Introduction 27 4 7 1 11 14 20 5 6 9 10 12 20 2 3 11 13 19 23 24 28 8 Materials and methods This cohort study was conducted in the rainforest of Suriname, South America. It was part of a large-scale project investigating the influence of dental treatment on the oral health of children (van Gemert-Schriks et al., submitted, 2007). Within the scope of that particular project, 380 6-year-old children were divided randomly among four different treatment groups. Material presented in the current article concerns only those children who received restorative treatment, according to the ART method, either in their primary or permanent dentitions. 4 7 23 29 1 Table 1 Evaluation criteria for the ART restorations Code Evaluation characteristics 00 Restoration present, correct 10 Restoration present, slight marginal defect/wear of surface (<0.5 mm). No repair needed. 11 Restoration present, gross marginal defect/wear of surface (>0.5 mm). Repair needed. 12 Restoration present, underfilled (>0.5 mm). Repair needed. 13 Restoration present, overfilled (>0.5 mm). Repair needed. 20 Secondary caries, discoloration in depth, surface hard and intact, caries within dentin. Repair needed. 21 Secondary caries, surface defect, caries within dentin. Repair needed. 30 Restoration not present, bulk fracture, moving or partial lost. Repair needed. 40 Inflammation of the pulp; signs of dentogenic infection (abscesses, fistulae, pain complaints). Restoration might still be in situ. Extraction needed. 50 Tooth not present because of extraction 60 Tooth not present because of shedding 70 Tooth not present because of extraction or shedding 90 Patient not present Statistical analysis Statistical analyses were performed using SPSS for Windows, version 12.0.1 (SPSS, Chicago, USA). All significant differences were detected at a 95% confidence level. Kaplan–Meier survival analyses were performed on the censored data of both single- and two-surface restorations. The significance of differences between survival curves was determined with log-rank tests. Possible confounding variables were taken into account using a Cox regression analysis. Results Materials and methods 17 2 U p df p Table 2 Baseline data for the ART restorations   Primary dentition Permanent dentition Number of filled surfaces 1 2 1 Number of restorations 133 342 54 N 61 147 34 Mean number of restorations per child (SD; range) 3.50 (1.61; 1–7) 3.64 (1.73; 1–8) 2.07 (0.97; 1–4) Dentist 1 43 (32.3%) 74 (21.6%) 12 (22.2%) 2 41 (30.8%) 84 (24.6%) 16 (29.6%) 3 34 (25.6%) 89 (26.0%) 7 (13.0%) 4 a 95 (27.8%) 19 (35.2%) Adjacent tooth present Yes 117 (88.0%) 303 (88.6%) 45 (83.3%) No 16 (12.0%) 39 (11.4%) 9 (16.7%) Contamination blood/saliva Yes 13 (9.8%) a 5 (9.3%) No 120 (90.2%) 232 (67.8%) 49 (90.7%) Venham behavior score 0 50 (37.6%) 78 (22.8%) 8 (14.8%) 1 44 (33.1%) 137 (40.1%) 27 (50.0%) 2 26 (19.5%) 82 (24.0%) 13 (24.1%) 3 13 (9.8%) 33 (9.6%) 6 (11.1%) 4 – a – 5 – – – SD a p 1 2 Fig. 1 a b  Fig. 2 Survival curve single-surface ART restorations, permanent dentition  3 Table 3 Failure characteristics for the ART restorations at 3 years   Primary dentition Permanent dentition 1 surface 2 surface 1 surface N 133 342 54 N 42 251 32 Failure score    Gross marginal defect (11) 21 (15.8%) 86 (25.1%) 13 (24.1%) Restoration present, underfilled (12) 1 (0.8%) 9 (2.6%) – Restoration present, overfilled (13) 2 (1.5%) 11 (3.2%) 2 (3.7%) Sec. caries, discoloration (20) – – – Sec. caries, surface defect (21) 5 (3.8%) 1 (0.3%) 14 (25.9%) Total or partial loss (30) 13 (9.8%) 120 (35.1%) 3 (5.6%) Pulpal inflammation (40) – 24 (7.0%) – Restoration missing, extracted (50) – – – Scores 60–90 were not included (censored data) df p 3 Fig. 3 Survival curves per dentist, multi-surface ART restorations primary dentition  To detect any confounding variables on the survival of the ART restorations, a Cox regression analysis was performed. No significant relation could be found, indicating that neither the presence or absence of an adjacent tooth, nor contamination with blood and/or saliva, nor the behavior of the child during the restorative phase of the treatment had an influence on the 3-year survival of the restorations in the primary dentition. No effect also could be found regarding the number of restorations per child. p Discussion In contrast with other studies, the results of this study show extremely low survival rates for both single- and two-surface ART restorations in the primary and permanent dentitions. An operator effect was observed for two-surface restorations only. Neither the behavior of the child during restoration, and the number of restorations per child, nor the contamination of preparations with blood or saliva had a significant influence on the survival of the restorations in this study. This field study was performed correctly and the statistical power was sufficiently high to detect at least medium effects. However, because it was part of a large-scale randomized controlled clinical trial, no comprehensive criteria were formulated beforehand regarding, for example, the number of restorations per patient, and the location and the size of the cavities. This aspect is inherent to many cohort studies and it does not imply an inferior study quality, but it limits a meaningful comparison with other survival studies. Although all possible efforts were exercised to trace the participating children over the evaluation period, 22 restorations (4.63%, eight children), all in primary molars, could not be evaluated at any of the recall visits. Either the children did not show up, or the teeth concerned had exfoliated before the first evaluation. These restorations were regarded as missing data and, therefore, excluded from further analysis. Twenty-six restorations (5.47%, 21 children) were “lost” for evaluation because the teeth either exfoliated or the child moved to another district during the course of the study, but after the first evaluation. These restorations (scores 60–90) were treated as censored data and not as true failures because they survived up to a certain moment. 4 9 15 21 26 28 7 Other patient-related factors that may influence the survival of the restorations are the behavior and saliva flow of the child. The survival of the ART restorations in this study was analyzed at the restoration level. This method requires independency of the restoration data and, with respect to the mentioned patient-related possible bias, this assumption could not be guaranteed. To control for this lack of independency, the survival analyses also were performed at the patient level, including only one randomly selected restoration per child. These analyses did not render higher survival rates. 6 13 25 26 13 16 22 5 6 10 26 7 18 26 30 31 The extremely low survival of the ART restorations observed in this study remains unexplained. Circumstances that were not recognized as possible interfering factors at the start of the study might have played an important role, including cultural and seasonal dietary influences. People living in the rainforest of Suriname eat seasonal fruits such as mangos and fruits of the fiber palm (Awarra). In particular, the latter may influence the survival of the restorations, given the frequency and method in which they are consumed. The authors have seen unusual wear patterns, also in adult dentitions, which might have been caused by excessive consumption of Awarras. A possible causality between these dietary habits and the survival of the ART restorations can only be disclosed by future controlled studies. 13 24 Conclusion The uncertain predictability for the success of ART may introduce further discussion about alternative treatment strategies, especially in those situations where choices have to be made with respect to a well-balanced, cost-effective package of basic oral health care. To gain insight into factors determining the cumulative success rate of ART restorations, future studies should focus in more detail on variables that could possibly contribute to the failure of restorations.