Longitudinal study of caries development from childhood to adolescenceAuthorsMs E. Hall-Scullin1, Consultant in Dental Public Health, NHS Ayrshire & Arran.Miss H. Whitehead2, Dental Professional Lead, Community Dental Service, Salford Royal NHS Foundation Trust.Prof K. Milsom3, Consultant in Dental Public Health, Department of Dental Public Health, Countess of Chester Hospital.Prof M. Tickle4, Professor of Dental Public Health and Primary Care, Division of Dentistry, University of Manchester.Dr T-L Su, Lecturer in Statistics, Division of Dentistry, University of ManchesterDr T. Walsh4, Reader in Biostatistics, Division of Dentistry, University of Manchester.

Address1Public Health Department, NHS Ayrshire & Arran, Afton House, Ailsa Hospital Campus, Dalmellington Road, Ayr, KA6 6AB, UK. 2Community Dental Service, Salford Royal NHS Foundation Trust, Pendleton Gateway, 1 Broadwalk, Salford, M6 5FX, UK. 3Department of Dental Public Health, 1829 Building, Countess of Chester Hospital, Chester, CH2 1UL, UK. 4Division of Dentistry, Faculty of Biology, Medicine and Health, University of Manchester, Oxford Road, Manchester, M13 9PL, UK.

Email Emma Hall-Scullin emma.hall-scullin@manchester.ac.uk Hilary Whitehead Hilary.Whitehead@srft.nhs.uk Ting-Li Su ting-li.su@manchester.ac.ukKeith Milsom Keith.Milsom@phe.gov.uk Martin Tickle martin.tickle@manchester.ac.uk Tanya Walsh tanya.walsh@manchester.ac.uk

Key words (from list provided)Child dentistry, Dental Public health, pediatric dentistry, prevention

Please provide one sentence that describes your article.

Adds to the literature on the longitudinal behaviour (incidence and prevalence) of dental caries in adolescence to inform oral health strategies.

Abstract word count 296 words tbc

Total word count 2882 words tbc

1

Total number of tables/figures 4 tbc

Number of references 18

Abstract

Introduction: The World Health Organization (WHO) has concluded that

globally, dental caries is the most important oral condition. To develop

effective prevention strategies requires an understanding of how this condition

develops and progresses over time, but there are few longitudinal studies of

caries onset and progression in children. Methods: The aim of the study was

to establish the pattern of caries development from childhood into

adolescence and to explore the role of potential risk factors (age, gender,

ethnicity and social deprivation). Of particular interest was the disease

trajectory of dentinal caries in the permanent teeth in groups defined by the

presence or absence of dentinal caries in the primary teeth. Intra-oral

examinations to assess oral health were performed at four time points by

trained and calibrated dentist examiners using a standardized, national

diagnostic protocol. Results: Clinical data were available from 6651 children.

Mean caries prevalence (% D3MFT>0) was 16.7% at the first clinical

examination (ages 7 to 9) increasing to 31.0%, 42.2% and 45.7% at

subsequent examinations. A population-averaged model (generalized

estimating equations) was used to model the longitudinal data. Estimated

mean values indicated a rising D3MFT count as pupils aged (consistent with

new teeth emerging) which was significantly higher (4.49 times, 95% CI 3.90

to 5.16) in those pupils with caries in their primary dentition than in those

without. Conclusion: This study is one of the few large longitudinal studies to

report the development of dental caries from childhood into adolescence.

Children who developed caries in their primary dentition had a very different

caries trajectory in their permanent dentition compared to their caries free

contemporaries. In light of these results, caries free and caries active children

should be considered as two separate populations, suggesting different

prevention strategies are required to address their different risk profiles.

2

Introduction Dental caries was confirmed by the WHO global review on oral health as the

single most important oral condition world-wide (Petersen et al. 2005).

Although caries prevalence is falling in many western countries, the disease

still affects a significant proportion of children. In the United States, NHANES

(2011-2012) reported that over 50% of children aged 6-18 years presented

with caries (Dye et al. 2015). In England, the Children’s Dental Health Survey

(2013) (Pitts et al. 2015) reported 40% of children aged 8 years had caries in

the primary teeth and 10% had caries in their permanent teeth. At age 12

years, prevalence of caries in the permanent teeth had increased to 26% and

to 39% in 15-year-olds. Oral health disparities by ethnicity (Dye et al. 2015)

and deprivation (Pitts et al. 2015) are well established and can persist

throughout the life course (Poulton et al. 2002). The consequences of caries

for children includes pain, swelling, problems eating and communicating,

impact on quality of life; and costs to society related to time off school;

reduced productivity and resources needed to treat the condition

(Casamassimo et al. 2009; Cunnion et al. 2010; Petersen 2005).

