Search

A retrospective study was planned in the Hérault (Mediterranean) region of France where bladder cancer mortality and incidence rates are high. In the present paper, variations in bladder cancer risk according to various smoking-related variables, in particular time of exposure and type of tobacco, are examined. This case-control study with 219 male incident cases and 794 male population controls randomized from electoral rolls was carried out in 1987-89. Trained interviewers obtained information on demographics, dietary habits (coffee, alcohol, artificial sweeteners, vegetables, spices, etc.), occupational exposures and detailed history of tobacco smoking (average number of cigarettes per day, number of years of smoking, age at which they began and/or quitted smoking, use of filter-tip and type of tobacco). The odds ratio (OR) for cigarette smokers versus non-smokers was greater than 5. Results for number of cigarettes daily, duration of smoking and lifetime smoking showed a highly significant dose-response relationship, which was confirmed when these variables were treated as continuous in a logistic regression model. Eighty-eight percent of the smokers used black tobacco. Quitting smoking did not result in a significant reduction in bladder cancer risk. Higher risks were associated with starting to smoke at an early age (OR before age 13 versus after age 21 = 3.42; 95% CI 1.07-10.9) and with black tobacco smoking (OR black versus blond = 1.63; 95% CI 0.73-3.64). Results suggest that black tobacco may be more harmful than blond tobacco and may have an early non-reversible role in bladder carcinogenesis.

This paper describes an automatic procedure for morphosemantic analysis and translation of compound medical terms. This analysis is of interest for the automatic indexation of medical discharge reports and summaries. Since words with the suffix -osis may have many different semantic interpretations, such -osis forms are taken as examples for a general method that avoids the difficulties in interpreting medical terms as reported in other studies.

This paper claims that, when evaluation of information systems is at stake, it is impossible in most cases to discard human subjectivity in the same way that one cannot discard quality of life when a therapeutic strategy is to be evaluated.

A meta-analysis of 10 randomized trials comparing steroid therapy with placebo or abstention was performed. Trials were pooled for severity of acute alcoholic hepatitis and time of analysis. Papers with survival analyses at one and three months were included. The robustness of this meta-analysis was analysed after exclusion of trials analysing survival at three months only, trials including mild forms, or atypical trials. Comparisons were made use the Der Simonian-Laird and Mantel-Haenszel-Peto methods. All results were statistically significant. However, if only one medium size trial were to be added, the results would no longer be significant. Taking into account the complications of steroid therapy, we conclude that further clinical trials are necessary to confirm the effectiveness of steroid therapy, especially for the group of patients with severe alcoholic hepatitis.

A case-control investigation about bladder cancer was carried out in Herault department with 219 incident cases from January 1987 to May 1989 and two control groups: 196 hospital controls randomized from hospital admission lists and 794 population controls randomized from census data. According to the comparison for those two groups, there are few socio-demographic differences; but hospital controls are more exposed to smoking, coffee, alcohol, artificial sweetners. So, the odds ratios with those hospital controls are under-estimated. We can evaluate the resulting biais.

Two GnRH agonist long-stimulation protocols, in association with HMG, are compared on a random basis. Group A (n = 53) was monitored daily from the 6th day of stimulation, whereas monitoring in group B (n = 55) began on the 11th day of stimulation. There was no significant difference in the numbers of follicles which matured, the numbers of collected oocytes, the numbers of embryos obtained in vitro and the clinical pregnancy rate. But the power of such a study is very weak (6%). It would have been necessary to include greater than 1000 patients in each group to obtain a bioequivalence test. Such a study is unrealistic for a single centre and multicentric studies are very difficult to achieve because of practical difficulties. A pooled analysis is perhaps the methodological answer.

Two hundred and nineteen patients with resected lung carcinoma were randomized 3 weeks after surgery between two treatment arms: a control group (110 cases) and an immunotherapy group (109 cases). The immunostimulant was a nonviable saprophytic mycobacterium, M. smegmatis, given monthly by subcutaneous injection in four sites. The two groups were equivalent in terms of prognostic factors, including a nonsignificant difference favoring the control group based on the N (node) classification. This interim analysis was carried out on June 1, 1981. Treatment comparison by the log-rank test did not show any significant differences between these two groups in regards to disease-free interval and overall survival. There was no significant difference between the two groups after stratification of the comparison according to the N classification or adjustment with a subset of eight prognostic parameters through the Cox model. The initially expected difference (20% 1-year survival) will probably not be achieved, given these interim results, but patients will continue to be treated and followed-up according to the protocol as to allow further evaluation of this nonspecific immunotherapy.

