Measuring the Certainty of a Decision Rule with Applications in Electrocardiography

As an extension of recent work, several types of uncertainty involved in a decision process are further clarified in this paper. The concept of allocation with ß-confidence and the ß-doubt matrix are illustrated by two electrocardiographic data sets. These data sets illustrate the importance and the novelty of the above concepts. We argue for augmenting the classical correct classification rate with an interval estimate. We think that interval estimates are essential in the area of prediction modelling to refrain the user from being too optimistic.

A randomized placebo controlled trial on the effects of simvastatin, a HMG-CoA reductase inhibitor, on blood lipids and fibrinolytic parameters

Coronary artery disease is frequently associated with disturbed blood lipids and with a deficient blood fibrinolytic capacity. In order to investigate a possible link between hypercholesterolemia and hypofibrinolysis, we have investigated the effect of simvastatin, a HMG-CoA reductase inhibitor, on blood lipids and fibrinolytic parameters in a double blind, randomized, placebo controlled study design. Twenty-four male patients, aged between 42 and 65 years, with angiographically confirmed coronary artery disease, were selected from a series of 731 consecutive patients on the basis of a fasting serum cholesterol level of more than 250 mg/dl and a plasma PAI-1 level of more than 60 ng/ml. Patients were randomly assigned to 20 mg of simvastatin daily (Group I), or placebo (Group II), for four weeks, followed by doubling of the dose for another four weeks. Blood samples were obtained at baseline and at 4 and 8 weeks. Groups I and II did not differ significantly at baseline. As expected, simvastatin produced a significant reduction in serum levels of total cholesterol (33 +/- 12 and 36 +/- 12 percent, mean +/- SD, after 4 and 8 weeks respectively, p less than 0.001 vs baseline), LDL-cholesterol (36 +/- 5 and 43 +/- 6 percent respectively, p less than 0.001 vs baseline) and apolipoprotein-B (20 +/- 29 and 23 +/- 30 percent respectively, p less than 0.05 vs baseline), whereas these parameters did not change significantly in the placebo group.(ABSTRACT TRUNCATED AT 250 WORDS)

Combined dynamic predictions using joint models of two longitudinal outcomes and competing risk data

Nowadays there is an increased medical interest in personalized medicine and tailoring decision making to the needs of individual patients. Within this context our developments are motivated from a Dutch study at the Cardio-Thoracic Surgery Department of the Erasmus Medical Center, consisting of patients who received a human tissue valve in aortic position and who were thereafter monitored echocardiographically. Our aim is to utilize the available follow-up measurements of the current patients to produce dynamically updated predictions of both survival and freedom from re-intervention for future patients. In this paper, we propose to jointly model multiple longitudinal measurements combined with competing risk survival outcomes and derive the dynamically updated cumulative incidence functions. Moreover, we investigate whether different features of the longitudinal processes would change significantly the prediction for the events of interest by considering different types of association structures, such as time-dependent trajectory slopes and time-dependent cumulative effects. Our final contribution focuses on optimizing the quality of the derived predictions. In particular, instead of choosing one final model over a list of candidate models which ignores model uncertainty, we propose to suitably combine predictions from all considered models using Bayesian model averaging.

A steep ramp test is valid for estimating maximal power and oxygen uptake during a standard ramp test in type 2 diabetes

A short maximal steep ramp test (SRT, 25 W/10 s) has been proposed to guide exercise interventions in type 2 diabetes, but requires validation. This study aims to (a) determine the relationship between Wmax and V˙O2peak reached during SRT and the standard ramp test (RT); (b) obtain test-retest reliability; and (c) document electrocardiogram (ECG) abnormalities during SRT. Type 2 diabetes patients (35 men, 26 women) performed a cycle ergometer-based RT (women 1.2; men 1.8 W/6 s) and SRT on separate days. A random subgroup (n = 42) repeated the SRT. ECG, heart rate, and V˙O2 were monitored. Wmax during RT: 193 ± 63 (men) and 106 ± 33 W (women). Wmax during SRT: 193 ± 63 (men) and 188 ± 55 W (women). The relationship between RT and SRT was described by men RT V˙O2peak (mL/min) = 152 + 7.67 × Wmax SRT1 (r: 0.859); women RT V ˙ O 2 p e a k (mL/min) = 603 + 4.75 × Wmax SRT1 (r: 0.771); intraclass correlation coefficients between first (SRT1) and second SRT Wmax (SRT2) were men 0.951 [95% confidence interval (CI) 0.899-0.977] and women 0.908 (95% CI 0.727-0.971). No adverse events were noted during any of the exercise tests. This validation study indicates that the SRT is a low-risk, accurate, and reliable test to estimate maximal aerobic capacity during the RT to design exercise interventions in type 2 diabetes patients.

Is tightly controlled disease activity possible with online patient-reported outcomes?

OBJECTIVE

To evaluate the performance of patient-reported outcomes (PRO) as primary indices for identification and prediction of a 28-joint Disease Activity Score (DAS28)>3.2 among patients with rheumatoid arthritis (RA).