To develop effective prevention strategies requires an understanding of how

caries develops and progresses over time, but there have been few, large

longitudinal cohort studies completed in children. Traditionally, observational

caries studies have been cross-sectional in design, which can be suggestive

of risk factors for disease, but are limited in terms of establishing relationships

between exposure and outcome over time. In particular, there have been few

longitudinal cohort studies bridging the transitional period between childhood

and adolescence. This is an important period in the life course in terms of

dental development, with the emergence of the permanent teeth, and social

development as children become more independent.

The primary aim of this study was to describe the prevalence and incidence of

caries in the permanent teeth in a cohort of children over a seven-year period,

with a view to establishing the pattern of caries development. The secondary

aim was to explore the role of potential risk factors (age, gender, ethnicity and

social deprivation) associated with caries. The disease trajectory of the

3

groups defined by the presence or absence of caries (into dentine) in the

primary teeth as a risk factor for caries (into dentine) in the permanent teeth

was of particular interest.

4

Materials and MethodsThis study aimed to clinically examine and follow up all school Year 3 and 4

pupils (aged 7 to 9) who attended any of the 207 state-funded primary schools

in East Lancashire in the North-West of England in February 2006. The

location was chosen for its comparatively high caries prevalence and the

absence of a fluoridated water program (Dental Observatory 2001; 2006). The

study was accepted for inclusion on the NIHR UK Clinical Research Network

Study Portfolio (Ref: 10315) and granted ethical approval by the National

Research Ethics Service (Ref: 11/NE/0006). The study is a human

observational study and conformed to the STROBE guidelines.

Following the invitation to participate, consent was obtained initially at the

school level, followed by parental and child consent. There were four waves of

data collection in the study, bridging the primary and secondary school

systems. Study consent was taken for waves one and two. For waves three

and four consent was separated into consent for a questionnaire of self-

reported risk factors and consent to the intra-oral clinical examination. For the

purpose of this study, only clinical examination data were analyzed. After the

first baseline clinical examination (CE1) at 7-9 years of age, three more

follow-up examinations were conducted when the same cohort reached mean

ages of 11 (CE2), 13 (CE3) and 15 (CE4) years. CE1 and CE2 were carried

out in primary schools; CE3 and CE4 were carried out in secondary (high)

schools.

Schools with less than 10 children in the age group under study were

excluded. All pupils registered in consenting schools in the age group under

study were invited to participate. Throughout the study we followed current

National Health Service guidance on consenting children for observational

studies (Rooney et al. 2010). Before each clinical examination, parents and

potential participants were sent Participant Information Sheets explaining the

study and consent process. In primary schools, participants who did not opt-

out and presented for examination were considered to have provided consent.

In secondary schools, while participants could opt-out, those who presented

were asked to give informed consent.

5

The school provided each pupil’s date of birth, gender and home postcode.

The postcode (geographic area with approximately 15 households) was used

to link to a small area measure of socio-economic position; the Index of

Material Deprivation quintiles 2010 (IMDQ) (IMD, 2010) (Mukund 2011; NPEU

2013). The Index of Multiple Deprivation, commonly known as the IMD, is the

official measure of relative deprivation for small geographical areas in

England. It combines information from seven domains including income,

employment, crime and health. It can be used to rank every small area in

England in quintiles from IMDQ1 (least deprived areas) to IMDQ5 (most

deprived areas). Ethnicity was reported by the parent according to nationally

agreed categories (white, Asian, black, Chinese, Mixed or other) (BASCD

2009). The first recorded gender; ethnicity and home postcode data were

used in the analyses.

At each clinical examination, caries was assessed using a UK national

diagnostic protocol by trained and calibrated dentist examiners (BASCD 2009;

Pine et al. 1997). A dentist recognized nationally as a “gold standard”

provided training and acted as the standard for calibration. The diagnostic

threshold for caries was visual assessment only of frank cavitation into

dentine (d3, D3). Calibration ranges were sensitivity 1.00 to 0.82 and specificity

1.00 to 0.99. Reproducibility for clinical caries scores was assessed using the

Kappa statistic with a re-examination of 10% of participants. Intra-examiner

Kappa was above 0.81 at each time point.

Data were analyzed with STATA 14. A descriptive analysis of pupils’ socio-

demographic characteristics (age, gender, ethnicity and social deprivation)

was carried. The characteristics of consenting and non-consenting pupils

were compared to assess representitivity of the pupils under study and the

potential impact of loss to follow-up.

The number of decayed (d/D), missing (m/M) and filled (f/F) teeth was

summed to calculate individual whole mouth d3mft values for primary and

D3MFT values for permanent dentition separately. D3MFT was recorded for

6

each individual at all four time points and d3mft was recorded for CE1 and

CE2 only. Using these two summary scores separately, each pupil was further

categorized into caries free (d3mft=0, D3MFT=0) and caries active (d3mft>0,

D3MFT>0) groups in order to calculate caries prevalence (percentage of

pupils caries active) and for subgroup analysis.

