Using machine learning or deep learning models in a hospital setting to detect inappropriate prescriptions: a systematic review

OBJECTIVES

The emergence of artificial intelligence (AI) is catching the interest of hospital pharmacists. A massive collection of health data is now available to train AI models and hold the promise of disrupting codes and practices. The objective of this systematic review was to examine the state of the art of machine learning or deep learning models that detect inappropriate hospital medication orders.

METHODS

A systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. MEDLINE and Embase databases were searched from inception to May 2023. Studies were included if they reported and described an AI model intended for use by clinical pharmacists in hospitals. Risk of bias was assessed using the Prediction model Risk Of Bias ASsessment Tool (PROBAST).

RESULTS

13 articles were selected after review: 12 studies were judged to have high risk of bias; 11 studies were published between 2020 and 2023; 8 were conducted in North America and Asia; 6 analysed orders and detected inappropriate prescriptions according to patient profiles and medication orders; and 7 detected specific inappropriate prescriptions, such as detecting antibiotic resistance, dosage abnormality in prescriptions, high alert drugs errors from prescriptions or predicting the risk of adverse drug events. Various AI models were used, mainly supervised learning techniques. The training datasets used were very heterogeneous; the length of study varied from 2 weeks to 7 years and the number of prescription orders analysed went from 31 to 5 804 192.

CONCLUSIONS

This systematic review points out that, to date, few original research studies report AI tools based on machine or deep learning in the field of hospital clinical pharmacy. However, these original articles, while preliminary, highlighted the potential value of integrating AI into clinical hospital pharmacy practice.

Evaluating the Portability of Rheumatoid Arthritis Phenotyping Algorithms: A Case Study on French EHRs

Previous work has successfully used machine learning and natural language processing for the phenotyping of Rheumatoid Arthritis (RA) patients in hospitals within the United States and France. Our goal is to evaluate the adaptability of RA phenotyping algorithms to a new hospital, both at the patient and encounter levels. Two algorithms are adapted and evaluated with a newly developed RA gold standard corpus, including annotations at the encounter level. The adapted algorithms offer comparably good performance for patient-level phenotyping on the new corpus (F1 0.68 to 0.82), but lower performance for encounter-level (F1 0.54). Regarding adaptation feasibility and cost, the first algorithm incurred a heavier adaptation burden because it required manual feature engineering. However, it is less computationally intensive than the second, semi-supervised, algorithm.

Psychopathological profile and sagittal alignment in low-back pain

INTRODUCTION

Low-back pain requires comprehensive care using a biopsychosocial model. The psychologic dimension plays an important role, but the link between sagittal alignment and a given psychopathological profile is little studied. The aim of this study was to analyze the psychopathological profiles and sagittal parameters of a population with low-back pain and to assess the link.

MATERIAL AND METHODS

205 patients, with a mean age of 49.6 years (range, 18-70 years), presenting chronic common low-back pain without radicular involvement, were included prospectively. Mood scores comprised: the self-administered "Hospital Anxiety and Depression Scale" (HAD), Hamilton Anxiety Scale (HAM-A), Hamilton Depression Scale (HAM-D) and Young Mania Rating Scale (YMRS). Radiological parameters, measured on lateral full-spine radiographs, included: L1-S1 lordosis, T1-T12 kyphosis, pelvic incidence, pelvic tilt, sacral slope, sagittal vertical axis (SVA), T1 slope, and Roussouly type.

RESULTS

Mean HAM-A score was 16.1; 54% of patients had scores ≥14, indicating anxiety disorder. Mean HAM-D score was 10.8; 55% of patients had scores ≥10, indicating depressive disorder. Mean YMRS score was 2.6; only 1 patient had a score ≥20, indicating manic disorder. The 112 patients with HAM-A score >14 showed mean 51.6° L1-S1 lordosis (p=0.356), 48.3° T1-T12 kyphosis (p=0.590), -4.3mm C7 SVA (p=0.900), and 29.3° T1 slope (p=0.451). In case of HAM-A <14, there were no significant differences. The 113 patients with HAM-D score >10 showed significant differences in T1-T12 kyphosis (mean 49.0°; p<0.05) and T1 slope (30.2°; p<0.05); mean L1-S1 lordosis was 50.5° (p=0.861) and C7 SVA 1.6mm (p=0.462). In case of HAM-D <10, T1-T12 kyphosis was 45.5° (p<0.05) and T1 slope 26.2° (p<0.05); mean lordosis was 50.9° (p=0.861) and mean C7 SVA -7.1mm (p=0.259). Multivariate analysis found no significant link between Roussouly type and psychiatric scores: HAD (p=0.715), HAM-A (p=0.652), and HAM-D (p=0.902).