METHODS

Patients with RA completed monthly online PRO [Health Assessment Questionnaire (HAQ), Rheumatoid Arthritis Disease Activity Index (RADAI), visual analog scale (VAS) fatigue] and were clinically assessed every 3 months using the DAS28. Simple descriptive statistics, logistic regression, and the Bayesian joint modeling approach were used to analyze the data. The Bayesian joint model combines the scores and changes in the scores of 3 PRO to predict a DAS28>3.2 at the subsequent timepoint.

RESULTS

A group of 159 patients with RA participated. Stratified summaries of the PRO by DAS28 categories at baseline provided incremental values of the PRO for more active disease. However, on an individual level, the DAS28 and the PRO fluctuated over time. The prediction of subsequent DAS score by a single instrument at single timepoints resulted in moderate sensitivity and specificity. Using the intercept and slope of the combined PRO of the first 3 measurements to predict the DAS28 state at 3 months resulted in a sensitivity of 0.81 and a specificity of 0.92. After 10-fold cross validation, the model had a sensitivity of 0.61 and specificity of 0.75 to identify patients with a DAS28>3.2.

CONCLUSION

PRO showed fluctuating levels of disease activity over time, while on a group level disease activity stayed the same. Using the changes in RADAI, HAQ, and VAS fatigue over time to predict future DAS28>3.2 resulted in moderate performance after the internal cross-validation of the model (sensitivity 0.61, specificity 0.75).

The history of Belgian assisted reproduction technology cycle registration and control: a case study in reducing the incidence of multiple pregnancy

STUDY QUESTION

What is the effect of a legal limitation of the number of embryos that can be transferred in an assisted reproductive technology (ART) cycle on the multiple delivery rate?

SUMMARY ANSWER

The Belgian national register shows that the introduction of reimbursement of ART laboratory costs in July 2003, and the imposition of a legal limitation of the number of embryos transferred in the same year, were associated with a >50% reduction of the multiple pregnancy rate from 27 to 11% between 2003 and the last assessment in 2010, without any reduction of the pregnancy rate per cycle.

WHAT IS KNOWN ALREADY

Individual Belgian IVF centres have published their results since the implementation of the law, and these show a decrease in the multiple pregnancy rate on a centre by centre basis. However, the overall national picture remains unpublished.

STUDY DESIGN, SIZE, DURATION

Cohort study from 1990 to 2010 of all ART cycles in Belgium (2685 cycles in 1990 evolving to 19 110 cycles in 2010), with a retrospective analysis from 1990 to 2000 and prospective online data collection since 2001.

PARTICIPANTS/MATERIALS, SETTING, METHODS

Registration evolved from paper written reports per centre to a compulsory online registration of all ART cycles. From 2001 up to mid-2009, data were collected from Excel spread sheets or MS Access files into an MS Access database. Since mid-2009, data collection is done via a remote and secured web-based system (www.belrap.be) where centres can upload their data and get immediate feedback about missing data, errors and inconsistencies.

MAIN RESULTS AND THE ROLE OF CHANCE

National Belgian registration data show that reimbursement of IVF laboratory costs in July 2003, coupled to a legal limitation in the number of embryos transferred in utero, were associated with a 50% reduction of the multiple pregnancy rate from 27 to 11% without reduction of the pregnancy rate per cycle, and with an increase in the number of fresh and frozen ART cycles due to improved access to treatment.

LIMITATIONS, REASONS FOR CAUTION

There is potential underreporting of complications of ART treatment, pregnancy outcome and neonatal health.

WIDER IMPLICATIONS OF THE FINDINGS

Over the 20 years of registration, the pregnancy rate has remained constant, despite the reduction in the number of embryos transferred, optimization of laboratory procedures and stimulation protocols, introduction of quality systems and implementation of the EU Tissue Directive over the period 2004-2010.

STUDY FUNDING/COMPETING INTEREST(S)

No external funding was sought for this study. None of the authors has any conflict of interest to declare.

An application of Bayesian growth mixture modelling to estimate infection incidences from repeated serological tests

Diagnoses of infectious diseases are often performed using antibody detection through enzyme-linked immunosorbent assay techniques. These data are usually dichotomized into positive and negative samples using a fixed cut-off and prevalences of infection are subsequently estimated assuming perfect correspondence between the dichotomized test results and infection status. In contrast to this approach, in this case study, we estimate the effect of distributing insecticide impregnated bednets to prevent Leishmania infection through mixture modelling of the original continuous data. We analyze the data from a cluster randomized intervention trial using a generalized latent variable model consisting of a longitudinal mixture model for the observed outcome and a Hidden Markov model for the underlying unobserved disease status to estimate the effect of an intervention. The response and structural models are jointly estimated in a Bayesian framework. This model has the advantage that it avoids the need to choose an arbitrary cut-off and allows for uncertainty in the infection status. In this paper, we describe the development of the model and selection of priors, the application to the motivating data, model checking and simulation results.

Suspected bacterial endophthalmitis following intravitreal anti-VEGF injection: case series and literature review

AIM

To report an up-to-date overview of all patients reported in the literature with suspected bacterial endophthalmitis following anti-VEGF injection. Secondly, to identify specific symptoms and signs to differentiate between infectious and noninfectious endophthalmitis.

METHODS

A Pubmed search retrieved 12 retrospective case series which had included a total of 118 patients with suspected bacterial endophthalmitis after anti-VEGF injection. Data of 15 patients from the Rotterdam Eye Hospital were added. Patients were divided into three groups: those who did not receive intravitreal antibiotics (group A), patients who received intravitreal antibiotics with biopsy-negative cultures (group B) and those with biopsy-positive cultures (group C).

