Use of longitudinal data analysis in the determination of a global score for total hip replacement indication in osteoarthritis patients

BACKGROUND

A composite index would help in medical practice for considering total hip replacement (THR) in hip osteoarthritis. In association with other parameters, such as age or general status, this index could serve to objectively determine the time to THR surgery, and might be used as an outcome measure in therapeutic trials. The objective of this study was to propose such an index, taking into account symptomatic, structural and therapeutic variables, and impact on lifestyle.

METHODS

A curve-fitting procedure and a marginal model were used to highlight the average change in the variables over time. A discrete Cox model, with time-dependent covariates. was then performed to study the association between the variables and hip surgery, leading to the construction of a composite index.

RESULTS

The smoothing and fitting analysis suggested takes into account not only the variables, but also the changes in their values during the preceding year, when studying the association between variables and hip surgery. Five covariates were retained in the Cox model, among them only one concerns the changes during the preceding year, showing that the values of the covariates are more important that their changes during the preceding year for THR indication. The coefficients of the model were transformed into new coefficients, so that their sum provided a severity score with individual values 0-100.

DISCUSSION

Based on the values of reliable variables recorded during the four last visits, this score can be regarded as a valid index able to objectively help with THR indications. This score might also be used as an outcome measure in therapeutic trials.

Longitudinal study of associations between perceived health status and self reported diseases in the French Gazel cohort

STUDY OBJECTIVE

Although perceived health status is an indicator widely used in epidemiological studies, its relation to various diseases is not well known. The objective of this study is to examine these relations in detail.

DESIGN

Marginal models used for a longitudinal study of the association between three health scales and 47 diseases among 12 164 men and 44 diseases among 4415 women.

SETTING

French Gazel cohort during the period from 1991 to 1996.

MAIN RESULTS

The general health status scale was significantly associated with 43 diseases among men, and 31 among women. Some of these significantly associated diseases were physical (for example, cancer and cerebrovascular accident) and others, psychological (for example, depression). The mental fatigue scale was more specifically associated with psychological disorders, including sleep problems, depression, and nervous diseases. Moreover, modifications in subjects' assessment of their health from one year to the next were generally associated with modifications in reported diseases.

CONCLUSION

Although the mechanism that relates the presence of a disease to perceived health status remains in question, these results show clearly that there is a close association between these two domains that justifies the use of perceived health as a proxy for self reported diseases.

Defining disease activity in ankylosing spondylitis: is a combination of variables (Bath Ankylosing Spondylitis Disease Activity Index) an appropriate instrument?

OBJECTIVE

Disease activity has been defined using a self-administered instrument, focusing on fatigue, axial pain, peripheral pain, enthesopathy and morning stiffness [Bath Ankylosing Spondylitis Disease Activity Index (BASDAI)]. This validated instrument is simple and takes 40 s to complete, but whether the index is an accurate expression of the component parts, or whether additional weighting would enhance its efficacy, is unclear.

METHODS

Four hundred and seventy-three patients with ankylosing spondylitis received placebo or active non-steroidal anti-inflammatory drug (NSAID) for 6 weeks, and changes between entry and completion were captured by BASDAI and the individual components. Principle component analysis (PCA) was used to explore the best combinations of variables in decreasing order of explained total dispersion and to assess whether a single sum (or algebraic expression) best defined disease activity status.

RESULTS

At entry, the correlation between BASDAI and the first axis was 0.99, 0.11 with the second, and zero thereafter. Data at study end and relating to change revealed a 100% correlation (R = 1) between the first axis and the sum, with zero for the remainder.

CONCLUSIONS

The data support BASDAI as being a valid and appropriate composite to define disease activity in ankylosing spondylitis. Developed as a simple sum of its components, BASDAI has excellent content validity.