CONCLUSION

More than 50% of patients with common low-back pain presented a mood disorder. Depressive disorder was associated with greater T1-T12 kyphosis and T1 slope. There was no relationship between psychiatric scores and overall sagittal alignment.

LEVEL OF EVIDENCE

II.

Environmental Drivers of Monkeypox Transmission in the Democratic Republic of the Congo

Monkeypox (MPX) is an emergent severe zoonotic disease resembling that of smallpox. To date, most cases of human MPX have been reported in the Democratic Republic of the Congo (DRC). While the number of cases has increased steadily in the DRC over the last 30 years, the environmental risk factors that drive the spatiotemporal dynamics of MPX transmission remain poorly understood. This study aimed to investigate the spatiotemporal associations between environmental risk factors and annual MPX incidence in the DRC. All MPX cases reported weekly at the health zone level over a 16-year period (2000-2015) were analyzed. A Bayesian hierarchical generalized linear mixed model was conducted to identify the spatiotemporal associations between annual MPX incidence and three types of environmental risk factors illustrating environment as a system resulting from physical, social and cultural interactions Primary forest (IRR 1.034 [1.029-1.040]), economic well-being (IRR 1.038 [1.031-1.047]), and temperature (IRR 1.143 [1.028-1.261]) were positively associated with annual MPX incidence. Our study shows that physical environmental risk factors alone cannot explain the emergence of MPX outbreaks in the DRC. Economic level and cultural practices participate from environment as a whole and thus, must be considered to understand exposure to MPX risk Future studies should examine the impact of these factors in greater detail.

COVID-19 Impact on Stroke Admissions during France's First Epidemic Peak: An Exhaustive, Nationwide, Observational Study

Introduction

The coronavirus disease 2019 (COVID-19) pandemic continues to have great impacts on the care of non-COVID-19 patients. This was especially true during the first epidemic peak in France, which coincided with the national lockdown. The aim of this study was to identify whether a decrease in stroke admissions occurred in spring 2020, by analyzing the evolution of all stroke admissions in France from January 2019 to June 2020.

Methods

We conducted a nationwide cohort study using the French national database of hospital admissions (Information Systems Medicalization Program) to extract exhaustive data on all hospitalizations in France with at least one stroke diagnosis between January 1, 2019, and June 30, 2020. The primary endpoint was the difference in the slope gradients of stroke hospitalizations between pre-epidemic, epidemic peak, and post-epidemic peak phases. Modeling was carried out using Bayesian techniques.

Results

Stroke hospitalizations dropped from March 10, 2020 (slope gradient: −11.70), and began to rise again from March 22 (slope gradient: 2.090) to May 7. In total, there were 23,873 stroke admissions during the period March–April 2020, compared to 29,263 at the same period in 2019, representing a decrease of 18.42%. The percentage change was −15.63%, −25.19%, −18.62% for ischemic strokes, transient ischemic attacks, and hemorrhagic strokes, respectively.

Discussion/Conclusion

Stroke hospitalizations in France experienced a decline during the first lockdown period, which cannot be explained by a sudden change in stroke incidence. This decline is therefore likely to be a direct, or indirect, result of the COVID-19 pandemic.