RESULTS

The median time between anti-VEGF injection and presentation with suspected bacterial endophthalmitis was 1 day in group A compared to 3 days in groups B and C. At presentation, patients of group A had a better median visual acuity (logMAR 1.0) compared to those in groups B and C (logMAR 2.1 and 2.5, respectively).

CONCLUSION

This study suggests that patients presenting with a visual acuity of 20/200 (logMAR 1.0) or less and later than 24 h after injection are more likely to have bacterial endophthalmitis. To prevent undertreatment in these patients, the threshold to proceed to vitreous biopsy and empirical intravitreous antibiotics should be low.

Assessment of caries experience in epidemiological surveys: a review

OBJECTIVES

To review aspects of methods for assessing caries experience (CE) in epidemiological surveys.

METHOD

A search of English language literature published between January 2000 and December 2008 was undertaken using 'epidemiology', 'dental caries' and 'assessment' as search terms. Information on criteria for CE assessment, materials and settings, diagnostic threshold, training of examiners and validation of the screening results was extracted from the reports.

RESULTS

Eighty-nine reports met the inclusion criteria. In 9 of the reports (10%) no reference was made to existing standardisation criteria for assessment of CE. Light condition applied (60 reports, 67%) and the use of a probe (60 reports, 67%) were frequently reported. Most reports mentioned that training and calibration of examiners took place, but the outcome of reliability checks were often not presented (48 reports, 54%). Only 28 of the reports (32%) specified that cleaning took place before the examination. Journals with Impact Factor (IF) provided specific information on methods more frequently than journals without. The WHO Basic Methods for Oral Health Surveys were most often applied (52 surveys, 58%). However, deviations from the original description were found especially for measurement and reporting of reliability measurement (24, 46% and 29, 56% respectively), type of probe used (27, 52%) and light condition (16, 31%). All of these hamper the (external) validity of the obtained results.

CONCLUSIONS

There is a clear need for improvement of the reporting and application of methods for assessing CE in epidemiological surveys. A check-list of aspects of methods to be included in reports of surveys assessing CE is proposed by the authors.

An application of Harrell's C-index to PH frailty models

Frailty models are encountered in many medical applications, yet little research has been devoted to develop measures that quantify the predictive ability of these models. In this paper, we elaborate on the concept of the concordance probability to clustered data, resulting in an 'Overall Conditional C-index' or bfC(O, C) and an 'Overall Marginal C-index' or C(O, M) . Both Overall C-indices can be split up into a 'Between Conditional' or C(B, C) and a 'Between Marginal C-index' or C(B, M) and into a 'Within Conditional' or C(W, C) and a 'Within Marginal C-index' or C(W, M) . For PH frailty models of the power variance family, C(W, C) and C(W, M) are equivalent resulting in one 'Within C-index' C(W) . We propose an application of Harrell's C-index to estimate the proposed indices within a likelihood and a Bayesian context and the performances of their point estimates and confidence/credible intervals are compared in an extensive simulation study. This simulation study shows that the point estimates of C(W) and C(B, M) perform good within both a likelihood and Bayesian context but that the point estimates of C(B, C) show less bias for the Bayesian approach than for the likelihood approach. The 95 per cent confidence/credible intervals also possess good coverage properties, given that the point estimates perform good. The performance of the C-indices is evaluated on a real data set.

Bayesian latent variable models for spatially correlated tooth-level binary data in caries research

Analysis of dental caries is traditionally based on aggregated scores, which are summaries of caries experience for each individual. A well-known example of such scores is the decayed, missing and filled teeth or tooth surfaces index introduced in the 1930s. Although these scores have improved our understanding of the pattern of dental caries, there are still some fundamental questions that remain unanswered. As an example, it is well believed among dentists that there are spatial symmetries in the mouth with respect to caries, but this has never been evaluated in a statistical sense. An answer to this question requires the analysis to be performed at subunits within the mouth, which necessitates the use of methods for correlated data. We propose a Bayesian generalized latent variable model coupled with an undirected graphical model to investigate the unique spatial distribution of tooth-level caries outcomes in the mouth. Data from the Signal Tandmobiel(®) study in Flanders, a dental longitudinal survey, are used to illustrate the methodology.

Measuring oral health behaviour in Flemish health care workers: an application of the theory of planned behaviour

OBJECTIVE

This study describes the development and validation of a questionnaire to measure the determinants of oral health related behaviour in health care workers, based on the Theory of Planned Behaviour (TPB).

RESEARCH DESIGN

A preliminary questionnaire was drafted containing 78 self report items measuring three behaviours related to oral health (i.e. dietary habits, oral hygiene habits and dental attendance), as well as the attitudes, perceived social norms and self-efficacy for each behaviour. The questionnaire was completed by 201 health care workers for the initial validation and 966 other health care workers for a replication.