Outcome variables in ankylosing spondylitis: evaluation of their relevance and discriminant capacity

The clinical status of ankylosing spondylitis (AS) can be defined by several domains (e.g., pain, function, metrology, laboratory) and subcomponents within each domain (e.g., pain using visual analog scale, Schober's within metrology). Our aim was (1) To define groups of highly correlated variables in order to determine the most relevant; and (2) to evaluate the capacity of different clinical and biological variables that best discriminate between placebo and active nonsteroidal drugs in AS. Patients with active AS (n=423) were followed prospectively over 6 weeks while receiving placebo (n=121) or active nonsteroidal antiinflammatory drugs (n=352). Eighteen variables were studied, including global assessment, pain, stiffness, functional indices, metrology, disease activity index, and laboratory markers. Statistics included (1) Evaluation of the relevance of the different domains by multivariate analysis (CART tree-structure classification; variable clustering); and (2) evaluation of the discriminant capacity by univariate analysis [i.e., differences in the standardized response mean (SRM) (mean change/SD) between placebo and active drug. A value > or =0.60 was considered relevant]. Four clusters were identified (patient's subjective perception, inflammatory symptoms, metrology, laboratory data) with multiple correlation R2 revealing the most relevant variables to be the Bath Ankylosing Spondylitis Functional Index (BASFI; 0.75), night pain (0.62), Schober's test (0.58), and platelet count (0.55), respectively, within each cluster. In terms of discriminant power (SRM) the patient perceived global status (0.84), lumbar pain (0.73), night pain (0.71), physician global assessment (0.66), and BASFI (0.65) were most relevant in the univariate analysis. Among the 4 most relevant domains are subjective perception, inflammatory symptoms, metrology, and laboratory. Multivariate analysis of the data reveals that the spinal pain and the patient global assessment are the variables that best discriminate between placebo and active nonsteroidal drug in short term studies.

Clinical relevance of C-reactive protein in axial involvement of ankylosing spondylitis

To evaluate C-reactive protein (CRP) as a potential useful criterion of symptomatic severity of ankylosing spondylitis (AS), we conducted both a cross sectional and a longitudinal (6 week) clinical study in 443 patients with axial involvement in AS. During the 6 weeks of the study, patients received either a placebo or an active nonsteroidal antiinflammatory drug (NSAID). At baseline, CRP was increased in 173 patients (39%). A multivariate analysis in which CRP was the dependent variable and all clinical assessment criteria (pain, range of motion, functional disability, hemoglobin, platelet count) the independent variables showed that range of motion and laboratory signs of inflammation were the most significant variables to explain the CRP values. A similar multivariate analysis conducted on the changes in the variables during the 6 weeks of the study concluded that night pain and laboratory signs of inflammation were the most significant variables explaining the changes in CRP values. The capacity of CRP to discriminate between an active NSAID and a placebo was moderate. This study suggests than an increase in CRP in patients with AS with axial involvement is not a rare phenomenon and might be correlated with the clinical severity of the disease. This outcome measure does not seem to be of great interest in the short term evaluation of fast acting drugs. However, the longterm clinical significance of such an increase in CRP remains to be investigated.

Ankylosing spondylitis: what is the optimum duration of a clinical study? A one year versus a 6 weeks non-steroidal anti-inflammatory drug trial

OBJECTIVE

To consider the relevance of the duration of a clinical trial in ankylosing spondylitis: long-term (i.e. 1 yr) vs short-term (i.e. 6 weeks) assessment of a non-steroidal anti-inflammatory drug (NSAID)-placebo controlled study.

METHODS

The design was a prospective, multicentre, double-blind, placebo-controlled study of 6 weeks duration with a 12 months double-blind extension. Study drugs were placebo (n = 121) or active NSAID (n = 352). A decrease of at least 50% in pain and/or global assessment and/or functional impairment during the study defined the response to treatment. The percentage of patients discontinuing the study drug over time (life table analysis) permitted the evaluation of both the efficacy and toxicity.

RESULTS

Among the 473 recruited patients, the percentage of responders was similar at 1 yr and week 6 with a highly statistically significant difference in favour of the active NSAID groups when compared to placebo (at 1 yr, 17% in the placebo group vs 37, 50 and 43% in the piroxicam 20 mg, meloxicam 15 mg and meloxicam 22.5 mg, respectively, for the patient's overall assessment) without any statistically significant difference between the three active groups. However, evaluation of the patients discontinuing the study drug during the 1 yr of the study permitted the detection of a statistically significant difference between the active NSAID groups. A lower percentage of patients taking meloxicam 22.5 mg had to discontinue the study drug when compared to either meloxicam 15 mg or piroxicam 20 mg (37% vs 53% and 53%, respectively, P < 0.05). By 52 weeks, drug-related upper gastrointestinal adverse events occurred in 13, 32, 20 and 18% in the placebo, piroxicam 20 mg, meloxicam 15 mg and meloxicam 22.5 mg groups, respectively. Some of the adverse events occurred only after week 6.