Caries incidence in the permanent dentition was explored using a population-

averaged (marginal) model with covariates of age, gender, deprivation,

ethnicity and caries absence or presence in the primary dentition. A negative

binomial regression using generalized estimating equations with mean

dispersion variance, unstructured correlation structure, and Huber White

sandwich estimator for standard errors was proposed as a suitable model to

accommodate the over-dispersion and longitudinal nature of the D3MFT count

data. A number of different models were investigated (See Appendix ‘Model

selection’ and Appendix Figure 2). Model fit was assessed through the visual

inspection of residuals, values of information criteria and deviance.

7

ResultsAt CE1 and CE2, 92% (190/207) of primary schools agreed to participate in

the study: Seven schools declined, three were due to close and seven had

fewer than ten pupils in the eligible age-group (See Figure 1). At CE3, 75%

(27/36) of secondary schools and at CE4 67% (24/36) of secondary schools

agreed to participate in the study. One school had closed and the remainder

cited concerns about disruption to academic activities as reasons for declining

to participate.

Figure 1 Study flowchart

The number of pupils providing study consent and consent to intra-oral

examination (CE3 and CE4 only) is summarized in Figure 1. At CE1: 518

pupils were absent from school on the day of examination or had left the

school before the clinical examination and therefore consent could not be

sought; 5470 of 5649 pupils who were present on the day of examination

consented (96.8%) and 179 withheld consent (118 parents and 61 pupils). At

CE2: 908 pupils were absent from school on the day of examination or had

left the school before the clinical examination; 5476 of 5618 pupils (97.5%)

and 142 withheld consent (83 parents and 59 children).

At CE3: 437 pupils were absent from school on the day of examination; 3443

of 3777 pupils (91.2%) consented to the study, with 69 parents and 265

children withholding consent; 3036 (80.4%) consented to the intra-oral exam,

with 69 parents and 672 children withholding consent for this aspect of the

study. At CE4: 756 pupils were absent from school on the day of examination;

3190 of 3295 pupils (96.8%) consented to the study, with 24 parents and 81

children withholding consent; 2733 (82.9%) consented to the intra-oral exam,

with 24 parents and 538 children withholding consent.

The socio-demographic profile of participants is shown in the Appendix Table

1. Due to the small numbers of pupils in some of the ethnic categories the

8

sparsely populated categories of Black (0.18%), Chinese (0.11%), Mixed or

other (0.99%) were combined with the Asian category (22.8%) to form a

binary variable of White (75.9%) and Asian or other (reference category)

(24.1%). A significant number of participants were from more deprived

backgrounds with over 60% of pupils from IMD quintiles 4 and 5.

The socio-demographic characteristics of consenting pupils were similar to

non-consenting pupils (Appendix Table 2). Mean age in years was very

similar for the consented and non-consented groups. For those children who

consented to an intra oral examination, there were slightly more males than

females at each data point. This was reversed in the non-consenting groups.

There were a higher proportion of children recording their ethnicity as ‘White’

in the non-consenting group than in the consenting group.

In line with our secondary aim of exploring the disease trajectory of the groups

defined by the presence or absence of caries (into dentine) in the primary

teeth as a risk factor for caries (into dentine) in the permanent teeth the socio-

demographic characteristics are presented by subgroup in Appendix Table 3.

In total 6651 pupils provided data from at least one clinical examination.

Caries prevalence for primary teeth (d3mft) at CE1 was 73.3% and 49.8% at

CE2; this lower prevalence was due to exfoliated primary teeth. Over the

seven-year period of study, caries prevalence for permanent teeth (D3MFT)

increased from 16.7% at CE1 to 31.0%, 42.2% and 45.7% at CE2, CE3 and

CE4 respectively (Table 1). Caries prevalence increased over time for every

quintile of deprivation (Appendix Figure 1).

The proportion of participants with caries in permanent teeth at every time

point was higher for those who had caries in their primary teeth at baseline

compared with those who were caries-free in their primary teeth at baseline;

and they had a higher mean D3MFT. At CE4, 22.1% of those who were caries

9

free in their primary dentition at baseline had developed caries in their

permanent dentition (mean D3MFT = 0.38, range 0 to 6, n = 606) compared

with 56.1% of those who had caries at baseline (mean D3MFT = 1.78, range 0

to 13, n = 1523). Figure 2 shows that those that are caries free in the primary

dentition at CE1 are more likely to stay as caries free when they reach

adolescence compared to those who have caries initially.

