Subgroup analysis and other (mis)uses of baseline data in clinical trials

A randomized comparison of repeat stenting with balloon angioplasty in patients with in-stent restenosis

OBJECTIVES

This randomized trial compared repeat stenting with balloon angioplasty (BA) in patients with in-stent restenosis (ISR).

BACKGROUND

Stent restenosis constitutes a therapeutic challenge. Repeat coronary interventions are currently used in this setting, but the recurrence risk remains high.

METHODS

We randomly assigned 450 patients with ISR to elective stent implantation (224 patients) or conventional BA (226 patients). Primary end point was recurrent restenosis rate at six months. Secondary end points included minimal lumen diameter (MLD), prespecified subgroup analyses, and a composite of major adverse events.

RESULTS

Procedural success was similar in both groups, but in-hospital complications were more frequent in the balloon group. After the procedure MLD was larger in the stent group (2.77 +/- 0.4 vs. 2.25 +/- 0.5 mm, p < 0.001). At follow-up, MLD was larger after stenting when the in-lesion site was considered (1.69 +/- 0.8 vs. 1.54 +/- 0.7 mm, p = 0.046). However, the binary restenosis rate (38% stent group, 39% balloon group) was similar with the two strategies. One-year event-free survival (follow-up 100%) was also similar in both groups (77% stent vs. 71% balloon, p = 0.19). Nevertheless, in the prespecified subgroup of patients with large vessels (> or =3 mm) the restenosis rate (27% vs. 49%, p = 0.007) and the event-free survival (84% vs. 62%, p = 0.002) were better after repeat stenting.

CONCLUSIONS

In patients with ISR, repeat coronary stenting provided better initial angiographic results but failed to improve restenosis rate and clinical outcome when compared with BA. However, in patients with large vessels coronary stenting improved the long-term clinical and angiographic outcome.

Inhalation aromatherapy during radiotherapy: results of a placebo-controlled double-blind randomized trial

PURPOSE

To determine whether the inhalation of aromatherapy during radiotherapy reduces anxiety.

PATIENTS AND METHODS

Three hundred thirteen patients undergoing radiotherapy were randomly assigned to receive either carrier oil with fractionated oils, carrier oil only, or pure essential oils of lavender, bergamot, and cedarwood administered by inhalation concurrently with radiation treatment. Patients underwent assessment by the Hospital Anxiety and Depression Scale (HADS) and the Somatic and Psychological Health Report (SPHERE) at baseline and at treatment completion.

RESULTS

There were no significant differences in HADS depression or SPHERE scores between the randomly assigned groups. However, HADS anxiety scores were significantly lower at treatment completion in the carrier oil only group compared with either of the fragrant arms (P =.04).

CONCLUSION

Aromatherapy, as administered in this study, is not beneficial.

Analysis of 1-year vertebral fracture risk reduction data in treatments for osteoporosis

One-year vertebral fracture risk reduction from clinical trials in adults with postmenopausal or glucocorticoid-induced osteoporosis is reviewed. Data were obtained by conducting a literature search of osteoporosis medications using the MEDLINE database, bibliographies of selected citations, and recent meeting abstracts. The methodologic quality of the trials was assessed using recently published criteria for ranking evidence. In prospective analyses, the 1-year risk of new morphometric vertebral fractures was reduced by risedronate 5 mg/d in two 3-year studies in postmenopausal women with prevalent vertebral fracture, and in two 1-year studies in patients with or at risk for glucocorticoid-induced osteoporosis. The 1-year risk of clinical vertebral fractures was reduced by alendronate and raloxifene in post hoc analyses. Reduction of morphometrically identified vertebral fracture risk is a more stringent therapeutic goal than clinical vertebral fracture risk. Therefore, more weight should be given to data from studies that use the morphometry to assess vertebral fracture incidence.

Response to montelukast among subgroups of children aged 2 to 14 years with asthma

BACKGROUND

Determining who responds to asthma therapies, particularly leukotriene modifiers, continues to be explored.

OBJECTIVE

We sought to identify patient characteristics predictive of response to montelukast.

METHODS

We used data from 2 clinical trials in which children with asthma received either montelukast or placebo. Symptoms, beta-agonist use, and unanticipated health resource use caused by asthma were recorded in validated daily diaries for children 2 to 5 (n = 689) and 6 to 14 (n = 336) years old. We defined primary end points of days without asthma in 2- to 5-year-old patients (24 hours without symptoms, beta-agonist use, or asthma attack) and change in percent predicted FEV(1) in 6- to 14-year-old children. Asthma attack was defined by the use of rescue oral corticosteroids or by an unscheduled visit to a medical provider. Patients were grouped according to baseline characteristics, such as family history of asthma, personal history of allergy, frequency of asthma symptoms, eosinophilia, and concomitant use of inhaled corticosteroids or cromolyn. We examined the stratum-specific effects of montelukast on the percentage of days without asthma, change in percent predicted FEV(1), asthma attack, and a variety of secondary symptom and FEV(1) end points.

RESULTS

We did not identify characteristics that predicted response to montelukast in either preschool or 6- to 14-year-old children. These findings were consistent across all symptom and FEV(1) outcomes. There was also no differential response to montelukast in either age group when asthma attack was the outcome.

CONCLUSION

The patient characteristics studied do not appear to provide an indication of who will benefit most from treatment with montelukast.