CONCLUSION

This study suggests that a 1 yr trial might be of optimum value compared to a 6 week assessment in order to define better the efficacy and tolerability of NSAIDs in ankylosing spondylitis.

Criteria for generalized and focal osteoarthritis

STUDY OBJECTIVE

to identify homogeneous, distinct subgroups of osteoarthritis patients based on distribution of osteoarthritis lesions.

PATIENTS-METHODS

STUDY DESIGN

cross-sectional, prospective, multicenter. Patients with osteoarthritis of the hip, knee, fingers, or spine.

DATA COLLECTED

clinical and radiologic features at 41 joint sites.

METHODS

multivariate statistical analysis including a) k-mean clustering analysis followed by ascending hierarchical classification b) and a tree-structured discriminant method to confirm and to characterize the subgroups obtained using the clustering method.

RESULTS

The 1021 patients were first classified into five categories with an error rate (obtained by cross-validation) of 7.6%. The tree obtained by segmentation took into account manifestations of osteoarthritis at the hands, knees and spine. Irrespective of the reason for seeking medical advice, patients with bilateral involvement of the fingers or with involvement of the spine and both femorotibial joints were classified as having generalized osteoarthritis; in contrast, focal osteoarthritis was defined as the absence of involvement of the fingers and of either the spine or the knees.

CONCLUSION

The statistical analysis provided a classification system that would be easy to use in everyday clinical practice. Prospective studies are needed to evaluate the potential clinical relevance of this system.

Evaluation of screening methods for Down's syndrome using bootstrap comparison of ROC curves

This paper concerns the prediction of fetal Down's syndrome in pregnant women. Down's syndrome is the most common congenital cause of severe mental retardation. We elaborate two predictive functions of trisomy 21, combining maternal age and a maternal serum marker. We evaluated them by means of receiver operating characteristic (ROC) curves which give a representation of sensitivity and specificity of a prediction model when varying the cutoff of the predictor on the whole spectrum. Since normal statistical methods for comparison of ROC curves rely on distributional assumptions which were not verified, we used bootstrapping of ROC curves as a check for the statistical significance of differences between the areas under the curves.

Longitudinal radiologic evaluation of osteoarthritis of the knee

Radiologic assessment appears to be an objective standard for longterm evaluation of osteoarthritis (OA) and it is inexpensive, quick, simple and noninvasive. We conducted a one year followup study of patients with OA of the knee to evaluate the reproducibility and validity of this evaluation. The intra and interobserver reproducibility of the radiologic variables were evaluated on 275 and 539 patients, respectively, and found to be satisfactory (intraclass coefficient of correlation above 0.70 for the evaluation of the space narrowing of the joint). During the one year followup study, there was a slight but statistically significant deterioration of the joint space narrowing, evaluated on a 6 grade scale on 360 patients (p less than 0.001). The changes in the joint space narrowing were more closely correlated with treatments received by the patients for OA (nonsteroidal antiinflammatory drug (NSAID) intake, synovial fluid aspiration) than with changes in the recorded clinical variables (pain on a visual analog scale, Lequesne algofunctional index). Moreover, some factors such as obesity, generalized OA and flares of OA appeared to be correlated with the deterioration of joint space narrowing. Further studies are necessary to confirm and/or explain the relationship between the deterioration of the joint space narrowing and such factors (i.e., NSAID intake, obesity, flares of OA, generalized OA), which were detected in this study.

[Mortality of dilated myocardiopathies as a function of continuation of alcohol drinking. Multivariate analysis concerning 236 patients]

A retrospective study was conducted in 236 consecutive patients with dilated cardiomyopathy to determine the characteristics of the disease in heavy alcohol drinkers (n = 110) and to study its outcome, using a Cox model, according to whether the patients were abstainers (n = 49) or continued to indulge drinking (n = 61). At the time of diagnosis, pulmonary pressures were higher in heavy drinkers than in abstainers. During a mean follow-up period of 39 +/- 27 months, 80 patients died of heart disease, viz.: 46 out of 127 non-alcoholic patients (36.2 per cent), 31 out of 61 heavy drinkers (50.8 per cent) and 3 out of 49 patients who had given up alcohol (6.12 per cent). Thus, independently of other parameters, abstinence is a highly significant (P less than 0.001) factor of favourable prognosis in dilated cardiomyopathy.