Figure 2 Distribution of D3MFT by d3mft caries status and clinical examination (CE)

Initial covariates for the regression model included: caries status in the

primary dentition at CE1, ethnicity (White or Asian or other), gender and

IMDQ (Appendix 1 Table 1). We investigated an interaction effect between

age and caries status in the primary dentition at CE1 but this was not

statistically significant (p = 0.09).

Table 1 Results from negative binomial using generalized estimating equation with unstructured correlation structure and mean dispersion variance

The regression analysis model was based on 14,535 records from 5048

pupils who contributed a D3MFT value from at least one of the four clinical

examinations, and values for all covariates in the model. All main effect

covariates were found to be statistically significant. Mean predicted values

indicated a rising D3MFT count as pupils aged (consistent with new teeth

emerging). These values were 4.5 times higher (95% CI 3.90 to 5.16) in those

pupils with caries in their primary dentition at CE1 than in those without. This

separation of the two trajectories can clearly be seen in Figure 3 where for

illustrative purposes covariates for gender, deprivation and ethnicity have

been fixed at Male, White and most deprived IMDQ. From the adjusted model

(Table 1) White, and Female are associated with higher D3MFT count; D3MFT

count is highly statistically significant with IMDQ (p<0.001) and increased with

increasing levels of deprivation as indicated by IMDQ.

10

Figure 3 Estimated mean D3MFT for Male, White, Index of Material Deprivation quintile 5 (most deprived) subgroup by Clinical Examination 1 d3mft caries status

In Figure 3, for illustrative purposes covariates for gender, ethnicity and

deprivation have been fixed at categories of Male, White and most deprived

IMDQ. Figures corresponding to other combinations of categories can be

plotting using the coefficients provided in Table 1.

11

DiscussionThe main purpose of the study was to describe the development of caries

over seven years in a cohort of children from ages 7 to 16 years. It is the first

of its type in the United Kingdom and one of the largest and longest duration

internationally. It is clear from these results that even in England, which has

the lowest caries rates for 12-year-olds in Europe (Malmo University 2016),

dental caries in disadvantaged areas of the country continues to be a major

public health problem, affecting over 50% of this study population. The

findings support previous studies that caries is significantly associated with

deprivation but that the disparity widens with age with a negative impact on

quality of individual and family life (Ravaghi et al. 2016). A key finding of our

study was the significant association between caries in the primary teeth

(ages 7-9 years) and caries in the permanent teeth. Our data show that the

development of caries in the primary dentition means that children embark on

a very steep and different disease trajectory in their permanent dentition to

children whose primary dentition is caries free.

The main strength of this study was the number of participants recruited and

followed up over a 7-year period. This large study with high quality continuous

data provides information on caries development and progression during a

period of the life course that we know little about. This is an important age

group as children transition from the primary to the permanent dentition and

start to become more independent. The change in social environment

experienced in these years will influence caries risk and have lifelong impact

on oral health. Furthermore, as this was a school-based study, eligibility was

not restricted to children and adolescents registered with or attenders of

general dental practices. This information provided by this whole population

approach improves our understanding of caries development and progression

in underserved populations.

The populations at baseline and follow-up were similar in terms of socio-

demographic values (gender, IMDQ and ethnicity) suggesting loss to follow up

had limited impact on the findings. The socio-demographic profile of the study

cohort was similar to the reference population of the region, with a significant

12

sub-population from an Asian ethnic background and high levels of

deprivation, both associated with increased caries prevalence, which explains

why the caries prevalence in the cohort was higher than national norms (Pitts

et al. 2015). A weakness of the study was loss to follow-up, particularly

between CE2 and CE3. The main reason for loss to follow up was that over

25 per cent of the secondary schools, accounting for more than 40% of pupils

in the study locality did not consent to take part. These schools refusing

access represented the larger schools in the area. Milsom et al (2011)

reported that epidemiological studies in the region showed that children from

larger schools had higher caries prevalence and levels of deprivation

(participation bias) so caries incidence and prevalence may have been under-

estimated in this study.

The caries assessment method used diagnosed caries at a late stage; caries

into dentine. This decision was taken as the study was designed to take a

public health perspective. Cavitated caries requires treatment and therefore

has implications for children and their parents and the healthcare system in

terms of costs and capacity. The caries assessment method we used is the

same method used in local and national disease surveillance programs and

therefore enabled comparison with recent and future cross sectional

epidemiological findings (Pitts et al. 1997). A finer level of caries diagnosis, for

example using the ICDAS stages 2 and 3, would have provided useful

additional information but would have added to the time required to undertake

the assessments and could have compromised recruitment and particularly

retention in secondary schools. Retaining inter-examiner agreement,

particularly in a study of this size, could also pose problems.

The findings of this study have implications for the design of oral health

prevention strategies. Our results show that deprivation remains an important

predictor of future caries, and that children who have caries in their primary

dentition follow a steep disease development trajectory in their permanent

dentition. The key predictor of caries in the permanent teeth was previous

caries experience. Children who have primary dentitions that are caries free

13

and caries active children should be considered as two different populations

for prevention strategies due to their very different risk profiles.