Cognitive reserve as one of the determinants of the progression of Alzheimer disease

To assess the validity of Yaakov Stern's theory, which postulates that the first cognitive deficits of Alzheimer disease are delayed in patients with high cognitive reserve, compared to those with low cognitive reserve, but this is subsequently associated with faster cognitive decline when cerebral lesions have reached a certain degree of severity. A retrospective study of 318 patients with Alzheimer disease without associated cerebral disorders, was performed at the Memory Clinic of the University Hospital of Strasbourg. The patients were divided into five groups according to their degree of cognitive reserve, as assessed by their level of education (primary, middle school, intermediate, high school and higher education). Cognitive status was assessed by scores on the Mini Mental Status Examination (MMSE) at the first examination and during follow-up lasting four years or longer. Slopes of cognitive decline were computed and compared between groups using ß linear regression. Patients in the higher education group had higher MMSE scores than those in the other groups at the first consultation, but a statistically significant steeper slope during the follow-up. Our results are consistent with Stern's theoretical model, which assumes that high cognitive reserve is accompanied by compensation mechanisms which hinder the onset of clinical cognitive decline in patients with Alzheimer disease, but that is then associated with a faster decline at a later stage when the reserve is exhausted.

Effect of emergency start and central venous catheter on outcomes in incident hemodialysis patients: a prospective observational cohort

BACKGROUND

Unfavorable conditions at hemodialysis inception reduce the survival rate. However, the relative contribution to outcomes of predialysis follow-up, symptoms, emergency start or central venous catheter (CVC) is unknown.

METHODS

We analyzed the determinants of survival according to dialysis initiation conditions in the nationwide REIN registry, using two methods based either on clinical classification or data mining. We divided patients into four groups according to dialysis initiation (emergency vs planned, symptoms or not, previous follow-up). "Followed planned starters" began dialysis as outpatients and with an arteriovenous fistula (AVF). "Followed symptomatic non-urgent starters" were patients who started earlier because of any non-urgent symptomatic event. "Followed urgent starters" had seen a nephrologist before inception but started dialysis in an emergency condition. "Unknown urgent starters" were patients without any follow-up and who had a CVC at inception.

RESULTS

"Followed urgent" starters had the lowest 2-year survival rate (66.8%) compared to "followed planned" (77.3%), "followed symptomatic non urgent" (79.2%), and "unknown urgent" (71.7%). Compared to other groups, the risk of mortality was lower in followed symptomatic non urgent (HR 0.86 95% CI 0.75-0.99) and higher in followed urgent starters (HR 1.05 (95% CI 0.94-1.18). In data mining Classification And Regression Tree regrouping in five categories, the lowest 2-year survival (52.3%) was in over 70-year-old starters with a CVC. The survival was 93.2% in under 57-year-old patients without active cancer, 82.5% in 57-70-year-old individuals without cancer, 72.4% in over 70-year-old patients without CVC and 61.4% in under 70-year-old subjects with cancer. The hazard ratio of data mining categories varied between 2.12 (95% CI 1.73-2.60) in 57-70-year-old subjects without cancer and 4.42 (95% CI 3.64-5.37) in over 70-year-old patients with CVC. Therefore, regrouping incident patients into five data mining categories, identified by age, cancer, and CVC use, could discriminate the 2-year survival in patients starting hemodialysis.

CONCLUSIONS

Although each classification captured different prognosis information, both analyses showed that starting hemodialysis on a CVC has more dramatic outcomes than emergency start per se.

PoleSat's Modelling a Reorganization of Vascular Catheterization Provision at the Territorial Hospital Grouping Level

Since a French organization (2016) has defined "the territorial hospital groupings", public hospitals must share medical-economic knowledge and decision-makers expect prospective analyses. PoleSat aims, quick hospital-catchment area modellings, completed by population analyses. Modellings are based on "diagnostic and interventional vascular catheterizations" acts and Nouvelle-Aquitaine, and they are carried out 3 times, through the graphical user interface's main-setting values, coupled with 3 activity-scenarios. Scenario results cannot confirm the NA02-Atlantique's H0. The experts have approved PoleSat's method as a robust help-tool; therefore they project to repeat its usages.

Polesat-Web-2018: A Simulation IT Tool with Immediate Prospective and Strategic Views of Hospital Spatial Planning

Medical geo-informatics allows the Health world to address major challenges thanks to attractive concepts, methods and user-friendly IT. PoleSat-web-2018 presents a decision support system - a modelling "variable geometry" IT tool for simulation of hospital spatial planning. The outputs enable quasi-instantaneous analytic visualization at several geographic levels. PoleSat-web-2018 provides prospective views of hospital catchments (by grouping, closing) and proves to be relevant for the French planners of the Ministry of Health.