Evaluation of a functional index and an articular index in ankylosing spondylitis

We describe an index of functional impairment and a system of scoring joint tenderness for use in the assessment of ankylosing spondylitis. The functional index consists of 20 questions and the articular index is based on the scoring of a total of 10 joint responses after movement or firm digital pressure. These indices are simple to establish and not time consuming. They have a high degree of intra- and interobserver reproducibility. The indices showed changes in short term clinical trials of antiinflammatory drugs; these changes were highly correlated with the patient's overall assessment of his own clinical condition.

Plasma ferritin in old age. Influence of biological and pathological factors in a large elderly population

The effects of biological (age, sex, weight) and pathological factors on plasma ferritin concentrations were documented in 776 unselected elderly patients aged 80.9 +/- 9.7 yr. A marked shift towards high values (159 +/- 142 micrograms/l) was observed in this elderly population together with the persistence of the well-known sex-related difference in ferritin levels (higher levels in men). Twenty-five percent of the population had high levels of ferritin (greater than or equal to 220 micrograms/l) but 75% of these high values (i.e. 18.5% of the population) could be readily explained by their known association with a particular pathology (inflammatory syndrome, renal failure, cardiovascular diseases, alcoholism). Only 6% of the population had unexplained high ferritin concentrations. Therefore, our data strongly suggest that the repeatedly reported increase of ferritin in the aged population is merely related to an age-associated pathology and may not be a normal physiological event occurring during the process of aging.

Biological and pathological factors affecting plasma gamma-glutamyl transpeptidase and alkaline phosphatase activity in the elderly

A high percentage of abnormally elevated plasma alkaline phosphatase (AP) and gamma-glutamyltranspeptidase (GGT) was found in a homogeneous population of 741 elderly subjects. The respective role of pathological and biological factors (age, sex, weight) upon the two plasma enzymatic activities has been analyzed. In disagreement with previous reports, it was observed that abnormal AP and GGT values could be readily explained by their association with specific diseases. The apparent relationship between abnormal levels of plasma AP and GGT activities and age, sex or weight reflected merely a different distribution of the pathology according to these biological parameters. In the absence of non-specific elevations of plasma AP and GGT activities with age, the usual reference values for these tests, although established on younger populations, still apply to the aged.

Age- and sex-associated modification of plasma melatonin concentrations in man. Relationship to pathology, malignant or not, and autopsy findings

Plasma melatonin concentrations were determined in 757 unselected elderly patients aged 80.9 +/- 9.7 years. The daytime (8-9.30 h) plasma levels of melatonin were in the so-called normal range in only one third of the population whereas 65% of the subjects had abnormal levels of the hormone, most often decreased i.e. less than 0.17 nmol/l (53%) and sometimes increased i.e. 0.43 nmol/l or higher (12%). A control group of healthy elderly subjects showed the same distribution as the entire population. A sex-difference with significantly higher levels of plasma melatonin in elderly women was observed. With respect to pathology and autopsy findings high levels of the hormone correlated with cancer, chronic renal failure, cardiovascular disease, biological inflammatory syndrome and diabetes. Low levels correlated with neurologic disease, tobacco or alcohol addiction. However, some of these relations were found to be sex-related as they were observed in women but not in men. Our data indicate that pineal function seems to be often altered in elderly human subjects and suggest potential diagnostic applications of melatonin determination.

[Measurement of arterial pressure in children. Influence of the resting time, inflatable surface and type of arm-band]

Measurements of arterial pressure in children should be standardized as much as possible. The influence on readings of three factors--pre-measurement resting time, area of the inflatable surface and type of cuff--was investigated in 151 children divided into several homogeneous groups. The study showed that reliable values can only be surface area, it would appear that no modern apparatus provides accurate readings for all age groups. The cuff itself is of little importance, but diastolic pressure values are slightly increased when cuffs with an unprotected bladder are used.