HER2 overexpression and doxorubicin in adjuvant chemotherapy for resectable breast cancer

PURPOSE

Human epidermal growth factor receptor 2 (HER2) overexpression was found to predict a good response in breast carcinoma patients treated with doxorubicin (Adriamycin [ADM]). Evidence from our recent study indicates that node-positive patients respond to cyclophosphamide, methotrexate, and fluorouracil (CMF) regardless of HER2 status. We address the issue of whether therapy regimens including CMF and ADM versus CMF alone have the same therapeutic effect in patients with HER2+ and HER2- tumors in terms of relapse-free survival (RFS) and overall survival (OS).

METHODS

Archival specimens of the primary tumors from 506 patients in a prospective clinical trial were stained with the anti-HER2 monoclonal antibody CB11. Originally, patients were randomly allocated to receive either 12 courses of intravenous CMF or eight courses of the same regimen followed by four cycles of ADM. RFS and OS were analyzed by a Cox model taking into account treatment, HER2 status, and the interaction between treatment and HER2 status, adjusting for the effect of other known clinical and biopathologic factors.

RESULTS

Analysis of survival rates indicates a possible differential effect of treatment in the patients grouped according to HER2 status. Improved RFS and OS were observed in the HER2+ subgroup after treatment with CMF plus ADM versus CMF alone. With a median follow-up of 15 years, the hazard ratio (HR) for RFS was 0.83 in HER2+ tumors and 1.22 in HER2- tumors. The effect of treatment was more evident on OS in HER2+ patients (HR = 0.61; CI, 0.32 to 1.16) than in HER2- patients (HR = 1.26).

CONCLUSION

Our data indicate that adding ADM to CMF might be beneficial for patients with HER2+ tumors.

Drotrecogin alfa (activated) (recombinant human activated protein C) for the treatment of severe sepsis

OBJECTIVE

To review the data supporting drotrecogin alfa (activated) for severe sepsis treatment.

DATA SOURCES

Published research and data from the Protein C Worldwide Evaluation in Severe Sepsis (PROWESS) trial.

DATA EXTRACTION AND SYNTHESIS

The coagulation cascade and intense inflammation play a central role in the development of organ failure due to severe sepsis. Drotrecogin alfa (activated) has anti-inflammatory, antithrombotic, profibrinolytic, and other properties that may explain the beneficial results seen in both animal models and humans with severe sepsis. Drotrecogin alfa (activated) produces a robust reduction in the mortality rate of patients with severe sepsis that is evident across nearly every subgroup examined in the phase III clinical trial and has an acceptable safety profile with bleeding during infusion as the only significant risk associated with therapy. The relative risk reductions for mortality seen in Gram-negative, Gram-positive, pneumonia, abdominal sources, shock, and nonshock are similar to the intent-to-treat population, 19.4%. Treatment also increases days alive and free from mechanical ventilation and shock.

CONCLUSIONS

Coagulopathy and systemic inflammation are almost universal in patients with severe sepsis. Treatment of this disorder with drotrecogin alfa (activated) directly addresses these derangements and substantially reduces morbidity and mortality rates with potential for bleeding during infusion as the only known risk.

Anabolic-androgenic steroids for alcoholic liver disease

BACKGROUND

Alcohol is one of the most common causes of liver disease in the Western World today. Randomised clinical trials have examined the effects of anabolic-androgenic steroids for alcoholic liver disease.

OBJECTIVES

The objectives were to assess the beneficial and harmful effects of anabolic-androgenic steroids for patients with alcoholic liver disease based on the results of randomised clinical trials.

SEARCH STRATEGY

The Cochrane Hepato-Biliary Group Controlled Trials Register, The Cochrane Controlled Trials Register on The Cochrane Library, MEDLINE, EMBASE, and full text searches were combined (all searched December 2001). Manufacturers and researchers in the field were also contacted.

SELECTION CRITERIA

Only randomised clinical trials studying patients with alcoholic steatosis, alcoholic fibrosis, alcoholic hepatitis, and/or alcoholic cirrhosis were included. Interventions encompassed anabolic-androgenic steroids at any dose or duration versus placebo or no intervention. The trials could be double blind, single blind, or unblinded. The trials could be unpublished or published and no language limitations were applied.

DATA COLLECTION AND ANALYSIS

All analyses were performed according to the intention-to-treat method. The statistical package (RevMan and MetaView) provided by the Cochrane Collaboration was used. The methodological quality of the randomised clinical trials was evaluated by components of methodological quality.

MAIN RESULTS

Combining the results of five randomised clinical trials randomising 499 patients with alcoholic hepatitis and/or cirrhosis demonstrated no significant effects of anabolic-androgenic steroids on mortality (relative risk (RR) 0.96, 95% confidence interval (CI) 0.72 to 1.28), liver related mortality (RR 0.83, 95% CI 0.60 to 1.15), complications of liver disease (RR 1.25, 95% CI 0.74 to 2.10), and liver histology. Further, anabolic-androgenic steroids did not significantly affect a number of other outcome measures. Anabolic-androgenic steroids were not associated with a significantly increased risk of non-serious adverse events but with the seldom occurrence of serious adverse events (RR 4.54, 95% CI 0.57 to 36.30).

REVIEWER'S CONCLUSIONS

This systematic review could not demonstrate any significant beneficial effects of anabolic-androgenic steroids on any clinically important outcomes (mortality, liver related mortality, liver complications, and histology) of patients with alcoholic liver disease.