Reliability, factor analysis and internal consistency calculation of the Insomnia Severity Index (ISI) in French and in English among Lebanese adolescents

Objectives

Our goal is to validate and to verify the reliability of the French and English versions of the Insomnia Severity Index (ISI) in Lebanese adolescents.

Study design

A cross-sectional study was implemented.

Methods

104 Lebanese students aged between 14 and 19 years participated in the study. The English version of the questionnaire was distributed to English-speaking students and the French version was administered to French-speaking students. A scale (1 to 7 with 1 = very well understood and 7 = not at all) was used to identify the level of the students' understanding of each instruction, question and answer of the ISI. The scale's structural validity was assessed. The factor structure of ISI was evaluated by principal component analysis. The internal consistency of this scale was evaluated by Cronbach's alpha. To assess test–retest reliability the intraclass correlation coefficient (ICC) was used.

Results

The principal component analysis confirmed the presence of a two-component factor structure in the English version and a three-component factor structure in the French version with eigenvalues > 1. The English version of the ISI had an excellent internal consistency (α = 0.90), while the French version had a good internal consistency (α = 0.70). The ICC presented an excellent agreement in the French version (ICC = 0.914, CI = 0.856–0.949) and a good agreement in the English one (ICC = 0.762, CI = 0.481–890). The Bland–Altman plots of the two versions of the ISI showed that the responses over two weeks' were comparable and very few outliers were detected.

Conclusion

The results of our analyses reveal that both English and French versions of the ISI scale have good internal consistency and are reproducible and reliable. Therefore, it can be used to assess the prevalence of insomnia in Lebanese adolescents.

•

Principal component analysis found two factors in the English version and three in the French version of the ISI.

•

The English version of the ISI showed an excellent internal consistency (α = 0.90).

•

Almost a perfect agreement was obtained with the French version of ISI where ICC = 0.914 (CI = 0.856–0.949).

•

The Bland–Altman plots of the two versions of ISI showed that the responses over two weeks’ time were comparable.

•

Both English and French versions of the ISI are reproducible and reliable.

Cervical sagittal alignment in adult hyperkyphosis treated by posterior instrumentation and in situ bending

BACKGROUND

In the normal adult spine, a link between thoracolumbar and cervical sagittal alignment exists, suggesting adaptive cervical positional changes allowing horizontal gaze. In patients with thoracic hyperkyphosis, cervical adaptation to sagittal global alignment might be different from healthy individuals. However, this relationship has not clearly been reported in hyperkyphotic deformity.

PURPOSE

The purpose of this study was to identify cervical sagittal alignment types observed on radiographs in young adults with thoracic hyperkyphosis. The relationship between cervical and thoracolumbar alignment as well as the effect of posterior instrumentation and adaptive positional changes of the mobile cervical segment were retrospectively analyzed.

PATIENTS AND METHODS

Twenty-three patients (32.7 years; 5-year follow-up) were included. Full spine radiographic measurements were: T1 slope, T1-T4 kyphosis, T4-T12 kyphosis, L1-S1 lordosis, pelvic incidence, pelvic tilt, sacral slope, SVA C7, SVA C2, lordosis between C0-C2, C2-C7, C2-C4 and C4-C7. A Bayesian model and Spearman correlation were used.

RESULTS

Two alignment types existed: cervical lordosis (group A) and cervical kyphosis (group B). Preoperatively, T4-T12 kyphosis and L1-S1 lordosis were significantly higher in group A: 76.6° versus 59.4° and -72.8° versus -65.8° (probability of>5° difference P (β>5)>0.95). Pelvic incidence was higher in group A (49.8° versus 44.2°) and C0-C2 lordosis in group B (-29.4° versus -21.6°). A significant correlation existed between: T4-T12 kyphosis and C2-C7 lordosis, L1-S1 lordosis and pelvic incidence, C2-C7 lordosis and T1 slope, C2-C7 lordosis and T1-T4 kyphosis. Postoperatively, T4-T12 kyphosis decreased by 33.1° P (β>5)=0.9995), L1-S1 lordosis decreased by 17.7° (P (β>5)=0.961), T1-T4 kyphosis increased by 14.1° (P (β>5)=0.973). SVA C2 (translation) increased by 13.8mm. C0-C2 lordosis (head rotation) remained unchanged. Six patients changed cervical alignment. PJK occurred in 15 patients, unrelated to cervical alignment or proximal instrumentation level.