This article presents an analysis of the method used in an experiment in computer-aided diagnosis, applied to the whole field of neurology.While the method is based on the classical Theorem of Bayes, it nevertheless contains a number of new elements:— By means of a self-learning procedure, it makes use of the patient sample as it grows in size;— there is an overall analysis of results as well as an individual analysis of each diagnosis.

The knowledge of human polymorphism provides an efficient tool for parenthood diagnosis. This paper describes an algorithm to diagnose fatherhood within a family. In genetic research, the validation of human families is useful since at least 7% of the families have to be rejected from such studies. First, the algorithm verfies whether the phenotype of the presumed father is compatible with the genotypes of all children under the assumption of assured motherhood. An original index coding allows us to write simple algorithms: automatic creation of genetic elements, search of compatible genotypes of parents and children. Finally, this algorithm calculates several indices used to estimate the likehood of the paternity of non-excluded man.

The fact that I have used for this article the title proposed by the organizers of this conference does not mean that I agree with the philosophy which inspires it: a philosophy of means instead of a philosophy of finalities. Indeed, the use of computers in health care is not unavoidable simply because computers do exist. This is a mistake that has frequently been made since the beginning, and which is probably the source of most of our illusions and delusions. Computers are tools, and no more, and we must and can resist the ‘instrumental reason’ criticized in depth by philosophers such as R. Garaudy [1], M. Henry [2], E. Morin [3], and scientists like J. Weizenbaum [4]. I would prefer to reverse the question of how to use computers in health care and answer the following ones: What are the main information problems in medicine and health? and What is the part of computer technology in their solution? Connected to these questions is the consideration of the mental, intellectual, cultural, and even ethical impact of the machinery on its designers and users. With regard to the latter, I will mention the benefits of a better methodological education and the threat of an excessive rationalistic drift. This ethical consideration will lead me back to a brief discussion of ‘instrumental reason’. This discussion will thus provide guidelines and caveats rather than true strategic recommendations. It will deal with attitudes more than with specific actions. Indeed, we cannot escape some kind of consideration of our humanity, and the mutual relationship between technology and our existence and actions.

Survivals of two series of CLL patients (99 from a retrospective series and 196 from a prospective series) were studied separately. The three main staging systems (Rai, Binet, Rundles) agreed well, but as far as survival is concerned, too many stages are defined. The authors performed a Cox multivariate analysis of survival in order to isolate important prognostic factors at diagnosis and to use them to build a simple three-stage classification. Thrombopenia and anemia appeared as the most important risk factors. Among the nonanemic and nonthrombopenic patients, the number of involved areas was clearly related to prognosis in the authors' two series. This study allowed the authors to propose a new classification in three prognostic groups. Group C: anemia (Hb less than 10 g) and/or thrombopenia (platelets less than 100,000/mm3); about 15% of the patients; median of 2 years. Group B: no anemia, no thrombopenia, three or more involved areas (counting as one each of the following: axillary, cervical, inguinal, lymph nodes, whether unilateral or bilateral, spleen and liver); about 30% of patients; median of 7 years. Group A: no anemia, no thrombopenia, less than three involved areas; about 55% of patients; the survival of this group does not seem different from that of the French population of the same age and sex distribution. This three-stage classification only requires clinical examination and routine hemogram, has a good prognostic value which was confirmed on the series of Montserrat and Rozman (146 patients), and should therefore be helpful in planning new clinical trials.

The automatic analysis program for one derivation of E.E.G. described here is made up of two parts:1.) The first part consists of the study of background rhythm. A temporal analysis is made of the signal to identify the waves of which it is composed. Bursts of rhythms are recognised. Then the principal background rhythms are determined by the analysis of histograms of amplitude and frequency. The results are given as the percentage of the total waves, the mean frequency and amplitude and the coefficient of variation of these quantities.2.) The second part is the recognition of paroxysmal patterns. By successive applications of a set of criteria the parts of the E.E.G. which may contain spikes or spike-and-waves are extracted. The criteria are:— existence of large amplitude or slow waves,— comparison of the frequency of these suspect patterns with the background frequency,— comparison of the amplitude of these waves with the mean amplitude of the E.E.G.,— study of the rapid rhythms which, are superimposed on the show waves.This analysis is concluded by the printing of the number of isolated spikes or spike-and-waves that were observed.This program, orientated towards pattern recognition, is suitable for the study of the background patterns as well as the pattern recognition of paroxysmal bursts.

Resume L'analyse multidimensionnelle des spectres E.E.G. est une methode qui permet une reduction importante des donnees fournies par l'analyse spectrale du trace. Son interet reside essentiellement dans l'etude des modifications du nuage de points selon la phase du trace enregistre ou selon les diverses derivations.