Treatment effects of eptifibatide in planned coronary stent implantation in patients with chronic kidney disease (ESPRIT Trial)

The role of platelet glycoprotein IIb/IIIa inhibitor therapy in patients with mild renal impairment is not well characterized. Our objective was to explore the associations of creatinine clearance (CrCl) with outcomes in a trial of eptifibatide therapy in patients who underwent percutaneous coronary intervention (PCI). We analyzed 48-hour and 30-day outcomes of patients enrolled in the Enhanced Suppression of the Platelet IIb/IIIa Receptor with Integrilin Therapy (ESPRIT) trial. Patients were randomly assigned to placebo or eptifibatide as an adjunct to stent implantation (1,755 with CrCl > or =60 ml/min and 289 with CrCl <60 ml/min). CrCl was calculated using the Cockcroft and Gault formula, and the associations of CrCl with outcomes were evaluated using logistic regression models. Patients with CrCl <60 ml/min were more likely to be older, women, hypertensive, and have a history of coronary artery bypass surgery, stroke, or peripheral vascular disease. The interaction of eptifibatide with CrCl had borderline significance for the 30-day outcome (p = 0.109). Treatment effect trended toward a greater magnitude in patients with lower CrCl (60 ml/min) (odds ratio 0.53, confidence interval 0.34 to 0.83) compared with those with higher CrCl (90 ml/min) (odds ratio 0.68, confidence interval 0.49 to 0.94). An accompanying increase in bleeding risk also was not apparent with lower CrCl. The treatment effect of eptifibatide is realized regardless of renal function and trends toward being greater in patients with mild renal impairment.

Drotrecogin alfa (activated) administration across clinically important subgroups of patients with severe sepsis

OBJECTIVE

To assess the effects of drotrecogin alfa (activated) therapy, a recombinant human activated protein C, across clinically relevant subpopulations in a randomized, phase 3, placebo-controlled study of patients with severe sepsis (recombinant human activated protein C worldwide evaluation in severe sepsis [PROWESS]).

DESIGN

Univariate and multivariable analysis of prospectively defined subgroups from the PROWESS study.

SETTING

A total of 164 medical centers in 11 countries.

PATIENTS

A total of 1,690 patients with severe sepsis.

MEASUREMENTS AND MAIN RESULTS

We report observed 28-day mortality rates for drotrecogin alfa (activated) and placebo patients for subgroups prospectively defined by demographic data, surgical status, type and site of infection, and clinical and biochemical measures of disease severity. We performed subgroup analyses to explore the consistency of the mortality benefit observed in the overall population and performed tests for both quantitative and qualitative interactions. To examine the magnitude of the treatment benefit with drotrecogin alfa (activated) across the underlying predicted risk of mortality spectrum, we used stepwise logistic regression on PROWESS placebo patients to generate a predicted risk of mortality model that simultaneously included many clinical and biochemical markers of mortality risk. Because drotrecogin alfa (activated) has anticoagulant properties, we also present analyses of bleeding and thrombotic events. Actual mortality rates were lower with drotrecogin alfa (activated) compared with placebo for nearly all prospectively defined subgroups. Both univariate and multivariable regression analyses showed a consistent relative risk reduction in 28-day mortality rates for drotrecogin alfa (activated). Larger absolute risk reductions were found with drotrecogin alfa (activated) in patients with a higher baseline predicted risk of mortality, and actual mortality rates were lower with drotrecogin alfa (activated) in all subgroups defined by disease severity measures where a > or = 20% placebo mortality was observed. Although discriminatory power was limited by few observed events, the increased absolute risk of experiencing a serious bleeding event with treatment did not seem to vary according to the baseline predicted risk of mortality.

CONCLUSIONS

The administration of drotrecogin alfa (activated) to patients with severe sepsis was associated with a significant survival benefit that tended to increase with higher baseline likelihood of death. Current data suggest that the increased risk of bleeding does not vary according to likelihood of death.

Sex-based differences in the effect of digoxin for the treatment of heart failure

BACKGROUND

The Digitalis Investigation Group trial reported that treatment with digoxin did not decrease overall mortality among patients with heart failure and depressed left ventricular systolic function, although it did reduce hospitalizations slightly. Even though the epidemiologic features, causes, and prognosis of heart failure vary between men and women, sex-based differences in the effect of digoxin were not evaluated.

METHODS

We conducted a post hoc subgroup analysis to assess whether there were sex-based differences in the effect of digoxin therapy among the 6800 patients in the Digitalis Investigation Group study. The presence of an interaction between sex and digoxin therapy with respect to the primary end point of death from any cause was evaluated with the use of Mantel-Haenszel tests of heterogeneity and a multivariable Cox proportional-hazards model, adjusted for demographic and clinical variables.

RESULTS

There was an absolute difference of 5.8 percent (95 percent confidence interval, 0.5 to 11.1) between men and women in the effect of digoxin on the rate of death from any cause (P=0.034 for the interaction). Specifically, women who were randomly assigned to digoxin had a higher rate of death than women who were randomly assigned to placebo (33.1 percent vs. 28.9 percent; absolute difference, 4.2 percent, 95 percent confidence interval, -0.5 to 8.8). In contrast, the rate of death was similar among men randomly assigned to digoxin and men randomly assigned to placebo (35.2 percent vs. 36.9 percent; absolute difference, -1.6 percent; 95 percent confidence interval, -4.2 to 1.0). In the multivariable analysis, digoxin was associated with a significantly higher risk of death among women (adjusted hazard ratio for the comparison with placebo, 1.23; 95 percent confidence interval, 1.02 to 1.47), but it had no significant effect among men (adjusted hazard ratio, 0.93; 95 percent confidence interval, 0.85 to 1.02; P=0.014 for the interaction).

CONCLUSIONS

The effect of digoxin therapy differs between men and women. Digoxin therapy is associated with an increased risk of death from any cause among women, but not men, with heart failure and depressed left ventricular systolic function.