DISCUSSION

Two cervical alignment types, lordotic or kyphotic, were observed thoracic hyperkyphosis patients. This alignment was mainly triggered by the amount of thoracic kyphosis and lumbar lordosis, linked to pelvic incidence. Moreover, the inclination of the C7-T1 junctional area plays a key role in the amount of cervical lordosis. The correction of T4-T12 kyphosis induced compensatory modifications at adjacent segments: T1-T4 kyphosis increase (PJK) and L1-S1 lordosis decrease. Global spino-pelvic alignment and head position did not change in the sagittal plane. The cervical spine tented to keep in its preoperative position in most patients.

LEVEL OF EVIDENCE

Level IV.

Pullout characteristics of percutaneous pedicle screws with different cement augmentation methods in elderly spines: An in vitro biomechanical study

BACKGROUND

Vertebroplasty prefilling or fenestrated pedicle screw augmentation can be used to enhance pullout resistance in elderly patients. It is not clear which method offers the most reliable fixation strength if axial pullout and a bending moment is applied. The purpose of this study is to validate a new in vitro model aimed to reproduce a cut out mechanism of lumbar pedicle screws, to compare fixation strength in elderly spines with different cement augmentation techniques and to analyze factors that might influence the failure pattern.

MATERIALS AND METHODS

Six human specimens (82-100 years) were instrumented percutaneously at L2, L3 and L4 by non-augmented screws, vertebroplasty augmentation and fenestrated screws. Cement distribution (2 ml PMMA) was analyzed on CT. Vertebral endplates and the rod were oriented at 45° to the horizontal plane. The vertebral body was held by resin in a cylinder, linked to an unconstrained pivot, on which traction (10 N/s) was applied until rupture. Load-displacement curves were compared to simultaneous video recordings.

RESULTS

Median pullout forces were 488.5 N (195-500) for non-augmented screws, 643.5 N (270-1050) for vertebroplasty augmentation and 943.5 N (750-1084) for fenestrated screws. Cement augmentation through fenestrated screws led to significantly higher rupture forces compared to non-augmented screws (P=0.0039). The pullout force after vertebroplasty was variable and linked to cement distribution. A cement bolus around the distal screw tip led to pullout forces similar to non-augmented screws. A proximal cement bolus, as it was observed in fenestrated screws, led to higher pullout resistance. This cement distribution led to vertebral body fractures prior to screw pullout.

CONCLUSION

The experimental setup tended to reproduce a pullout mechanism observed on radiographs, combining axial pullout and a bending moment. Cement augmentation with fenestrated screws increased pullout resistance significantly, whereas the fixation strength with the vertebroplasty prefilling method was linked to the cement distribution.

Cancer incidence and survival in adolescents and young adults in France, 2000-2008

This study aimed to describe cancer incidence (2000-2008) and survival (2000-2004) in France in adolescents and young adults (AYA). All cases of cancer diagnosed in 15-24 years, recorded by all French population-based registries (14% of the French population), over the 2000-2008 period, were included. Incidence change over time was described with the conventional annual percentage change (cAPC). The survival of cases diagnosed (2000-2004) was estimated using Kaplan-Meier method. A total of 1022 in adolescents and 1396 in young adults were diagnosed. Overall incidence rates were 219.4/10(6) in 15-19 year olds and 293.1/10(6) in 20-24 year olds. The most frequently diagnosed cancers in male AYA were malignant gonadal germ-cell tumors and Hodgkin's disease, and were melanoma, thyroid carcinoma, and Hodgkin's disease in females. The age-standardized rates appeared stable over time in AYA, with a cAPC of +2.0% (P = 0.68). The 5-year overall survival for all cancers was different between genders and age groups, with 78.8% (95%CI: 75.6-82.0) for males and 85.2% (95%CI: 82.2-88.1) for females (P = 0.01), and 78.5% (95%CI: 75.0-82.1) in 15-19 year olds and 84.3% (95% CI: 81.6-87.0) in 20-24 year olds (P = 0.02). Noteworthy, the frequency and the distribution of tumor types in AYA are unique and different from the observed at any other age group. Survival in French AYA has improved over time. Epidemiological data might reflect major trends in the risk factors and preventive interventions. Thus, further research into etiology of cancers affecting AYA should become key priorities for cancer control among AYA.