OUTCOMES

A principal component analysis with Procrustes rotation toward an a priori three component structure on the original sample yielded high congruence measures for reported dental attendance, but not for dietary habits and oral hygiene habits. Subsequent exploratory Varimax rotations and discarding of redundant items resulted in three component solutions explaining 43% of the variance in dietary habits, 57% in oral hygiene habits and 66% in dental attendance, respectively. For all three behaviours, these components corresponded to the dimensions of the TPB. Internal consistency of the scales was satisfactory, with Chronbach's alpha's ranging from 0.51 to 0.87. Scale scores accounted for a significant proportion of the variance in the intention to avoid sweet snacks, to brush teeth, and to attend dental check-ups, and of the frequency of consumption of sweet drinks and frequency of brushing. A confirmatory factor analysis on the larger sample of 966 health care workers provided excellent goodness of fit indices, confirming the construct validity of the scales.

Semi-parametric modelling of the distribution of the baseline risk in meta-analysis

In meta-analysis of clinical trials, often meta-regression analyses are performed to explain the heterogeneity in treatment effects that usually exist between trials. A popular explanatory variable is the risk observed in the control group, the baseline risk. The relationship between the treatment effect and the baseline risk is investigated by fitting a linear model that allows randomness on the true baseline risk by assuming a normal distribution with unknown mean and variance. However, the normality assumption could be too strong to adequately describe the underlying distribution. Therefore, we developed a new semi-parametric method that relaxes the normality assumption to a more flexible and general distribution. We applied a penalized Gaussian mixture distribution to represent the baseline risk distribution. Furthermore, a bivariate hierarchical model is formulated in order to take into account the correlation between the baseline and treatment effect. To fit the proposed model, a penalized likelihood function is maximized by an Expectation Maximization (EM) algorithm. We illustrate our method on a number of simulated data sets and on a published meta-analysis data set.

Caries patterns in primary dentition in 3-, 5- and 7-year-old children: spatial correlation and preventive consequences

This study aimed to contribute to the descriptive information of oral health status in the primary dentition, especially concerning the distribution and spatial correlation of lesions. Data were obtained from two surveys. In the Signal-Tandmobiel project 4,468 7-year-old children in Flanders (Belgium) were selected by a stratified clustered random sample. In the Tandje de Voorste - Smile for Life (TDV) project, data were obtained from 1,291 3-year-old and 1,315 5-year-old children. The children were examined by trained dentists, using standardized and widely accepted criteria, based on the British Association for the Study of Community Dentistry criteria. At the population level, symmetry in the prevalence of caries experience across the midline was tested at the tooth and tooth surface levels using generalized estimating equations and alternating logistic regression (ALR) approaches. Individual symmetry was tested using an approach described by Hujoel et al. [J Dent Res 1994;73:1575-1580]. Descriptive observations suggested a symmetrical distribution of caries experience at the population level. The null hypothesis of symmetry could not be rejected at a 0.05 level, suggesting that caries experience might be symmetric in the deciduous dentition. Based on the ALR approach, 2 x 2 associations of caries experience at the tooth and surface levels, both in 5- and 7-year-olds, appeared to be strongest for the left-right pairs in the mandible, followed by the left-right pairs in the maxilla. At the individual level the hypothesis of random caries pattern was rejected (p < 0.0001). In conclusion, at the population level the null hypothesis of left-right symmetry could not be rejected, while at the individual level lesions tended to cluster on one side of the mouth.

The clinicopathological and prognostic relevance of cytokeratin 7 and 19 expression in hepatocellular carcinoma. A possible progenitor cell origin

AIMS

Cytokeratin (CK) 7 and CK19 expression, present in hepatic progenitor cells (HPCs) and in cholangiocytes but not in normal hepatocytes, has been reported in some hepatocellular carcinomas (HCCs); however, the incidence and relevance of this expression in HCC in Caucasians is not known. Therefore, our aim was to study the occurrence and clinicopathological characteristics of HCC expressing CK7 and/or CK19 in 109 Caucasian patients.

METHODS AND RESULTS

The expression of hepatocellular differentiation markers (Hepar, canalicular polyclonal carcinoembryonic antigen), biliary/progenitor cell markers (CK7, CK19), alpha-fetoprotein (AFP), p53 and beta-catenin in HCC was semiquantitatively assessed by immunohistochemistry. Of 109 HCCs, 78 were CK7-/CK19- (72%), 13 CK7+/CK19- (12%), seven CK7-/CK19+ (6%), 11 CK7+/CK19+ (10%). CK19 expression was significantly associated with elevated serum AFP (400 ng/ml) (P = 0.023), tumour AFP expression (P < 0.0001), presence in serum of anti-hepatitis B core (P = 0.016), less fibrosis in non-neoplastic parenchyma (P = 0.009) and less nuclear beta-catenin expression (P = 0.021). CK7 expression was significantly associated with elevated serum bilirubin (> 2 mg/dl) (P = 0.0005) and less nuclear beta-catenin expression (P = 0.003). HCC expressing CK19 had a higher rate of recurrence (P = 0.009, hazard ratio 12.5, n = 31) after liver transplantation compared with CK19- tumours.

CONCLUSIONS

In our series, 28% of HCCs contained cells expressing CK7 and/or CK19. They potentially derive from HPCs. The higher recurrence rate of CK19+ HCC after transplantation suggests a worse prognosis for these HCCs compared with CK19- HCC.