There is a correlation between parental oral health, parental behaviors and

locus of control, which can impact on the oral health risk of young children

(Duijster et al. 2015). Parallel qualitative research on this study population

(need the reference (Hall-Scullin et al. 2015) showed that participants in

adolescence perceived that the example set by their parents in early

childhood was the dominant influence on their oral health. The Dunedin

project (Thomson et al. 2004) reported an association across the life course

between unfavorable dental health beliefs and poorer oral health and

demonstrated that it is possible to change the behaviors between

adolescence and adulthood but these behavioral changes can be unstable.

The study suggests that prevention of caries in early childhood is critically

important to prevent children embarking on a ‘fast train’ disease trajectory,

which will have lifelong consequences, but the evidence base for effective

interventions to instill risk-reducing behavior in very young children is lacking.

This observational study raised two significant research questions; ‘how do

we keep very young children caries free?’ and secondly ‘once a child

develops caries how to we slow down and stop the process?’ The same

interventions may not provide the answers to both questions. For example the

assessment of caries in the primary teeth (caries free or not) could be used to

identify at-risk children for a targeted caries prevention program such as

fissure sealants as soon as permanent teeth erupt.

The findings of this study illustrate the significant long-term consequences of

developing caries in early childhood. The widening of the social gradient of

disease with age also emphasizes the need for dental public health

improvement strategies to address the social determinants of dental disease.

AcknowledgementsThe authors would like to thank the staff and participants at schools in East

Lancashire and Blackburn with Darwen. We also wish to acknowledge the

14

Salaried Dental Service, Lancashire Care NHS Foundation Trust led by

Catherine Morley; Joe Fielding, Uriana Boye, Angela Willesey and Sarah

Procter who acted as Gold Standard examiners for the calibration and training

events; and Janet Neville at The Dental Observatory. This work was

supported by Colgate-Palmolive. Support cost funding was obtained from

Cumbria and Lancashire Comprehensive Local Research Network (CLCLRN).

The authors declare there are no competing interests.

15

ReferencesBASCD. 2009. Oral health survey of 12 year old children in England

2008/2009. National protocol. NHS Dental Epidemiology Programme.

Casamassimo PS, Thikkurissy S, Edelstein BL, Maiorini E. 2009. Beyond the

dmft: The human and economic cost of early childhood caries. Journal

of the American Dental Association. 140(6):650-657.

Cunnion DT, Spiro A, 3rd, Jones JA, Rich SE, Papageorgiou CP, Tate A,

Casamassimo P, Hayes C, Garcia RI. 2010. Pediatric oral health-

related quality of life improvement after treatment of early childhood

caries: A prospective multisite study. Journal of Dentistry for Children

(Chicago, Ill ). 77(1):4-11[abstract only].

Regional dental health 12 year olds 2000/01(England). 2001. [accessed

04/10/2014]. http://www.dental-observatory.nhs.uk/Default.aspx?

crumb=00020007015710E000BB.

Regional dental health 5 year olds 2005/06 (England). 2006. [accessed

04/10/2014]. http://www.dental-observatory.nhs.uk/Default.aspx?

crumb=00020007015610E000BB.

Duijster D, de Jong-Lenters M, Verrips E, van Loveren C. 2015. Establishing

oral health promoting behaviours in children – parents’ views on

barriers, facilitators and professional support: A qualitative study. BMC

Oral Health. 15:157.

Dye BA, Thornton-Evans G, Li X, Iafolla TJ. 2015. Dental caries and sealant

prevalence in children and adolescents in the United States, 2011–

2012. CDC National Centre for Health Statistics.

Hall-Scullin E, Goldthorpe J, Milsom K, Tickle M. 2015. A qualitative study of

the views of adolescents on their caries risk and prevention

behaviours. BMC Oral Health. 15(1):141.

16

Malmo University. 2016. Country oral health profiles who regions table

European Union and European Economic Area (Euro incl. DMFT for

12-year-olds).

Mukund L. 2011. The English Indices of Deprivation 2010. In: Government

CaL, editor.: Department for Communities and Local Government.

NPEU. 2013. IMD postcode search tool. University of Oxford.

Petersen PE. 2005. The burden of oral disease: Challenges to improving oral

health in the 21st century. Bulletin of the World Health Organization.

83(1):3.

Petersen PE, Bourgeois D, Ogawa H, Estupinan-Day S, Ndiaye C. 2005. The

global burden of oral diseases and risks to oral health. Bulletin of the

World Health Organization. 83(9):661-669.

Pine C, Pitts N, Nugent Z. 1997. British Association for the Study of

Community Dentistry (BASCD) guidance on the statistical aspects of

training and calibration of examiners for surveys of child dental health.