National estimate of multiple sclerosis incidence in France (2001–2007)

Background:

In France, the incidence of multiple sclerosis (MS) is not well known, and MS is one of the 30 long-term illnesses for which patients are covered for 100% of their health care costs.

Objective:

To estimate the incidence of MS in France and its geographic variations.

Methods:

We estimated the national rate for notification of MS to the main French health insurance system, and its confidence interval (CI), between November 2000 and October 2007, which covers 87% of the population. We analysed geographic variations using a Bayesian approach.

Results:

Between November 2000 and October 2007, among a covered population of 52,449,871, some 28,682 individuals were registered as having MS. After age standardization according to the European population, the notification rate for MS was 6.8 per 100,000 (6.7–6.9), 9.8 (9.7–10.0) in women and 3.7 (3.6–3.8) in men. When the under-notification rate (11.5% and 29%) was taken into account, the notification rate per 100,000 inhabitants was estimated between 7.6 and 8.8. The notification rate was higher in north-eastern France, and lower on the Atlantic coast and in the Alps as well as on both sides of the Rhône River.

Conclusions:

This study, conducted on a representative French population, provides for the first time national estimates of MS incidence between November 2000 and October 2007.

Left-sided predominance of hypodontia irrespective of cleft sidedness in a French population

OBJECTIVE

Individuals with oral clefts exhibit considerably more dental anomalies than do individuals without clefts. Our aim was to evaluate the prevalence of tooth agenesis in a sample composed of 124 children (81 boys and 43 girls, mean age 12.5 years) with clefts registered with the Cleft Palate Center in Strasbourg (France).

DESIGN

Cleft types and dental agenesis were assessed by clinical and radiographic examination. Cleft types were divided into four groups according to the clinical extent of the cleft (cleft lip only [CL, 12.9%], cleft lip and alveolus [CLA, 4%], cleft lip and palate [CLP, 49.2%], and cleft palate only [CP, 33.9%]).

RESULTS

Of the subjects 63% had evidence of hypodontia: maxillary lateral incisors (54%) and upper and lower premolars (32%) were the most common missing teeth. The number of dental ageneses associated with CP (54%) and CLP (79%) was significantly higher than that with CL (33%). All these anomalies were found in proportionately higher frequencies as the severity of the cleft increased, and we found left side predominance for hypodontia (p < .01) irrespective of cleft sidedness.

CONCLUSIONS

Both right-sided and left-sided clefts were more frequently correlated with left-sided dental agenesis. This left-sided prevalence suggests that common signaling malfunctions might be involved, both in dental development anomalies and cleft genesis.

Risk for non Hodgkin's lymphoma in the vicinity of French municipal solid waste incinerators

BACKGROUND

Dioxin emissions from municipal solid waste incinerators are one of the major sources of dioxins and therefore are an exposure source of public concern. There is growing epidemiologic evidence of an increased risk for non-Hodgkin's lymphoma (NHL) in the vicinity of some municipal solid waste incinerators with high dioxin emission levels. The purpose of this study was to examine this association on a larger population scale.

METHODS

The study area consisted of four French administrative departments, comprising a total of 2270 block groups. NHL cases that had been diagnosed during the period 1990-1999, and were aged 15 years and over, were considered. Each case was assigned a block group by residential address geocoding. Atmospheric Dispersion Model System software was used to estimate immissions in the surroundings of 13 incinerators which operated in the study area. Then, cumulative ground-level dioxin concentrations were calculated for each block group. Poisson multiple regression models, incorporating penalized regression splines to control for covariates and dealing with Poisson overdispersion, were used. Five confounding factors were considered: population density, urbanisation, socio-economic level, airborne traffic pollution, and industrial pollution.

RESULTS

A total of 3974 NHL incident cases was observed (2147 among males, and 1827 among females) during the 1990-1999 time period. A statistically significant relationship was found at the block group level between risk for NHL and dioxin exposure, with a relative risk (RR) of 1.120 (95% confidence interval [CI] 1.002 - 1.251) for persons living in highly exposed census blocks compared to those living in slightly exposed block groups. Population density appeared positively linked both to risk for NHL and dioxin exposure. Subgroup multivariate analyses per gender yielded a significant RR for females only (RR = 1.178, 95% CI 1.013 - 1.369).