Effect of caries experience in primary molars on cavity formation in the adjacent permanent first molar

This study sought to address the influence of a sound versus affected first and/or second deciduous molar on the incidence of visible caries experience in the adjacent permanent first molar. Clinical and questionnaire data were obtained from a 6-year prospective oral health screening project in Flanders, Belgium (Signal-Tandmobiel project), where 4,468 children (born in 1989) were examined annually during their primary school time. A multiple survival model allowing for dependent data with possible censoring was applied. The impact of timing of tooth emergence (determining the period at risk), gender, presence of sealants and reported oral hygiene habits was also considered. Cavity formation in permanent first molars was clearly influenced by the status of the adjacent primary molars;the effect of the second deciduous molar was most pronounced. Moreover, if both deciduous molars experienced caries and the child presented with poor oral hygiene, a peak in cavity formation of the permanent first molar 1-2 years after emergence was noticed. On the other hand, if a child presented with good oral hygiene, no peak was observed; caries risk increased slightly over time. No significant benefit from restoring primary molars could be demonstrated, possibly because of methodological limitations.

Analysis of caries experience taking inter-observer bias and variability into account

In larger oral health surveys, clinical measurements are often obtained using several examiners. This raises the issue of inter-observer variability in measurement. Often, the problem is dealt with by reporting kappa values obtained in a calibration exercise. In the present study, the limitations of this statistic are presented, and an alternative, based on a Bayesian approach, is proposed. When the alternative approach was applied to caries experience data obtained in an oral health screening survey in seven-year-old Flemish children (Signal Tandmobiel study), it could be ruled out that the observed geographic East-West gradient was due to bias induced by variability in scoring of the different dental examiners involved. The proposed method offers an opportunity to refine existing analytical approaches and is relevant to any health outcome study.

Variables associated with oral hygiene levels in 7-year-olds in Belgium

OBJECTIVES

The present investigation aimed to delineate the extent of the correlation between each of a number of possible associated variables and the amount of plaque (as measured by a plaque index) present in 7-year-old Flemish children and to investigate the potential relation between caries experience and the amount of dental plaque.

PARTICIPANTS

A sample of 4,468 children representative of children living in Flanders and born in 1989 was obtained. Their mean age was 7.1 (SD = 0.4) years with a boys/girls ratio of 51.8%-48.2%.

METHODS

Analyses were performed with ordinal plaque index outcome, considering the children's school (i.e. class) as a random effect. Random effect models were estimated with the SAS procedure Proc NLMIXED.

RESULTS

From the multiple logistic regression analysis, it could be concluded that the older the subjects the more dental plaque was observed (OR = 1.35; CI 1.08-1.68; p = 0.0081). Moreover, four explanatory variables could be identified as significant determinants for dental plaque in the mixed dentition: 'start of brushing after the age of 2' (OR = 1.161; CI: 1.09-1.23; p < 0.0001), 'brushing frequency of < 1x/day' (OR = 1.369; CI: 1.012-1.66; p = 0.0019), 'more than two in between meal snacks' (OR = 1.18; CI: 1.023-1.36; p = 0.0111) and 'daily intake of sugar containing drinks' (OR = 1.24; CI: 1.08-1.42; p = 0.0023). Further, being male was associated with a higher amount of plaque (OR = 1.16; CI: 1.00-1.34; p = 0.0506) and finally, a weak but significant correlation (Spearmans' rho: r = 0.23, p < 0.0001) between caries experience and the amount of cervical plaque was shown.

CONCLUSION

The amount of cervical plaque in the mixed dentition was significantly related to variables associated with oral hygiene and dietary habits. A low but significant correlation between the amount of dental plaque and caries experience was observed.

Low back pain in Flemish adolescents and the role of perceived social support and effect on the perception of back pain

AIM

To document the prevalence of low back pain (LBP) in Flemish adolescents and to recognize the association between perceived social support and affect and medical consultation and reduction of activities.

METHODS

The study comprised 620 adolescents. A questionnaire was used to elicit the characteristics of LBP. Perceived social support and affect were investigated using the Personal Resource Questionnaire (PRQ) and the Positive Affect Negative Affect Scale (PANAS), respectively. Comparisons were made between adolescents with LBP and adolescents without LBP. Within the LBP group, answers from adolescents appealing for medical advice were compared with those who did not seek advice. Another comparison was done between adolescents who reduced their activities and those who did not. Descriptive statistics, the median test for two samples, a chi2 test or Fisher's exact probability test and logistic regression analysis were used for data analysis.

RESULTS

Month prevalence of LBP was 24.7%. Thirteen adolescents visited a physician or received treatment, pain severity being the sole factor associated with the decision. Fifteen adolescents reduced their sports activities and 11 reduced or stopped other leisure activities. Pain severity and negative affect were the main associated factors.

CONCLUSION

The results of the present study encourage further research on the role of pain perception and the influence of psychosocial factors on back pain in adolescents.

Modelling tooth emergence data based on multivariate interval-censored data

Studies on emergence of (permanent) teeth are published regularly in the dental literature. Besides descriptive statistics (mean or median values) on emergence times, comparisons between boys and girls are of interest. Gender comparisons are intersubject analyses, but also intrasubject questions, like 'Is there a left-right symmetry with respect to the mean (median) emergence times?' are of interest. Studies on emergence times most often are faced with interval-censored data. We will extend a GEE-type test proposed by Huster et al. for bivariate right-censored data to the multivariate setting with interval-censored data. Central to our paper is to provide appropriate statistical models to resolve some dental questions on emergence. The analyses are based on data from the longitudinal Signal-Tandmobiel study.