The role of covariates in estimating treatment effects and risk in long-term clinical trials

This paper reviews previously published work showing that the impact of including covariates in models used to estimate the magnitude of treatment effects in long-term clinical trials is different from what would be predicted from results for the normal linear model. Typically, models with and without covariates cannot simultaneously be valid. A case is made for the use of data from clinical trials to model the future risk and potential benefits of treatment in individual subjects. The methods and results are illustrated using data from the West of Scotland Coronary Prevention Study.

Subgroup analysis, covariate adjustment and baseline comparisons in clinical trial reporting: current practice and problems

Clinical trial investigators often record a great deal of baseline data on each patient at randomization. When reporting the trial's findings such baseline data can be used for (i) subgroup analyses which explore whether there is evidence that the treatment difference depends on certain patient characteristics, (ii) covariate-adjusted analyses which aim to refine the analysis of the overall treatment difference by taking account of the fact that some baseline characteristics are related to outcome and may be unbalanced between treatment groups, and (iii) baseline comparisons which compare the baseline characteristics of patients in each treatment group for any possible (unlucky) differences. This paper examines how these issues are currently tackled in the medical journals, based on a recent survey of 50 trial reports in four major journals. The statistical ramifications are explored, major problems are highlighted and recommendations for future practice are proposed. Key issues include: the overuse and overinterpretation of subgroup analyses; the underuse of appropriate statistical tests for interaction; inconsistencies in the use of covariate-adjustment; the lack of clear guidelines on covariate selection; the overuse of baseline comparisons in some studies; the misuses of significance tests for baseline comparability, and the need for trials to have a predefined statistical analysis plan for all these uses of baseline data.

Does a homeopathic ultramolecular dilution of Thyroidinum 30cH affect the rate of body weight reduction in fasting patients? A randomised placebo-controlled double-blind clinical trial

OBJECTIVE

To test whether an ultramolecular dilution of homeopathic Thyroidinum has an effect over placebo on weight reduction of fasting patients in so-called 'fasting crisis'.

DESIGN

Randomised, placebo-controlled, double-blind, parallel group, monocentre study.

SETTING/LOCATION

Hospital for internal and complementary medicine in Munich, Germany.

SUBJECTS

Two hundred and eight fasting patients encountering a stagnation or increase of weight after a weight reduction of at least 100 g/day in the preceding 3 days.

INTERVENTION

One oral dose of Thyroidinum 30cH (preparation of thyroid gland) or placebo.

OUTCOME MEASURES

Main outcome measure was reduction of body weight 2 days after treatment. Secondary outcome measures were weight reduction on days 1 and 3, 15 complaints on days 1-3, and 34 laboratory findings on days 1-2 after treatment.

RESULTS

Weight reduction on the second day after medication in the Thyroidinum group was less than in the placebo group (mean difference 92 g, 95% confidence interval 7-176 g, P=0.034). Adjustment for baseline differences in body weight and rate of weight reduction before medication, however, weakened the result to a non-significant level (P=0.094). There were no differences between groups in the secondary outcome measures.

CONCLUSIONS

Patients receiving Thyroidinum had less weight reduction on day 2 after treatment than those receiving placebo. Yet, since no significant differences were found in other outcomes and since adjustment for baseline differences rendered the difference for the main outcome measure non-significant, this result must be interpreted with caution. Post hoc evaluation of the data, however, suggests that by predefining the primary outcome measure in a different way, an augmented reduction of weight on day 1 after treatment with Thyroidinum may be demonstrated. Both results would be compatible with homeopathic doctrine (primary and secondary effect) as well as with findings from animal research.

Selection and description of cancer clinical trials participants--science or happenstance?

BACKGROUND

The selection of clinical trials participants is a critical step in study design, because it affects the generalizability of recommendations made on the basis of trial results and public acceptance of medical research. The authors assessed the heterogeneity of subgroups in cancer treatment and prevention trials and the analysis of subgroups in the evaluation of trial outcomes.

METHODS

The authors reviewed published reports (1990-2000) of cancer prevention and treatment trials from 11 journals. They report here on all Phase III cancer treatment and prevention trials that had at least 100 participants and were conducted among adults in the United States. A structured abstract was developed and used to extract data from the 261 published reports. Descriptive summaries of the abstracted data provided the information included in this systematic review.

RESULTS

Age and gender of study participants were reported in more than 90% of these trials, whereas fewer than 30% of the trials reported race or ethnicity. Gender was reported as an explicit criterion for participant selection primarily in studies of gender specific malignancies. Race and ethnicity were reported as explicit selection criteria for participant selection for five of the prevention trials and for none of the cancer treatment trials. The 105 treatment trials that reported including both men and women had 42,355 participants, and 38.6% of those participants were women. The 26 prevention trials that reported including both men and women had 73,995 participants, and 34.7% of those participants were women. Fifty-seven treatment trials reported participant ethnic diversity: There were 45,815 participants, with 10.5% African-American participants and with less than 1% Hispanic, Asian, or Native American participants. Seventeen prevention trials reported participant ethnic diversity: There were 91,741 participants, with 5.5% African-American participants, 1.7% Hispanic participants, and less than 1% Asian or Native American participants.

CONCLUSIONS

Cancer treatment and prevention trial reports provide scant information about participant race and ethnicity. Such studies use participant selection criteria that do not define diverse subgroups, and few subgroup analyses are conducted. Improvements in the selection, reporting, and analysis of clinical trials participants are needed.