In acute lymphoblastic leukemia, the present therapeutic progress is due to a better use of the known chemotherapy: drugs association, different ways of administration, adaptation to severity… The therapeutic adaptation is evidently based on the identification of efficient pronostic factors. Taking the remission duration as a response variable, we undertook a prognostic study of 1288 children with ALL divided into 4 groups, each receiving homogeneous treatments.The pronostic factors of the 08 LAL 74 protocol (third chronological group) are more numerous than those of the two first groups and the prognostic information is more important. The therapeutic adaptation and the modification in the nature of the first relapse could be taken as a possible explanation but this explanation cannot be proved in the absence of an experimental design. The 10 LAL 76 new protocol leads us to invalidate this hypothesis and gives moreover some results in contradiction with those of the 08 LAL 74 protocol for identical therapeutic subgroups. This fact has not yet been explained. So, the analysis of these 4 groups validates only the well-known pejorative prognostic value of a leucocytosis greater than 35000 and /or the existence of a tumoral syndrom. Therefore, the present study shows the real interest of a more rigorous methodological approach : validation of prognostic results, definition of protocols in order to show the interest of treatment adaptation to severity and inclusion of larger samples of patients.

Hospital medical records meet four main requirements: patient foliow-up, communication between physicians whether inside or outside the hospital, medical research and analysis of the medical activity of departments [19].

The classical concepts of mortality and morbidity, upon which the major health indexes are based are revisited regarding their main limitations. This paper investigates the recent attempts to improve these aspects: extension of the concept of morbidity by the introduction of disability measurements and the development of a “positive health” concept. Indexes based on these concepts are described.

It is desirable to assess the value of special investigations by objective criteria. Conditional probabilities are suitable for this purpose. Sensitivity and specificity indicate the prevalence of a sign in a disease. Predictive values estimate the probability of a disease when a sign is present and the probability of absence of disease when the sign is absent. Comparison between the informational value of a sign or examination with concepts of cost or of risk to the patient is then possible. This made be used as a basis for determining a strategy for special investigations in the establishment of a diagnosis.

Survival is an objective criteria, reliable, easy to measure and easy to analyse. But this criteria, the most important in almost all therapeutic trials, is a very rough one and not all fitted to many questions raised in cancer treatment decision making. We suggest a new criteria, combining quality and quantity of survival, which is equivalent in years of survival at a maximal quality level to the time of survival at a varying level of a cancer patient. This criteria could be usefull in the majority of "palliative" clinical trials (advanced cancers, recurrences, metastatic evolution...).

The physician has always had a great interest in prognostic knowledge: enability to predict the evolution, to better know the physiopathogical mechanism and to adapt treatment according to the prognostic status. This important medical need leads to perform numerous prognostic studies, sometimes coupled with clinical trials. Our experience in this field leads us to suggest the research of new prognostic factors easy to measure in the daily activity, and, to go further in the development of statistical methods well-adapted to the explanation of more complex response variables. In fact, first, the present prognostic explanation is weak for most diseases, and second the medical reality is often over simplified. Whatever the progress may be, though, the important prognostic variability (due to sampling, environment…) requires the validation of results of prognostic studies.

The authors have designed a multicriteria framework for the analysis of health indexes in order to provide a tool for the comparison of population health status indexes and to establish a typology of these indexes regarding their use for public health. This framework allows for (a) shedding some light on the conceptualization of health underlying each index, (b) understanding the nature of the phenomenon which are effectively measured, (c) understanding the principles of the construction of the index, (d) giving some details concerning its main properties, (e) and finally understanding its possible uses. Using this multicriteria framework for analyzing and comparing 16 different population health status indexes, the authors suggest some guidelines for choosing an index regarding its future utilization (identification of priorities, resources allocation, evaluation). Finally, the limits of the utilization of health status indexes within the process of decision making in public health are underlined.

Mortality statistics are underutilized in France. Authors show that it is possible to use them for small population (125 000). Classical statistical methods are utilized. The interpretation of the results is difficult with regard to the quality of National data. Comparison data conclude that mortality rates are lower within the industry for lung and gastrointestinal disease, suicide, injuries and that three health problems are more prevalent: alcholism, injuries for young men and cancer.

Two problems are envisaged: --the relationship between the "statistical decision" and the individual decision. Long before the existence of decision-making aids, medical knowledge has always been of a statistical nature. Ever since medicine became an organised science, the treating of a patient and not just the illness has been a constant problem. At present, it is by no means certain that we have the right to apply the strategy recommended by a well organised therapeutic trial, with our eyes shut. If for no other reason, simply because the criteria of judgment of the trial may not necessarily be the criterion best adapted to the patient concerned. --how to implicate in decision and evaluation non-quantitative criteria such as the quality of survival. Having recalled the distinction made by Wood--impairment, disability, handicap--an attempt is made to specify the state of sanometric research providing physicians with health indexes that are finer than the simple Boolean criteria of Life/Death. In conclusion, the question can be raised as to how pertinent, it is to consider death as the lowest level of health, and as to what role should be left to the patient in the choice of decision-making criteria for problems directly concerning his own fate.