CONCLUSION

This study, in line with previous results obtained in the vicinity of the incinerator located in Besançon (France), adds further evidence to the link between NHL incidence and exposure to dioxins emitted by municipal solid waste incinerators. However, the findings of this study cannot be extrapolated to current incinerators, which emit lower amounts of pollutants.

[Cancer incidence and survival among adolescents and young adults in France (1978-1997)]

Malignancies are rare young French adults but represent the third significant cause of death in the cohort of 15-24 years of age. The aim of this study was to investigate incidence and survival rates of French adolescents and young adults with cancer. All cases of cancer occuring over a 20-year period (1978-1997) in the cohort of patients aged 15 to 24, were obtained from nine population-based registries (10 % of the French population). Basal cell carcinomas of the skin were excluded. 1161 and 1884 cases were recorded in adolescents and young adults, respectively. Overall incidence rates (IR) were 161.4/10(6) in adolescents aged 15-19 years (M/F ratio = 1.3), and 252.6/10(6) in young adults aged 20-24 years (M/F ratio : 1.2). During the 1978-97 period, the IRs appeared stable over the years, +0.4 % [CI95 % = -2.3 ; +3.1] (p = 0.79) for adolescents and +1.7 % [CI95 % = -4.0 ; +7.3] (p = 0.57) for young adults. Five-year overall survival rates were 69.1 % [CI95 % = 66.4-71.8] for adolescents and 74.5 % [CI95 % = 72.3-76.7] for young adults. The 5-year survival rate for patients 15-24 years improved from 62.0 % (CI95 % = 57.5-66.5) in 1978-82 to 80.2 % (IC95 % = 77.7-82.8) in 1993-97. Noteworthy, results in adolescents and young adults are poor compared to the ones from their younger counterparts, especially in patients with acute lymphoblastic leukemia, non-Hodgkin lymphoma, Ewing's sarcoma, osteosarcoma, rhabdomyosarcoma, and astrocytoma. Further studies are warranted to elucidate whether these differences are due to intrinsic biological properties of the tumor or to differences in clinical practices in the two populations.

Socioeconomic and geographic determinants of survival of patients with digestive cancer in France

Using a multilevel Cox model, the association between socioeconomic and geographical aggregate variables and survival was investigated in 81 268 patients with digestive tract cancer diagnosed in the years 1980-1997 and registered in 12 registries in the French Network of Cancer Registries. This association differed according to cancer site: it was clear for colon (relative risk (RR)=1.10 (1.04-1.16), 1.10 (1.04-1.16) and 1.14 (1.05-1.23), respectively, for distances to nearest reference cancer care centre between 10 and 30, 30 and 50 and more than 90 km, in comparison with distance of less than 10 km; P-trend=0.003) and rectal cancer (RR=1.09 (1.03-1.15), RR=1.08 (1.02-1.14) and RR=1.12 (1.05-1.19), respectively, for distances between 10 and 30 km, 30 and 50 km and 50 and 70 km, P-trend=0.024) (n=28 010 and n=18 080, respectively) but was not significant for gall bladder and biliary tract cancer (n=2893) or small intestine cancer (n=1038). Even though the influence of socioeconomic status on prognosis is modest compared to clinical prognostic factors such as histology or stage at diagnosis, socioeconomic deprivation and distance to nearest cancer centre need to be considered as potential survival predictors in digestive tract cancer.

Combining expert ratings and exposure measurements: a random effect paradigm

The aim of this paper is to present a paradigm for combining ordinal expert ratings with exposure measurements while accounting for a between-worker effect when estimating exposure group (EG)-specific means for epidemiological purposes. Expert judgement is used to classify the EGs into a limited number of exposure levels independently of the exposure measurements. The mean exposure of each EG is considered to be a random deviate from a central exposure rating-specific value. Combining this approach with the standard between-worker random effect model, we obtain a nested two-way model. Using Gibbs sampling, we can fit such models incorporating prior information on components of variance and modelling options to the rating-specific means. An approximate formula is presented estimating the mean exposure of each EG as a function of the geometric mean of the measurements in this EG, between and within EG standard deviations and the overall geometric mean, thus borrowing information from other EGs. We apply this paradigm to an actual data set of dust exposure measurements in a steel producing factory. Some EG-specific means are quite different from the estimates including the ratings. Rating-specific means could be estimated under different hypotheses. It is argued that when setting up an expert rating of exposures it is best done independently of existing exposure measurements. The present model is then a convenient framework in which to combine the two sources of information.