Baseline comparisons in surgical trials

INTRODUCTION

Baseline measurements are an essential component of clinical trials. They should establish that the groups involved are clinically equivalent so that any observed differences can be attributed to the intervention under evaluation. The objective of this study was to review the use of baseline comparisons in surgical trials.

METHODS

A standardized format was used to evaluate 206 surgical trials published within 10 prestigious journals between January 1997 and December 1999.

RESULTS

One hundred and fifty-one (73%) of the trials used the first table in the article to record baseline data. More than one-quarter of the trials declared less than five items and one-third of the trials inappropriately used 'P-values' as a measure of baseline equivalence. Only nine of the 54 multicentre studies (17%) mentioned the individual centres as possible confounding factors.

DISCUSSION

Greater attention needs to be paid to baseline comparisons in surgical trials. One way of developing this type of critical ability is to use a checklist while reading a surgical trial of interest. Good resources include the CONSORT Statement by a group of prominent international biostaticians and the CLEAR (Critical Literature Evaluation and Research) Courses run by the Royal Australasian College of Surgeons.

Issues related to subgroup analysis in clinical trials

Interpretation of subgroup findings is a difficult task. The attempt of this article is to clarify confusions on subgroup analysis and to give some practical suggestions on how to avoid mistakes in interpreting subgroup outcome. We believe that the correct interpretation of subgroup findings is closely related to the intrinsic statistical property and validity of the subgroup analysis. A systematic discussion on subgroup analysis from a statistical point of view will be helpful to clinical trial practitioners.

Nursing intervention studies: a descriptive analysis of issues important to clinicians

When reading a report of an intervention study, clinicians are interested in knowing: whether the intervention is effective, with whom it is effective, how much benefit it produces, and whether associated, adverse outcomes occur. Recommendations have been made in the research literature regarding how to conduct and report intervention studies so as to produce knowledge regarding these questions. This descriptive study was conducted to estimate the frequency with which these recommendations are being used in nursing intervention studies. Data pertinent to five research questions were extracted from 84 experimental and quasi-experimental study reports published between 1998 and 2000. Seventeen percent of the studies used a design that could statistically test for variation in intervention effect depending on the level of an individual characteristic. However, a test of interaction was actually conducted in only 8% of the studies. The magnitude of the intervention's effect was addressed in 38% of the study reports. Providing the proportion of persons in the intervention group who attained a discrete outcome was the most frequently used way of showing intervention magnitude. Associated, adverse outcomes were examined in 23% of the studies, and were most often measured as continuous variables. The low level of use of recommended methods leads the author to suggest dialogue between clinicians and researchers to determine if intervention studies are being conducted and reported in ways that produce knowledge that is useful to clinicians.

Anabolic-androgenic steroids for alcoholic liver disease: a Cochrane review

OBJECTIVES

The objectives were to assess the beneficial and harmful effects of anabolic-androgenic steroids for alcoholic liver disease.

METHODS

The Cochrane Hepato-Biliary Group Controlled Trials Register, The Cochrane Library, MEDLINE, EMBASE, and full text searches were combined. Only randomized clinical trials studying patients with alcoholic liver disease were included. Interventions encompassed anabolic-androgenic steroids at any dose or duration versus placebos or no intervention. The statistical package (RevMan and MetaView) provided by The Cochrane Collaboration was used.

RESULTS

Five randomized clinical trials (including mainly men with alcoholic hepatitis and/or cirrhosis) were identified. Only one trial was assessed as adequate regarding all methodological quality components. Anabolic-androgenic steroids versus placebos or no intervention demonstrated no significant effects on mortality (relative risk [RR] = 0.96, 95% CI = 0.72-1.28), liver-related mortality (RR = 0.83, 95% CI = 0.60-1.15), complications to the liver disease (RR = 1.25, 95% CI = 0.74-2.10), liver histology, and a number of other outcome measures. Anabolic-androgenic steroids were not associated with a significantly increased risk of nonserious adverse events, but with the seldom occurrence of serious adverse events (RR = 4.54,95% CI = 0.57-36.30).

CONCLUSIONS

This systematic review could not demonstrate any significant beneficial effects of anabolic-androgenic steroids on any clinically important outcomes of patients with alcoholic liver disease.

How should meta-regression analyses be undertaken and interpreted?

Appropriate methods for meta-regression applied to a set of clinical trials, and the limitations and pitfalls in interpretation, are insufficiently recognized. Here we summarize recent research focusing on these issues, and consider three published examples of meta-regression in the light of this work. One principal methodological issue is that meta-regression should be weighted to take account of both within-trial variances of treatment effects and the residual between-trial heterogeneity (that is, heterogeneity not explained by the covariates in the regression). This corresponds to random effects meta-regression. The associations derived from meta-regressions are observational, and have a weaker interpretation than the causal relationships derived from randomized comparisons. This applies particularly when averages of patient characteristics in each trial are used as covariates in the regression. Data dredging is the main pitfall in reaching reliable conclusions from meta-regression. It can only be avoided by prespecification of covariates that will be investigated as potential sources of heterogeneity. However, in practice this is not always easy to achieve. The examples considered in this paper show the tension between the scientific rationale for using meta-regression and the difficult interpretative problems to which such analyses are prone.

Thrombolytic treatment of acute ischemic stroke

Intravenous recombinant tissue-type plasminogen activator is approved by the US Food and Drug Administration for treating acute ischemic stroke within 3 hours of onset of symptoms. Initiation of thrombolysis within 90 minutes of onset of symptoms is a treatment goal supported by current studies. Postmarketing data suggest that the risk of intracranial hemorrhage may be unacceptably high when recombinant tissue-type plasminogen activator is given to patients who would not have been eligible for enrollment in the pivotal phase 3 clinical trials. Further studies of local intra-arterial thrombolysis and improved selection of patients with advanced brain imaging are expected in the future, but the emphasis at present should be on rapid identification, evaluation, and treatment of appropriate patients with intravenous therapy.