The value of prognostic information and some of the statistical aspects of prognostic studies are discussed. In the first part, seven areas of medical activity where knowledge of prognostic factors is used are identified: prediction of disease evolution, treatment decisions, design and analysis of clinical trials, comparison of non-randomized series of patients, assessment of additional clinical examinations and laboratory tests, understanding of the physio-pathology of disease, screening. In the second part, the nature of the data collected in prognostic studies is discussed, and a few statistical methods useful for the identification of prognostic factors are outlined. The paper concludes with an example of a prognostic study in Chronic Lymphocytic Leukemia.

This work describes an experiment to assist differential diagnosis in Neurology. It seeks to furnish an evaluation of the contribution of the commonly used laboratory tests in a retrospective study.

A quantitative analysis of the diagnostic contribution of the classical special investigation normally used in hospital neurology. Six examinations were studied: laboratory blood investigations, electroencephalogramm, lumbar puncture, optic fundus, standard skull X-ray, standard spine X-ray. The analysis involved 548 patients in whom 2219 examinations were performed with a yield of 26 "corrected" diagnosis (a diagnosis in which there was a clinical error corrected in the light of the results of these examinations).

The discovery of the ABO system is at the origin of the study of human polymorphism the latest developments of which have permitted such breakthrough as blood transfusion and organ transplants.

The domain of public health is now one of the major public concerns in industrialized as well as in less developed countries. The main reasons for this are the ever rising costs of health care, the technical and political difficulties in allocating scarce resources between the different social sectors — education environment control and so forth, health being only one of these sectors — and the great complexity of the health care system, where one can always find a great number of parameters interfering with each other, most of them being poorly known and controlled, even in the smallest subsystems such as, for instance, a hospital department.

Survivals of two series of CLL patients (99 from a retrospective series and 196 from a prospective series) were studied separately. The three main staging systems (Rai, Binet, Rundles) agreed well, but as far as survival is concerned, too many stages are defined. The authors performed a Cox multivariate analysis of survival in order to isolate important prognostic factors at diagnosis and to use them to build a simple three-stage classification. Thrombopenia and anemia appeared as the most important risk factors. Among the nonanemic and nonthrombopenic patients, the number of involved areas was clearly related to prognosis in the authors' two series. This study allowed the authors to propose a new classification in three prognostic groups. Group C: anemia (Hb less than 10 g) and/or thrombopenia (platelets less than 100,000/mm3); about 15% of the patients; median of 2 years. Group B: no anemia, no thrombopenia, three or more involved areas (counting as one each of the following: axillary, cervical, inguinal, lymph nodes, whether unilateral or bilateral, spleen and liver); about 30% of patients; median of 7 years. Group A: no anemia, no thrombopenia, less than three involved areas; about 55% of patients; the survival of this group does not seem different from that of the French population of the same age and sex distribution. This three-stage classification only requires clinical examination and routine hemogram, has a good prognostic value which was confirmed on the series of Montserrat and Rozman (146 patients), and should therefore be helpful in planning new clinical trials.

It is a truism to claim that informatics applied to health has known nearly as many delusions as hopes. If we try to sort and discriminate the cases of success (accounting, tomodensitometry..) from those with half success (hospital information systems, computer-aided decision making, computer-aided instruction), we are driven to face some fundamental problems: What is information and what is knowledge? Is it correct — and if yes, to which extent — to consider human brain as an information processing machine? Can the meaning of a text or a situation be completely enclosed inside a formal syntactic representation? Is human language uniquely a way of conveying information? What is the part of intuition in our way of behaving and solving problems intelligently? Can we restrict our mental activity to a pure rational process? Which are the weak and strong aspects of our intelligence and memory? Which are the weak and strong aspects of the computer’s ‘intelligence’ and memory?

The knowledge of human polymorphism provides an invaluable tool in the analysis of parenthood. This paper deals with the fatherhood diagnosis within a family (presumed father, mother, children) and describes a computer program, called Fratrie, which identifies parenthood rejections and computes different decisional indices. The analysis of these indices points out a neglected index, the percentage of men rejected with respect to the (mother, children) set. We propose a new decisional strategy using this index.

Medicine is a discipline of judgment and action. At each moment of his professional life, the physician must suggest decisions and actions to his patient. In order to do this, he must gather some pertinent information and extract in the most logical and the surest way arguments allowing him to achieve his objectives.