[Factors determining dental plaque and their relation to the assessment of caries among 7-year-old Flemish children]

The present investigation aimed to assess the impact of a number of variables on the variation on the amount of dental plaque present in the mixed dentition of 7-year-old Flemish children and to investigate a potential correlation between caries experience and the amount of dental plaque. Analyses were performed with ordinal plaque index as outcome measure, considering the children's school (i.e. class) as a random effect. This random effect turned out to be of great importance when dealing with large clustered samples. From the multiple logistic regression analysis, it was concluded that the older the subjects the more dental plaque was observed (or = 1.35; CI: 1.08-1.68; P = 0.0081). Four exploratory variables, could be identified as important risk indicators for dental plaque accumulation in the mixed dentition: 'start of brushing after the age of 2 years' (or = 1.161; CI: 1.09-1.23; P < 0.0001), 'brushing frequency of < 1 x/day (or = 1.37; CI: 1.12-1.66; P = 0.0019), more than two in between meal snacks (or = 1.18; CI: 1.03-1.36; P = 0.02) and daily intake of sugar containing drinks (or = 1.24; CI: 1.08-1.42; P = 0.002). Further, being male was linked to a higher amount of plaque (or = 1.16; CI: 1.00-1.34; P = 0.0506). Finally a weak but significant correlation (spearman's rho: r = 0.23; P < 0.0001) between caries experience and the amount of cervical plaque was shown.

[In the interests of taking one's time even when one reads a scientific article! Critical review of the literature on tooth eruption]

Nowadays everybody has access to the medical literature, not only through the libraries of the universities, but even more conveniently via Medline and Pubmed. It is very tempting only to read abstracts and summaries instead of "wasting time" by reading the whole article. Reading the article itself however enables the reader to select only these articles that have a good methodological and statistical basis. In this contribution several topics will be discussed that may help the reader to read scientific articles with a critical mind: definitions, quality of the data, study concepts, composition of the sample, control group, statistical analysis, sample size and interpretation of the results. Different items are exemplified with articles that deal with the emergence of permanent teeth.

[Analysis of dental caries risk factors in the primary dentition]

The aim of the present study was to assess indicators shown to be associated with the prevalence of caries in the primary dentition of 7-year-old Flemish schoolchildren. Cross-sectional first year data of the Signal-Tandmobiel survey were analysed (n = 4468). Gender, age, oral hygiene habits, use of fluorides, dietary habits, geographical factors and social factors were involved in logistic regression models. There was a significant difference (P < 0.05) in caries experience determined by the geographical spread, with an explicit trend of caries declining from the east to the west. A start of brushing before the age of 2 and a brushing frequency of at least once a day need to be encouraged, while the use of sugar containing drinks and snacks between meals needs to be restricted to a maximum of 2 per day. The mean dmf-t values were the lowest for the most advantaged children and where threefold higher in the least advantaged children. There was a cumulative effect of decreasing social level and oral health habits on the caries prevalence. Social inequalities in oral health among children exist in Flanders.

Assessing risk indicators for dental caries in the primary dentition

The aim of the present study was to assess indicators shown to be associated with the prevalence of caries in the primary dentition of 7-year-old Flemish schoolchildren. Cross-sectional first year data of the longitudinal Signal-Tandmobiel survey were analysed (n=4468). Gender, age, oral hygiene habits, use of fluorides, dietary habits, geographical factors and parental modelling were the considered predictors. From the multiple logistic regression analysis, including schools as a random effect, and after adjusting for the confounding variables-educational system and province (stratification variables), gender and age-it became clear that the following risk indicators remained significant (at 5% level) for the presence of caries: frequency of toothbrushing (P=0.05) with an OR 1.24 for brushing less than once a day, age at start of brushing (P<0.001) with an OR=1.22 for a delay of 1 year, regular use of fluoride supplements (P<0.001) with an OR=1.54 for no use, daily use of sugar-containing drinks between meals (P<0.001) with an OR=1.38, and number of between-meals snacks (P=0.012) with an OR=1.22 for using more than 2 between-meal snacks. There was a significant difference (P<0.05) in caries experience determined by the geographical spread, with an explicit trend of caries declining from the east to the west. In a model with an ordinal response outcome, the daily use of sugar-containing drinks between meals had a more pronounced effect when caries levels were high. From this study it became obvious that, in Flemish children, an early start of brushing and a brushing frequency of at least once a day need to be encouraged, while the use of sugar-containing drinks and snacks between meals needs to be restricted to a maximum of 2 per day. Geographical differences need to be investigated in more detail.

Parental occupational status related to dental caries experience in 7-year-old children in Flanders (Belgium)

OBJECTIVES

The study aimed to investigate differences in the oral health of 7-year-old Flemish children by the parents' occupational class and to explore behavioural and demographic factors linked to social inequalities in oral health.

PARTICIPANTS

A stratified cluster sample of 1,500 primary school children in Flanders, born in 1989.

METHOD

The clinical oral examination was performed in the first year of a longitudinal oral health project in Flanders (1996-2001) by 16 calibrated examiners. Information on occupational level of the parents and on oral health related behaviour was gained, respectively, by the school health care centres and the parents of the children using a structured questionnaire.