A BASCD coordinated dental epidemiology programme quality

standard. Community Dental Health. 14:18.

Pitts N, Chadwick B, Anderson T. 2015. Report 2: Dental disease and

damage in children England, Wales and Northern Ireland. Children’s

Dental Health Survey 2013. Online: National Statistics Office.

Pitts N, Evans D, Pine C. 1997. British Association for the Study of

Community Dentistry (BASCD) diagnostic criteria for caries prevalence

surveys-1996/97. Community Dent Health. 14(Suppl 1):6 - 9.

Poulton R, Caspi A, Milne BJ, Thomson WM, Taylor A, Sears MR, Moffitt TE.

2002. Association between children's experience of socioeconomic

disadvantage and adult health: A life-course study. The Lancet.

360(9346):1640-1645.

17

Ravaghi V, Holmes RD, Steele JG, Tsakos G. 2016. The impact of oral

conditions on children in England, Wales and Northern Ireland 2013. Br

Dent J. 221(4):173-178.

Rooney E, Davies G, Neville J, Robinson M, Perkins C, Bellis M. 2010. NHS

dental epidemiology programme for England oral health survey of 12

year old children 2008/2009. Summary of caries prevalence and

severity reaults.

Thomson WM, Poulton R, Milne BJ, Caspi A, Broughton JR, Ayers KMS.

2004. Socioeconomic inequalities in oral health in childhood and

adulthood in a birth cohort. Community Dentistry & Oral Epidemiology.

32(5):345-353.

18

List of 1 table and 3 figures

Figure 1 Study flowchart

Figure 2 Distribution of D3MFT by d3mft caries status and clinical examination (CE)

Figure 3 Estimated mean D3MFT for Male, White, Index of Material Deprivation quintile 5 (most deprived) subgroup by Clinical Examination 1 d3mft caries status

Table 1 Results from negative binomial using generalized estimating equation with unstructured correlation structure and mean dispersion variance

19

Figure 1 Study flowchart

20

21

Figure 2 Distribution of D3MFT by d3mft caries status and clinical examination (CE)

22

Figure 3 Estimated mean D3MFT for Male, White, Index of Material Deprivation quintile 5 (most deprived) subgroup by Clinical Examination 1 d3mft caries status

Footnote: Illustrative estimated mean D3MFT values for male pupils of self-

reported White ethnicity living in an area of greatest deprivation. Markers (+)

indicate mean observed values at CE1 through CE4.

23

Table 1 Results from negative binomial using generalized estimating equation with unstructured correlation structure and mean dispersion variance

Exponentiated coefficient

Semi robust standard

error§

95% CI

Age 1.27*** 0.01 1.26, 1.28Caries free (d3mft = 0) 1Caries active (d3mft >0) 4.49*** 0.32 3.90, 5.16Asian or other 1White 1.24*** 0.07 1.12, 1.38Male 1Female 1.11* 0.05 1.02, 1.20IMDQ 1 §§ 1IMDQ 2 1.06 0.13 0.83, 1.36IMDQ 3 1.34* 0.17 1.04, 1.72IMDQ 4 1.39** 0.17 1.10, 1.77IMDQ 5 1.77*** 0.21 1.41, 2.23Constant 0.01*** 0 0.00, 0.01

§ Huber/White/sandwich estimator of variance used which produces valid standard errors even if the correlations within group are not as hypothesized by the specified correlation structure and accounts for clustering in the data§§ IMDQ p<0.001* p<0.05, ** p<0.01, *** p<0.001

24

Appendix Title

Longitudinal study of caries development from childhood to adolescenceAuthorsMs E. Hall-Scullin1, Consultant in Dental Public Health, NHS Ayrshire & Arran.Miss H. Whitehead2, Dental Professional Lead, Community Dental Service, Salford Royal NHS Foundation Trust.Prof K. Milsom3, Consultant in Dental Public Health, Department of Dental Public Health, Countess of Chester Hospital.Prof M. Tickle4, Professor of Dental Public Health and Primary Care, Division of Dentistry, University of Manchester.Dr T-L Su, Lecturer in Statistics, Division of Dentistry, University of ManchesterDr T. Walsh4, Reader in Biostatistics, Division of Dentistry, University of Manchester.

Address1Public Health Department, NHS Ayrshire & Arran, Afton House, Ailsa Hospital Campus, Dalmellington Road, Ayr, KA6 6AB, UK. 2Community Dental Service, Salford Royal NHS Foundation Trust, Pendleton Gateway, 1 Broadwalk, Salford, M6 5FX, UK. 3Department of Dental Public Health, 1829 Building, Countess of Chester Hospital, Chester, CH2 1UL, UK. 4Division of Dentistry, Faculty of Biology, Medicine and Health, University of Manchester, Oxford Road, Manchester, M13 9PL, UK.