Study control, violators, inclusion criteria and defining explanatory and pragmatic trials

Important differences between explanatory and pragmatic studies were originally argued by Schwartz and Lellouch. Three important differences between the two types of study involve study control, study violators and inclusion criteria. It was originally argued that explanatory studies are highly controlled, and pragmatic studies may be looser and more like 'real life'. It was argued that an explanatory study should only analyse those receiving treatment, and a pragmatic study would analyse all randomized patients. Explanatory trials are said to use homogeneous groups, and pragmatic studies have less selection (better generalizability). Some suggestions are put forward to update the original distinctions between these two attitudes for future study design. Poor study control is undesirable (but might be necessary) and should not be welcomed as pragmatic. The intention-to-treat strategy is now considered as standard for nearly all trials. Homogeneity is a red herring for studies in humans. Inclusion criteria should be minimized and they should not be used to justify claims of representativeness. Routine criticism of randomized controlled trials for being unrepresentative is unwarranted. We should accept that most trials in humans are 'explanatory'. The division line should be moved, so that pragmatic studies are in the domain of non-therapeutics and complex treatments.

On the variability of covariate adjustment. experience with Koch's method for evaluating the absolute difference in proportions in randomized clinical trials

Recently Koch et al. suggested an interesting approach for covariate adjustment in randomized clinical trials. Here we report our experience with the approach in Assent II, a large randomized trial comparing 30-day mortality rates of two thrombolytic treatments. Data from the Gusto-I study and a recent successor to the Assent II study will also be used. Further, we balance the advantage of covariate adjustment (with Koch's method) to the extra efforts and resources necessary. In this context we highlight the fact that there is a potential risk of using a covariate-adjusted analysis as primary analysis and show that covariate adjustment has the least effect when it is needed most. Analytic results will be given to quantify this risk and a simulation study illustrates our findings.

Ursodeoxycholic acid for primary biliary cirrhosis

BACKGROUND

Primary biliary cirrhosis is a rare autoimmune liver disease and an effective treatment has been difficult to establish. Some randomised clinical trials have found an effect of ursodeoxycholic acid for primary biliary cirrhosis.

OBJECTIVES

Evaluate the beneficial effects and adverse effects of peroral ursodeoxycholic acid for primary biliary cirrhosis versus placebo or no intervention.

SEARCH STRATEGY

The Controlled Trials Register of The Cochrane Hepato-Biliary Group, The Cochrane Library, MEDLINE, EMBASE and the full text of the identified studies were searched until April 2001. The electronic searches were done by entering the search terms 'ursodeoxycholic acid', 'UDCA', 'primary biliary cirrhosis', and 'PBC'.

SELECTION CRITERIA

Randomised clinical trials evaluating ursodeoxycholic acid administered perorally at any dose versus placebo or no intervention in patients with primary biliary cirrhosis diagnosed by any method. Only trials using an adequate method for randomisation were included, regardless of blinding and language.

DATA COLLECTION AND ANALYSIS

The methodologic quality of the randomised clinical trials was evaluated by components and the Jadad-score. The following outcomes were extracted: mortality, liver transplantation, pruritus, other clinical symptoms (jaundice, portal pressure, (bleeding) oesophageal varices, ascites, hepatic encephalopathy, hepato-renal syndrome, autoimmune conditions), liver biochemistry, liver function, liver biopsy findings, quality of life, and adverse events. All analyses were performed according to the intention-to-treat method.

MAIN RESULTS

A total of 16 randomised clinical trials evaluating ursodeoxycholic acid against placebo (n = 15) or no intervention (n = 1) in 1422 patients were identified. The median Jadad-score was 3 (range 1-5). A number of trials described as double blind had problems with the blinding. Neither mortality (odds ratio = 0.94; 95% confidence interval (CI) 0.60 to 1.48), liver transplantation (odds ratio = 0.83; 95% CI 0.52 to 1.32), mortality or liver transplantation (odds ratio = 0.90; 95% CI 0.65 to 1.26), pruritus, fatigue, autoimmune conditions, quality of life, liver histology, or portal pressure were significantly affected by ursodeoxycholic acid (given in doses of 8-15 mg/kg/day for three months to five years). However, ursodeoxycholic acid significantly (P < 0.05) reduced ascites, jaundice, and biochemical variables such as serum bilirubin and liver enzymes. Ursodeoxycholic acid was not significantly associated with adverse events. Including data after patients had been switched onto open label ursodeoxycholic acid confirmed the findings regarding the lack of a significant effect of ursodeoxycholic acid on mortality and mortality or liver transplantation. A significant (P = 0.04) effect was, however, observed on the incidence of liver transplantation (odds ratio = 0.68; 95% CI 0.48 to 0.98).

REVIEWER'S CONCLUSIONS

Ursodeoxycholic acid has a marginal therapeutic effect for primary biliary cirrhosis. On the positive side, ursodeoxycholic acid has few side effects. The general usage of ursodeoxycholic acid for primary biliary cirrhosis needs reevaluation.