Cumulative effects of age and pathology on plasma carcinoembryonic antigen in an unselected elderly population

Plasma CEA was measured in 776 institutionalized elderly patients, 593 women and 183 men, whose mean age was 80.8 +/- 9.7 yr. Of these subjects 39 were found to be free of any pathology and were considered as a control group. All the other subjects had a variety of diseases, malignant or not, known or not known to be associated with elevated CEA. This study showed that elderly people in their 80s, in apparent good health, had higher levels of CEA (3.0 +/- 1.4 ng/ml) than younger people. These levels were shown to be significantly increased in a large number of non-malignant diseases. No correlation could be found between elevated CEA values and autopsy finding or drug administration. Using the chi square test, the number of patients with CEA levels greater than 5 ng/ml was found to be significantly higher in chronic renal failure and cancer. Mortality was also found to be higher in patients with CEA levels greater than 5 ng/ml during a 25-month follow-up. Higher levels of CEA in elderly subjects in apparent good health, as well as increase of these levels in a large set of non-malignant pathology, must thus be kept in mind when interpreting data arising from old people because of both the false-positive (healthy elderly subjects) and the cumulative effects of polypathology on plasma CEA.

DISBAR: a program for two-group discriminant analysis on mixed variables after binary coding

The program DISBAR is particularly adapted to medical data which often concern several measurements of different types (mixed variables) collected on patients divided into 2 groups according to a binary response variable such as survival or non-survival. DISBAR leads to a simple geometrical representation of the observations providing an easy allocation rule for a new case. The computer time required to run the program DISBAR is negligible in comparison with the one necessary to perform any other discriminant analysis of binary variables. DISBAR combines in practice computational facility with good results and easy interpretation.

[Value of mediastinoscopy and mediastinotomy in exploration of the mediastinum (author's transl)]

The value of the information provided by mediastinoscopy or cervical mediastinotomy was evaluated in 150 cases by means of tests measuring diagnostic efficiency with calculated agreement coefficient. An accurate histological diagnosis was made in 100 out of 116 mediastinal syndromes (adenopathy, mediastinal or paramediastinal tumors). In cases with negative results exploratory surgery is required. The operability of bronchial carcinoma was assessed in 34 cases selected for their location and radiological extension. A negative mediastinoscopy makes it very unlikely that an inoperable cancer will be found. In bronchial carcinoma, mediastinoscopy is recommended whenever the risk attending excision of a tumor with lymph node involvement is to be weighed against the high operative risk in a debilitated patient. The usefulness of these exploratory procedures is enhanced by anterior mediastinotomy in cases with anterior tumor and in some left-sided bronchial cancers.

Two methods of discrimination on qualitative variables

The methods of discrimination on qualitative variables are generally based on the estimation of density distribution in each group yk (k = 1, 2, ..., K). Two kinds of estimation are proposed and the correspondent computer program is presented. The results of a medical example are compared with those obtained by using a logistic model used by different authors when dealing with discrimination problems.

Discriminant analysis: methods and program

Because of the disparity in the results obtained using different methods of discriminant analysis, we have written and used a unique program to test them. These methods can be divided into two groups corresponding to a classical probabilistic approach (i.e., Bayesian methods) or a topological approach (i.e., Sebestyen method).

[Local tolerance of subcutaneous heparin. Double-blind comparison of calcium and sodium heparins (author's transl)]

One hundred and twelve subcutaneous injections of concentrated heparin (25,000 IU/ml), half in sodium heparin and half in calciparine, both with an identical heparin cation, were given in doses varying from 0.2 ml to 0.5 ml by a single operator in a double-blind study to assess the comparative local tolerance of the two heparin salts. The occurrence of bruising at the point of injection, more frequent with sodium heparin than with Calciparine, proved to be the distinguishing factor. Overall after 48 hours, 31 hematomas were observed with the sodium salt versus 20 with calciparine. McNemar's test of comparing matched proportions confirms this result (p = 0.03) and, with the order of the two treatments taken into consideration, J.J. Gart's exact test gives p = 0.0101.

[Criteria for the evaluation of the antirheumatic (antiinflammatory) drugs in man]

Comparison of 5 antiinflammatory drugs and placebo in a double blind, short term, cross-over trial conducted in patients suffering from rheumatoid arthritis. 15 measurements were used and studied. With classical unidimensional statistical tests it was not possible to discriminate active drugs between them. Multidimensional analysis (correspondance analysis and discriminant analysis) are more sensitive and allow us to distinguish active drugs with only 11 criterion.