25

First (Baseline) Clinical Examination 1 (n =

5470)Clinical Examination 2 (n = 5475) Clinical Examination 3 (n = 3036) Clinical Examination 4 (n = 2733)

Median Date 19/03/2006 18/02/2009 10/10/2011 16/05/2013

D3MFT = 0

n (%)

D3MFT > 0

n (%)

D3MFT

Mean (sd)

D3MFT = 0

n (%)

D3MFT > 0

n (%)

D3MFT

Mean (sd)

D3MFT = 0

n (%)

D3MFT > 0

n (%)

D3MFT

Mean (sd)

D3MFT = 0

n (%)

D3MFT > 0

n (%)

D3MFT

Mean (sd)

Caries prevalence

(D3MFT) 4559 (83.35) 911 (16.65)

0.28 (0.73)

[D3MFT > 0

1.70 (0.90)]

3780 (69.04) 1695 (30.96)

0.64 (1.20)

[D3MFT > 0

2.08 (1.29)]

1756 (57.84) 1280 (42.16)

1.04 (1.67)

[D3MFT > 0

2.48 (1.75)]

1484 (54.30) 1249 (45.70)

1.36 (2.08)

[D3MFT > 0

2.98 (2.16)]

Caries status

(d3mft) at

CE1

d3mft = 0 1398 (95.69) 63 (4.31) 0.06 (0.34) 1120 (89.24) 135 (10.76) 0.16 (0.54) 520 (80.12) 129 (19.88) 0.34 (0.80) 472 (77.89) 134 (22.11) 0.38 (0.85)

d3mft > 0 3161 (78.85) 848 (21.15) 0.36 (0.82) 2140 (61.94) 1315 (38.06) 0.81 (1.31) 828 (48.06) 895 (51.94) 1.32 (1.80) 668 (43.86) 855 (56.14) 1.78 (2.27)

Age (years) (SD) 8.04 (0.30) 8.08 (0.31) 10.96 (0.30) 10.97 (0.30) 13.61 (0.34) 13.61 (0.35) 15.20 (0.34) 15.18 (0.34)

Gender Male 2404 (84.29) 448 (15.71) 0.26 (0.70) 2001 (70.31) 845 (29.69) 0.62 (1.18) 917 (59.47) 625 (40.53) 0.97 (1.61) 774 (55.44) 622 (44.56) 1.36 (2.10)

Female 2155 (82.31) 463 (17.69) 0.31 (0.77) 1779 (67.67) 850 (32.33) 0.67 (1.22) 839 (56.16) 655 (43.84) 1.10 ( 1.73) 710 (53.10) 627 (46.90) 1.37 (2.06)

Ethnicity Asian or

other1074 (81.43) 245 (18.57) 0.31 (0.75) 915 (66.64) 458 (33.36) 0.69 (1.21) 475 (55.62) 379 (44.38) 1.11 (1.70) 359 (52.87) 320 (47.13) 1.41 (2.07)

White3484 (83.95) 666 (16.05) 0.28 (0.73) 2849 (69.86) 1229 (30.14) 0.62 (1.19) 1278 (58.76) 897 (41.24) 1.02 (1.65) 1119 (54.83) 922 (45.17) 1.34 (2.08)

Index of

Material

Deprivation

quintiles

2010

(IMDQ)

IMDQ 1 302 (89.61) 35 (10.39) 0.16 (0.57) 279 (80.87) 66 (19.13) 0.37 (0.91) 128 (71.11) 52 (28.89) 0.54 (1.08) 123 (72.78) 46 (27.22) 0.62 (1.32)

IMDQ 2 830 (88.39) 109 (11.61) 0.19 (0.58) 712 (76.48) 219 (23.52) 0.44 (0.98) 360 (69.23) 160 (30.77) 0.63 (1.17) 311 (62.08) 190 (37.92) 0.95 (1.72)

IMDQ 3 503 (84.11) 95 (15.89) 0.28 (0.71) 435 (72.50) 165 (27.50) 0.54 (1.11) 197 (54.27) 166 (45.73) 1.11 (1.68) 176 (54.66) 146 (45.34) 1.29 (2.02)

IMDQ 4 944 (84.59) 172 (15.41) 0.26 (0.70) 760 (69.66) 331 (30.34) 0.59 (1.10) 355 (58.29) 254 (41.71) 1.04 (1.66) 297 (54.50) 248 (45.50) 1.33 (1.97)

IMDQ 5 1933 (79.91) 486 (20.09) 0.35 (0.81) 1559 (63.79) 885 (36.21) 0.80 (1.35) 702 (52.70) 630 (47.30) 1.25 (1.85) 547 (47.94) 594 (52.06) 1.68 (2.30)