Trials within trials: confirmatory subgroup analyses in controlled clinical experiments

Subgroup analyses remain a popular and necessary component of controlled clinical trials. However, lack of prospective specification, inadequate sample size, inability to maintain power, and the cumulative effect of sampling error can complicate their interpretation. This article demonstrates that clinical trial design tools that would allow the medical community to draw confirmatory and not just exploratory conclusions from specific subgroup evaluations are available to methodologists. Distinct from the use of a treatment by subgroup interaction term, this methodology provides an evaluation of the effect of an intervention within a particular subgroup stratum prospectively declared to be of interest to the investigators. The necessary prespecification of stratum-specific type I error rates, when combined with (1) a stratum-specific event rate in the subgroup, (2) a stratum-specific primary endpoint, (3) a stratum-specific endpoint precision, and/or (4) a stratum-specific efficacy, satisfies the requirements for a subgroup stratum's "stand-alone" interpretation at the trial's conclusion.

Is the role of steroids in acute spinal cord injury now resolved?

Steroids have long been used in the context of acute spinal cord injury but the evidence for doing so is limited. The second National Acute Spinal Cord Injury Study trial had the potential to provide such evidence for the first time, as this was a placebo controlled, prospective, randomized trial. From the outset, however, some clinicians found the methodology and consequently the results unsatisfactory. This concern has been revisited within the evidence-based framework of critical appraisal of the accumulation of clinical studies. High-dose methylprednisolone cannot be justified as a standard treatment in acute spinal cord injury within current medical practice.

Multilevel models for meta-analysis, and their application to absolute risk differences

Meta-analysis can be considered a multilevel statistical problem, since information within studies is combined in the presence of potential heterogeneity between studies. Here a general multilevel model framework is developed for meta-analysis to combine either summary data or individual patient outcome data from each study, and to include either study or individual level covariates that might explain heterogeneity. Classical and Bayesian approaches to estimation are contrasted. These methods are applied to a meta-analysis of trials of thrombolytic therapy after myocardial infarction. Subgroups within the trials were available, categorized by the time delay until treatment, so that a three-level random effects model that includes time delay as a covariate is proposed. In addition it was desired to represent the treatment effect as an absolute risk reduction, rather than the conventional odds ratio. We show how this can be achieved within a Bayesian analysis, while still recognizing the binary nature of the original outcome data.

HIV protease inhibitor substitution in patients with lipodystrophy: a randomized, controlled, open-label, multicentre study

BACKGROUND

Lipodystrophy, dyslipidaemia and insulin resistance often complicate protease inhibitor-containing antiretroviral therapy. The aims of this study were to determine if these are reversible with continued HIV suppression following protease inhibitor substitution.

METHODS

Eighty-one HIV protease inhibitor recipients (78 male; mean antiretroviral therapy, 55 months) with predominant peripheral lipoatrophy, HIV RNA < 400 copies/ml plasma for at least the preceding 6 months, and no prior abacavir, non-nucleoside analogue or adefovir therapy were randomized 3 : 2 to continue nucleoside analogues and substitute protease inhibitor(s) with abacavir, nevirapine, adefovir and hydroxyurea (n = 49) or to continue all therapy (n = 32) with an option to switch at week 24. The primary endpoints were total body fat and HIV RNA at week 24. Other assessments were regimen safety, regional body composition, metabolic parameters, quality of life, and CD4 T-lymphocyte counts to week 48.

RESULTS

There was a greater decline in total body fat in the switch group than in the continue group (-1.6 and -0.4 kg, respectively at week 24; P = 0.006). This comprised greater declines in limb and subcutaneous abdominal fat, and in intra-abdominal fat of patients with moderate or severe abdominal fat accumulation. Viral suppression was similar, despite 18 (37%) switch group patients ceasing at least one study drug by week 24 because of adverse events. Total cholesterol and triglycerides declined more in the switch group (both P < 0.002). High density lipoprotein cholesterol increased significantly in both groups at week 48 (P < 0.02). There was no change for any glycaemic parameter.

CONCLUSIONS

In predominantly lipoatrophic patients, switching from HIV protease inhibitor therapy lead to improved lipids and less intra-abdominal fat, but also to less peripheral fat, and had minimal effect on insulin resistance. Virological control in these heavily pretreated patients was unaffected, despite frequent switch drug cessations.

New strategies for clinical trials in patients with sepsis and septic shock

OBJECTIVE

The difficulty in identifying new treatment modalities that significantly reduce the mortality and morbidity rates associated with sepsis has highlighted the need to reevaluate the approach to clinical trial design. The United Kingdom Medical Research Council convened an International Working Party to address these issues.

DATA SOURCES

The subject areas that were to be the focus of discussion were identified by the co-chairs, and group leaders were nominated. Preconference reading material was circulated to group members.

STUDY SELECTION AND DATA EXTRACTION

Small-group discussion fed into an iterative process of feedback from plenary sessions, followed by the formulation of recommendations. Finally, each working group prepared a summary of its recommendations and these are reported herein.

DATA SYNTHESIS

There were five key recommendations. First, investigators should no longer rely solely on the American College of Chest Physicians/Society of Critical Care Medicine definitions of sepsis or sepsis syndrome as the basis of trial entry. Entry criteria should be based on three principles: a) All patients should have infection; b) there should be evidence of a pathologic process that represents a biologically plausible target for the proposed intervention, for example, an abnormal circulating level of a biological marker pertinent to the study drug; and c) patients should fall into an appropriate category of severity (usually severe sepsis). Second, investigators should use a scoring system for organ dysfunctions that has been validated and that can be incorporated into all sepsis studies; agreement on the use of a single system would simplify comparisons between studies. Third, the primary outcome measure generally should be mortality rates, but under appropriate circumstances major morbidities could be considered as primary end points. Regardless of choice of the duration to primary end point, patients should be followed for > or =90 days. Fourth, sample size needs to be based on a realistic assessment of achievable effect size based on knowledge of the at-risk population. Fifth, subgroups should be few in number and should be defined a priori on the basis of variables present before randomization.