Dose-response slope of forced oscillation and forced expiratory parameters in bronchial challenge testing

In population studies, the provocative dose (PD) of bronchoconstrictor causing a significant decrement in lung function cannot be calculated for most subjects. Dose-response curves for carbachol were examined to determine whether this relationship can be summarized by means of a continuous index likely to be calculable for all subjects, namely the two-point dose response slope (DRS) of mean resistance (Rm) and resistance at 10 Hz (R10) measured by the forced oscillation technique (FOT). Five doses of carbachol (320 microg each) were inhaled by 71 patients referred for investigation of asthma (n=16), chronic cough (n=15), nasal polyposis (n=8), chronic rhinitis (n=8), dyspnoea (n=8), urticaria (n=5), post-anaphylactic shock (n=4) and miscellaneous conditions (n=7). FOT resistance and forced expiratory volume in one second (FEV1) were measured in close succession. The PD of carbachol leading to a fall in FEV1 > or = 20% (PD20) or a rise in Rm or R10 > or = 47% (PD47,Rm and PD47,R10) were calculated by interpolation. DRS for FEV1 (DRSFEV1), Rm (DRSRm) and R10 (DRSR10) were obtained as the percentage change at last dose divided by the total dose of carbachol. The sensitivity (Se) and specificity (Sp) of DRSRm, DRS10 delta%Rm and delta%R10 in detecting spirometric bronchial hyperresponsiveness (BHR, fall in FEV1 > or = 20%) were assessed by receiver operating characteristic (ROC) curves. There were 23 (32%) "spirometric" reactors. PD20 correlated strongly with DRSFEV1 (r=-0.962; p=0.0001); PD47,Rm correlated significantly with DRSRm (r=-0.648; p=0.0001) and PD47,R10 with DRSR10 (r=-0.552; p=0.0001). DRSFEV1 correlated significantly with both DRSRm (r=0.700; p=0.0001) and DRSR10 (r=0.784; p=0.0001). The Se and Sp of the various FOT indices to correctly detect spirometric BHR were as follows: DRSRm: Se=91.3%, Sp=81.2%; DRSR10: Se=91.3%, Sp=95.8%; delta%Rm: Se=86.9%, Sp=52.1%; and delta%R10: Se=91.3%, Sp=58.3%. Dose-response slopes of indices of forced oscillation technique resistance, especially the dose-response slope of resistance at 10Hz are proposed as simple quantitative indices of bronchial responsiveness which can be calculated for all subjects and that may be useful in occupational epidemiology.

Sick-building symptoms in office workers in northeastern France: a pilot study

STUDY OBJECTIVES

To verify that sick building symptoms are present in north-eastern France office workers; to try to identify new confounding factors.

METHODS

The design was that of a cross-sectional study with control group. We studied with the same methods the personnel of an air-conditioned building (n=425), and of a naturally ventilated building (n=351). Air temperature and humidity, bacterial and fungal densities were measured by the same technical staff in the two buildings. A standard questionnaire on irritative and respiratory symptoms, personal and family history, and lifestyle was completed by the participants.

RESULTS

In univariate analysis, exposure to air-conditioning was associated with an increased prevalence of symptoms (odds ratios-OR-between 1.54 and 2.84). A significant increase in sickness absence was also found among subjects working in air-conditioned offices. As a series of factors were suspected to interfere with these associations, logistic regression was applied. This method confirmed exposure to be an independent determinant of 7 symptoms, and also identified two determinants not previously described: a family history of respiratory diseases and "do-it-yourself' activities.

IN CONCLUSION

we found the sick building symptoms to be present in a group of French office workers exposed to air-conditioning. We confirmed the influence of a number of confounding factors and described two further confounders - do-it-yourself activities at home and a history of familial respiratory disease.