Appendix Table 1 Per cent (%) caries free (D3MFT=0) and caries active (D3MFT>0) with mean D3MFT and standard deviation (sd)by clinical and socio-demographic variables at each clinical examination (CE)

26

Clinical Examination (CE)

CE1 N (%) CE2 N (%) CE3 N (%) CE4 N (%)

No consent Consent No

consent Consent No consent Consent No

consent Consent

Age (years)

Mean age (sd) N/A 8.0 (0.3) N/A 11.0

(0.3)13.6 (0.4)

13.6 (0.3)

15.2 (0.4)

15.2 (0.3)

Gender

Male 83 (46.4)

2852 (52.1)

58 (40.8)

2846 (52.0)

365 (49.3)

1542 (50.8)

250 (44.5)

1396 (51.1)

Female 96 (53.6)

2618 (47.9)

84 (59.2)

2630 (48.0)

376 (50.7)

1494 (49.2)

312 (55.5)

1337 (48.9)

Ethnicity

Asian or other

10 (5.7)

1319 (24.1)

14 (9.9)

1373 (25.2)

119 (16.6)

854 (28.2)

71 (12.8)

679 (25.0)

White 164 (94.3)

4150 (75.9)

128 (90.1)

4079 (74.8)

596 (83.4)

2175 (71.8)

483 (87.2)

2041 (75.0)

Index of Material Deprivation quintiles 2010 (IMDQ)

IMDQ 1 19 (10.7)

337 (6.2)

11 (8.0)

345 (6.4)

30 (4.1)

180 (6.0)

12 (2.2)

169 (6.3)

IMDQ 2 41 (23.0)

939 (17.4)

32 (23.3)

931 (17.2)

114 (15.7)

520 (17.3)

74 (13.8)

501 (18.7)

IMDQ 3 23 (12.9)

598 (11.1)

16 (11.7)

601 (11.1)

81 (11.20

363 (12.1)

80 (14.9)

322 (12.0)

IMDQ 4 40 (22.5)

1116 (20.6)

41 (29.9)

1091 (20.2)

152 (21.0)

609 (20.3)

116 (21.6)

545 (20.4)

IMDQ 5 55 (30.9)

2419 (44.7)

37 (27.0)

2444 (45.2)

348 (48.0)

1332 (44.3)

256 (47.6)

1141 (42.6)

Appendix Table 2 Socio-demographic characteristics of consenting and non-consenting pupils

27

First (Baseline) Clinical

Examination

N (%)

d3mft =0 d3mft > 0

Age (years)

Mean age (sd)8.0

(0.29)1461

8.1 (0.30)4009

Gender

Male 83 (46.4)

2852 (52.1)

Female 96 (53.6)

2618 (47.9)

Ethnicity

Asian or other 177 (12.1)

1142(28.5)

White 1284 (87.9)

2866 (71.5)

Index of Material Deprivation quintiles 2010 (IMDQ)

IMDQ 1 157(10.8)

180(4.55)

IMDQ 2 360(24.7)

579(14.6)

IMDQ 3 190(13.1)

408(10.3)

IMDQ 4 279(19.2)

837(21.6)

IMDQ 5 468(32.2)

1951(49.3)

Appendix Table 3 Characteristics of the participants by caries status (d3mft) at first (baseline) clinical examination (CE1)

28

Appendix Figure 1 Graph of caries prevalence (D3MFT) at each data point for each Index of Material Deprivation quintiles (IMDQ)

29

Model selectionVarious modeling approaches were explored including Poisson regression

and Negative Binominal regression (with either mean dispersion variance

(which corresponds to Cameron and Trivedi’s NB2 variance function) or

constant dispersion variance (which corresponds to the NB1 variance

function). All approaches took into account the longitudinal structure of the

data using Generalized Estimating Equations with either independent,

unstructured or exchangeable correlation structures).

Initial covariates for the regression models included all main effects: caries

status in the primary dentition at first (baseline) clinical examination (CE1),

ethnicity (White or Asian or other), gender and IMDQ. We investigated an

interaction effect between age and caries status in the primary dentition at

CE1 but this was not statistically significant (p = 0.09). For improved model fit

and ease of interpretation we removed the interaction term from a final model,

leaving only main effects.

There was statistically significant evidence of over dispersion (mean greater

than variance) in the data, with the estimated dispersion parameter alpha =

1.811 statistically significantly away from zero and indicating that a mean

dispersion variance function was appropriate. Alongside Appendix Figure 2

this provided visual evidence that Negative Binominal demonstrated a

sensible model fit to the data that was superior to the Poisson regression. We

therefore opted for Negative Binominal with Generalized Estimating Equation

as a sensible analysis strategy.

30

Appendix Figure 2 Observed and predicted values from Poisson and Negative Binomial models

31