CONCLUSIONS

Important changes in several aspects of trial design may improve the quality of clinical studies in sepsis and maximize the chance of identifying effective therapeutic agents.

Viable Malignant Cells After Primary Chemotherapy for Disseminated Nonseminomatous Germ Cell Tumors: Prognostic Factors and Role of Postsurgery Chemotherapy—Results From an International Study Group

PURPOSE: To assess the value of postsurgery chemotherapy in patients with disseminated nonseminomatous germ-cell tumors (NSGCTs) and viable residual disease after first-line cisplatin-based chemotherapy.

PATIENTS AND METHODS: The outcome of 238 patients was reviewed. Tumor markers had normalized in all patients before resection. A multivariate analysis of survival was performed on 146 patients.

RESULTS: The 5-year progression-free survival (PFS) rate was 64% and the 5-year overall survival (OS) rate was 73%. Three factors were independently associated with both PFS and OS: complete resection (P < .001), < 10% of viable malignant cells (P = .001), and a good International Germ Cell Consensus Classification (IGCCC) group (P = .01). Patients were assigned to one of three risk groups: those with no risk factors (favorable group), those with one risk factor (intermediate group), and those with two or three risk factors (poor-risk group). The 5-year OS rate was 100%, 83%, and 51%, respectively (P < .001). The 5-year PFS rate was 69% (95% confidence interval [CI], 62% to 76%) and 52% (95% CI, 40% to 64%) in postoperative chemotherapy recipients and nonrecipients, respectively (P < .001). No significant difference was detected in 5-year OS rates. After adjustment on the three prognostic factors, postoperative chemotherapy was associated with a significantly better PFS (P < .001) but not with better OS. Patients in the favorable risk group had a 100% 5-year OS, with or without postoperative chemotherapy. Postoperative chemotherapy appeared beneficial in both PFS (P < .001) and OS (P = .02) in the intermediate-risk group but was not statistically beneficial in the poor-risk group.

CONCLUSION: A complete resection may be more critical than recourse to postoperative chemotherapy in the setting of postchemotherapy viable malignant NSGCT. Immediate postoperative chemotherapy or surveillance alone with chemotherapy at relapse may be reasonable options depending on the completeness of resection, IGCCC group, and percent of viable cells. Validation is necessary.

Significance of the clinical course and early upper gastrointestinal studies in predicting complications associated with repair of esophageal atresia

BACKGROUND/PURPOSE

In many centers, use of upper gastrointestinal (UGI) contrast studies in the early postoperative period after esophageal atresia (EA) repair is considered routine. Indications for this are many, including searching for existing problems and predicting future complications. However, most major complications, both early and late, usually are identified clinically before any radiologic studies. The purpose of this study was to investigate factors that may anticipate the development of postoperative complications after EA repair, looking particularly at the predictive value of routine early postoperative UGI studies.

METHODS

A total of 111 consecutive cases of EA were identified retrospectively over a 10-year period from 2 major Canadian pediatric health centers. One hundred one were associated with a distal tracheoesophageal fistula (TEF), of which, 90 had repairs. Ninety-seven percent of these had a UGI study at a median of 9.1 postoperative days (range, 2 to 23) before consideration of oral feeding. Charts were reviewed looking at patient variables, surgical factors, early UGI findings, and postoperative courses. Complications that required intervention were noted, including anastomotic leaks, gastroesophageal reflux (GER), strictures, and recurrent and missed fistulae. All initial UGI studies were reexamined by 1 of 2 pediatric radiologists. Logistic regression was used to examine relationships between these clinical and radiologic variables and outcomes.

RESULTS

Of the variables analyzed, univariate analysis showed clinically significant leaks to be associated with intraoperative factors (subjective degree of anastomotic tension, and the use of myotomies) and early postoperative clinical evidence suggesting a leak. In a multivariate model, all remained independently significant except for the use of myotomies. Later development of clinically significant GER also was associated with the degree of tension. It had no relationship, however, with findings of dysmotility, esophageal shortening, or reflux at the initial UGI study. Development of a stricture requiring dilatations or resection was associated with a history of clinically evident GER only; no relationships were seen with a history of an anastomotic leak or any other clinical, operative, or radiographic variables. Missed or recurrent fistulae were all suspected clinically before radiologic confirmation.

CONCLUSIONS

Early and late complications after repair of EA can be identified and potentially anticipated based on clinical findings at the time of repair and during the postoperative period. The use of early "routine" UGI studies, with no suspicion of a problem, has little value in terms of predicting complications or future clinical course.

Clinical trials in psychiatry: do results apply to practice?

OBJECTIVE

The purpose of this communication is to alert psychiatrists to the difficulties of translating results of group difference obtained from large, randomized clinical trials to the treatment of individual patients.

METHOD

Reported discrepancies between a) clinical trial participants and general psychiatric patients, b) clinical trial investigators and general clinicians, and c) study trial and usual clinic conditions were assessed.

RESULTS

The results confirm that important differences exist in all 3 areas.

CONCLUSIONS

Recommendations for researchers include more complete assessments of factors that account for individual difference, an appraisal of outcomes more important to patients than symptom scores, and the use of statistical methods that permit the evaluation of individual difference. Recommendations for clinicians include a careful differentiation of results obtained in different phases of clinical trials and a clear appreciation of the different purposes of those trials. Clinicians should also appreciate that short-term effectiveness is not the same as long-term outcome and that aggregate scores may not apply to individual patients.