliva buffer, saliva secretion | Model 1 Sens 0.73, Spec 0.60 PPV 0.45, NPV 0.83, LR + 1.80, LR- 0.45 AUC 0.75 Model 2 Sens 0.84, Spec 0.47 PPV 0.41, NPV 0.86, LR + 1.60, LR − 0.36 AUC 0.73 Model 3 Sens 0.79, Spec 0.51 PPV 0.42, NPV 0.85, LR + 1.60, LR − 0.41 AUC 0.75 Model 4 Sens 0.77, Spec 0.49 PPV 0.41, NPV 0.83, LR + 1.50, LR − 0.45 AUC 0.75 Model 5 Sens 0.90, Spec 0.20 PPV 0.34, NPV 0.82, LR + 1.10, LR − 0.50 AUC 0.72 | Model 1 LR + 1.83, (1.52–2.19) LR − 0.45 (0.33–0.61) Model 2 LR + 1.58 (1.38–1.82) LR − 0.34 (0.22–0.52) Model 3 LR + 1.61 (1.38–1.88) LR − 0.41 (0.29–0.59) Model 4 LR + 1.51 (1.30–1.76) LR − 0.47 (0.33–0.66) Model 5 LR + 1.13 (1.03–1.22) LR − 0.50 (0.28–0.90) | Accuracy of caries prediction in school children was significantly impaired when Cariogram was applied without enumeration of salivary tests. ms enumeration seemed to be most important of the salivary variables | Data collected as lbc counts used as a measure of cariogenic diet Caries prevalence at baseline 40% Caries increment (2 years) 31% |
Holgerson, 2009 [32] | N = 7 Modified Cariogram • Life-style situational (diet frequency, oral hygiene, fluorides) • Physiological (related diseases) • Oral biological (ms in saliva—counts with four thresholds) • Caries experience (above average for age group) • Other predictors (clinical judgement) | Sens 0.46, Spec 0.88 PPV 0.90, NPV 0.40, LR + 3.7, LR − 0.6 | LR + 3.83 (1.39–10.58) LR − 0.61 (0.40–0.94) | Modified Cariogram in 2-year-old children resulted in high sensitivity for future caries, but the method lacked accuracy and precision | Caries prevalence at baseline 3% Caries increment (5 years) 71% Clinical judgement added after result of Cariogram was obtained |
Pang, 2021 [19] | N = 7 Predictors described below Pang 2021 in studies of model development described above | Total AUC 0.73 (0.68–0.79) Low caries risk (DMFT ≥ 1 caries lesion) Sens 0.29, Spec 0.63, PPV 0.29, NPV 0.63 Moderate caries risk Sens 0.34, Spec 0.66, PPV 0.48, NPV 0.52 High caries risk Sens 0.59, Spec 0.68, PPV 0.74, NPV 0.53 Very high caries risk Sens 0.66, Spec 0.58, PPV 0.90, NPV 0.22 | Low caries risk LR + 0.78 (0.57–1.08) LR − 1.13 (0.96–1.33) Moderate caries risk LR + 1.00 (0.73–1.38) LR − 1.00 (0.85–1.18) High caries risk LR + 1.86 (1.39–2.48) LR- 0.60 (0.49–0.74) Very high caries risk LR + 1.54 (1.22–1.94) LR − 0.60 (0.47–0.76) | This caries risk prediction model can accurately identify a high-risk population. The model can be utilized as a powerful tool at community level | Sample different for studies of model development and model validation Caries prevalence at baseline 40% Caries increment (1.7 years) 63% ICDAS codes 3–6, Pitts and Ekstrand, [35] |
Root caries | |||||
Hayes, 2017 [33] | N = 9 Model 1 Cariogram • Life-style situational (diet content, diet frequency, fluoride programme) • Physiological (related diseases) • Oral biological (ms ≥ 105 counts per ml saliva, saliva secretion, saliva buffer, plaque amount) • Caries experience (normal caries experience: mean DMFT score 22.4 ± 5.3 range 17–28) Model 2 N = 8 Model 1 excluding ms in saliva Model 3 N = 8 Model 1 excluding saliva buffer Model 4 N = 8 Model 1 excluding saliva secretion Model 5 N = 6 Model 1 excluding ms in saliva, saliva buffer, saliva secretion | Model 1 Cariogram Sens 0.79, Spec 0.63, PPV 0.40, NPV 0.90 AUC 0.77 Model 2 Model 1 excluding ms Sens 0.74, Spec 0.73, PPV 0.48, NPV 0.90 AUC 0.80 Model 3 Model 1 excluding saliva buffer Sens 0.79, Spec 0.58, PPV 0.39, NPV 0.89 AUC 0.76 Model 4 Model 1 excluding saliva secretion Sens 0.79, Spec 0.58, PPV 0.39, NPV 0.89 AUC 0.77 Model 5 Model 1 excluding ms, saliva buffer, saliva secretion Sens 0.73, Spec 0.66, PPV 0.42, NPV 0.88 AUC 0.79 | Model 1 Cariogram LR + 2.12 (1.71–2.63) LR − 0.34 (0.21–0.54) Model 2 Model 1 excluding ms LR + 2.78 (2.14–3.62) LR − 0.35 (0.23–0.52) Model 3 Model 1 excluding saliva buffer LR + 1.88 (1.53–2.29) LR − 0.37 (0.23–0.59) Model 4 Model 1 excluding saliva secretion LR + 1.88 (1.53–2.29) LR − 0.37 (0.23–0.59) Model 5 Model 1 excluding ms, saliva buffer, saliva secretion LR + 2.13 (1.68–2.69) LR − 0.41 (0.28–0.61) | Cariogram may be clinically useful in determining future root caries in independently living older dentate adults (aged ˃ 65) | Data collected as lbc counts used as a measure of cariogenic diet Caries prevalence at baseline NR Caries increment (2 years) 25% |