RESULTS

The mean dmf values between the several SES groups, assessed by occupational status of the parents, were significantly different (P<0.001). The mean dmft and dmfs values were the lowest for the most advantaged children (1.3/2.7) and were threefold higher in the least advantaged children (3.9/9.1). The prevalence of caries-free children was 2.5 times higher among the highest SES families compared with the lowest SES families. In a logistic regression model adjusted for the stratification factors, the excess risk for caries in children increased with decreasing occupational level of the parents. There was a cumulative effect of decreasing occupational level and oral health habits on the caries prevalence.

CONCLUSION

The results of the study allow the supposition that social inequalities in oral health among children exist in Flanders. Environmental factors and individual behavioural factors partially explain why social class might influence oral health.

The value of a baseline caries risk assessment model in the primary dentition for the prediction of caries incidence in the permanent dentition

To establish a reliable screening method for caries prediction and to identify predominant risk factors, this study tested whether a cross-sectional caries risk model assessed at age 7 could be used to predict future caries onset in the permanent first molars at age 10 in 3,303 children born in 1989. As prediction variables, assessing the believed risk, baseline data at age 7 on oral health status, oral hygiene level, oral health behaviour and sociodemographic factors were used. The real risk, based on data collected for the first permanent molars during the follow-up, was assessed by different approaches. Cumulative incidence during the 3-year observation period was 31.6%, ranging from 22.4% in the believed low-risk group to 43.2% in the believed high-risk group. A stepwise logistic regression analysis was performed with net caries increment as outcome measure, adjusted for the real time at risk, using eruption times. Baseline dmfs and occlusal and buccal plaque indices were highly significant for having a high caries increment in permanent first molars with respective odds ratios of 1.07, 1.43 and 1.35. Brushing less than once a day and the daily use of sugar-containing drinks between meals were confirmed as risk factors (OR 2.43 and 1.25, respectively). The logistic regression analysis provided a sensitivity of 59-66% and a specificity of 65.7-72.8%, which indicates that the risk marker did not have an important predictive power. None of the socio-demographic and behavioural variables had enough predictive power at community level to be useful for identifying caries susceptible children. Even the power of dmfs at baseline must be considered modest.

Modelling bivariate ordinal responses smoothly with examples from ophthalmology and genetics

A non-parametric implementation of the bivariate Dale model (BDM) is presented as an extension of the generalized additive model (GAM) of Hastie and Tibshirani. The original BDM is an example of a bivariate generalized linear model. In this paper smoothing is introduced on the marginal as well as on the association level. Our non-parametric procedure can be used as a diagnostic tool for identifying parametric transformations of the covariates in the linear BDM, hence it also provides a kind of goodness-of-fit test for a bivariate generalized linear model. Cubic smoothing spline functions for the covariates are estimated by maximizing a penalized version of the log-likelihood. The method is applied to two studies. The first study is the classical Wisconsin Epidemiologic Study of Diabetic Retinopathy. The second study is a twin study, where the association between the elements of twin pairs is of primary interest. The results show that smoothing on the association level can give a significant improvement to the model fit.

Sensitivity analysis for nonrandom dropout: a local influence approach

Diggle and Kenward (1994, Applied Statistics 43, 49-93) proposed a selection model for continuous longitudinal data subject to nonrandom dropout. It has provoked a large debate about the role for such models. The original enthusiasm was followed by skepticism about the strong but untestable assumptions on which this type of model invariably rests. Since then, the view has emerged that these models should ideally be made part of a sensitivity analysis. This paper presents a formal and flexible approach to such a sensitivity assessment based on local influence (Cook, 1986, Journal of the Royal Statistical Society, Series B 48, 133-169). The influence of perturbing a missing-at-random dropout model in the direction of nonrandom dropout is explored. The method is applied to data from a randomized experiment on the inhibition of testosterone production in rats.

Treatment of reinfarction after thrombolytic therapy for acute myocardial infarction: an analysis of outcome and treatment choices in the global utilization of streptokinase and tissue plasminogen activator for occluded coronary arteries (gusto I) and assessment of the safety of a new thrombolytic (assent 2) studies

BACKGROUND

Early reinfarction after thrombolytic therapy is associated with adverse outcomes and increased mortality. Among patients with reinfarction in the 1992 Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries (GUSTO I) and the 1998 Assessment of the Safety of a New Thrombolytic (ASSENT 2) trials, we investigated temporal and regional differences in the use of repeat thrombolysis, revascularization (angioplasty and/or bypass surgery), or conservative measures and the outcomes of each management strategy.

METHODS AND RESULTS

Data from the 4% of patients (n=2301) who experienced reinfarction after thrombolytic therapy were studied. Baseline characteristics, 30-day mortality, and incidence of total and hemorrhagic strokes were compared among the 3 treatment groups. The 30-day mortality did not differ between the repeat thrombolysis and revascularization groups (P=0.72), and it was significantly lower among patients treated by these 2 strategies than in those treated conservatively (11% and 11% versus 28%, respectively; P<0.001). Stroke rates did not differ significantly between the 3 treatment strategies (P=0.49). From 1992 to 1998, the percentage of reinfarction patients treated with repeat thrombolysis decreased from 29.3% to 18.5% in US centers and from 51.4% to 41.9% in all other centers (P<0.001). In contrast, use of revascularization procedures increased from 33.5% to 47.9% in US centers and from 8.1% to 23.0% in all other centers (P<0.001).

CONCLUSIONS

Repeat thrombolysis and revascularization are associated with significantly lower mortality among reinfarction patients. Randomized trials are necessary to assess the exact risks and benefits of rethrombolysis versus interventional revascularization in this subset of high-risk patients presenting with reinfarction after thrombolytic therapy.

Estimating the power of compliance-improving methods

Measuring patient compliance with prescribed or protocol-specified regimens of drug administration is an important topic in clinical trial research. One of the most useful expressions of compliance is the proportion of days the prescribed dose was taken correctly, but the distribution of this variable is often bimodal. We show that a two-component mixture of beta distributions can be fitted to these data. The two components can be interpreted as the groups of poorly and satisfactorily complying patients, respectively. This representation was helpful in the calculation of the power in a recent trial of a method to improve the compliance of patients with a lipid-lowering drug regimen. Control Clin Trials 2000;21:540-551

Fraud in medical research: an international survey of biostatisticians. ISCB Subcommittee on Fraud

The characteristics of scientific fraud and its impact on medical research are in general not well known. However, the interest in the phenomenon has increased steadily during the last decade. Biostatisticians routinely work closely with physicians and scientists in many branches of medical research and have therefore unique insight into data. In addition, they have methodological competence to detect fraud and could be expected to have a professional interest in valid results. Biostatisticians therefore are likely to provide reliable information on the characteristics of fraud in medical research. The objective of this survey of biostatisticians, who were members of the International Society for Clinical Biostatistics, was to assess the characteristics of fraud in medical research. The survey was performed between April and July 1998. The participation rate was only 37%. We report the results because a majority (51%) of the participants knew about fraudulent projects, and many did not know whether the organization they work for has a formal system for handling suspected fraud or not. Different forms of fraud (e.g., fabrication and falsification of data, deceptive reporting of results, suppression of data, and deceptive design or analysis) had been observed in fairly similar numbers. We conclude that fraud is not a negligible phenomenon in medical research, and that increased awareness of the forms in which it is expressed seems appropriate. Further research, however, is needed to assess the prevalence of different types of fraud, as well as its impact on the validity of results published in the medical literature.

An overview of group sequential methods in longitudinal clinical trials

During the last decade, several papers have been published on group sequential methods in general and on sequential longitudinal clinical trials in particular. This paper gives an overview of the proposed methods, emphasizing longitudinal clinical trials. Furthermore, it tries to answer some practical questions that may arise during the conduct of interim analyses in longitudinal trials. Simulations have been carried out to obtain insight in these practical considerations.

Disease mapping models: an empirical evaluation. Disease Mapping Collaborative Group

The analysis of small area disease incidence has now developed to a degree where many methods have been proposed. However, there are few studies of the relative merits of the methods available. While many Bayesian models have been examined with respect to prior sensitivity, it is clear that wider comparisons of methods are largely missing from the literature. In this paper we present some preliminary results concerning the goodness-of-fit of a variety of disease mapping methods to simulated data for disease incidence derived from a range of models. These simulated models cover simple risk gradients to more complex true risk structures, including spatial correlation. The main general results presented here show that the gamma-Poisson exchangeable model and the Besag, York and Mollie (BYM) model are most robust across a range of diverse models. Mixture models are less robust. Non-parametric smoothing methods perform badly in general. Linear Bayes methods display behaviour similar to that of the gamma-Poisson methods.

The value of the number-needed-to-treat method in antiepileptic drug trials

PURPOSE

There is controversy between clinicians and statisticians on the appropriateness of the number needed to treat (NNT) as a summary statistic to report the effectiveness of a treatment. We examine the two viewpoints and make proposals concerning the reporting of clinical trial results.

METHODS

In the context of antiepileptic treatments, we explain the two different viewpoints and illustrate the use of the odds ratio, relative risk, absolute difference, and NNT on the results of randomized clinical trials with topiramate (TPM). Special attention is paid to the use of these summary statistics in meta-analyses. Here, the NNT is the expected number of patients one would need to treat to achieve a single occurrence of a specified good outcome (e.g., 50% reduction in seizure rate) in comparison to no (or placebo) treatment.

RESULTS

Although the NNT is readily interpretable in some instances, it exhibits undesirable statistical behavior in other cases. In particular, confidence intervals for the NNT may split into two intervals and extend to positive and negative infinity when treatment efficacy is not clearly established by the data. Meta-analyses cannot be sensibly conducted directly on the NNT scale.

CONCLUSIONS

Although other measures, such as the odds ratio, have been more commonly used in the context of meta-analyses, clinicians prefer the NNT because it gives them a clearer clinical interpretation of the effectiveness of a (new) treatment. On the other hand, statisticians do not recognize the value of the NNT, as it has undesirable statistical properties. Some reconciliation between the two views could be achieved when the clinicians acknowledge the weak aspects of the NNT and when statisticians realize that statistical appropriateness is not the same as clinical relevance. It is suggested that the NNT be used as a secondary reporting tool not on an equal footing with the classic scales.