Amy Salter

Performance of the Modified Poisson Regression Approach for Estimating Relative Risks From Clustered Prospective Data

Modified Poisson regression, which combines a log Poisson regression model with robust variance estimation, is a useful alternative to log binomial regression for estimating relative risks. Previous studies have shown both analytically and by simulation that modified Poisson regression is appropriate for independent prospective data. This method is often applied to clustered prospective data, despite a lack of evidence to support its use in this setting. The purpose of this article is to evaluate the performance of the modified Poisson regression approach for estimating relative risks from clustered prospective data, by using generalized estimating equations to account for clustering. A simulation study is conducted to compare log binomial regression and modified Poisson regression for analyzing clustered data from intervention and observational studies. Both methods generally perform well in terms of bias, type I error, and coverage. Unlike log binomial regression, modified Poisson regression is not prone to convergence problems. The methods are contrasted by using example data sets from 2 large studies. The results presented in this article support the use of modified Poisson regression as an alternative to log binomial regression for analyzing clustered prospective data when clustering is taken into account by using generalized estimating equations.

Prevalence, awareness, treatment, and control of hypertension in rural China: results from Shandong Province.

Objectives Hypertension is an important public health problem in rural China with a rapidly increasing prevalence noted in recent years. This study estimates the prevalence, awareness, treatment and control of hypertension in a rural population in Shandong Province, China. Methods A cross-sectional study was conducted in rural Shandong Province, China, in April 2007 using multistage cluster sampling. A total of 16 364 rural residents aged 25 years and more were interviewed and examined. Two blood pressure (BP) measurements were obtained using a standardized mercury sphygmomanometer after a 5-min seated rest. Information on history of hypertension was obtained using a standard questionnaire. Hypertension was defined as mean systolic BP (SBP) at least 140 mmHg and/or diastolic BP (DBP) at least 90 mmHg, and/or self-reported current use of antihypertensive medication. Results Overall, 43.8% of the population had hypertension. Among hypertensive patients, only 26.2% were aware of their hypertension, 22.2% were currently undergoing antihypertensive treatment, and 3.9% achieved BP control (<140/90 mmHg). Lack of knowledge about hypertension and the importance of BP control were associated with poor compliance with nonpharmacological and pharmacological treatments. Conclusions In the study population, the prevalence of hypertension is high, but levels of awareness, treatment and control are unacceptably low. There is an urgent need for comprehensive integrated strategies to improve prevention, detection and treatment of hypertension in rural areas in Shandong Province, China.

Competing risks survival analysis applied to data from the Australian Orthopaedic Association National Joint Replacement Registry

Background and purpose The Kaplan-Meier (KM) method is often used in the analysis of arthroplasty registry data to estimate the probability of revision after a primary procedure. In the presence of a competing risk such as death, KM is known to overestimate the probability of revision. We investigated the degree to which the risk of revision is overestimated in registry data. Patients and methods We compared KM estimates of risk of revision with the cumulative incidence function (CIF), which takes account of death as a competing risk. We considered revision by (1) prosthesis type in subjects aged 75–84 years with fractured neck of femur (FNOF), (2) cement use in monoblock prostheses for FNOF, and (3) age group in patients undergoing total hip arthroplasty (THA) for osteoarthritis (OA). Results In 5,802 subjects aged 75–84 years with a monoblock prosthesis for FNOF, the estimated risk of revision at 5 years was 6.3% by KM and 4.3% by CIF, a relative difference (RD) of 46%. In 9,821 subjects of all ages receiving an Austin Moore (non-cemented) prosthesis for FNOF, the RD at 5 years was 52% and for 3,116 subjects with a Thompson (cemented) prosthesis, the RD was 79%. In 44,365 subjects with a THA for OA who were less than 70 years old, the RD was just 1.4%; for 47,430 subjects > 70 years of age, the RD was 4.6% at 5 years. Interpretation The Kaplan-Meier method substantially overestimated the risk of revision compared to estimates using competing risk methods when the risk of death was high. The bias increased with time as the incidence of the competing risk of death increased. Registries should adopt methods of analysis appropriate to the nature of their data.

Risk of hospitalization for hip fracture and pneumonia associated with antipsychotic prescribing in the elderly: a self-controlled case-series analysis in an Australian health care claims database.

AbstractBackground: Antipsychotics are commonly used in the elderly to treat the behavioural symptoms of dementia. Randomized controlled trial data on the safety of antipsychotics are limited and little is known about the long-term effects of these medicines. Observational studies have investigated the risk of hip fracture and pneumonia associated with the use of antipsychotics, but varying results may be due to lack of control for unmeasured confounding. Objective: The aim of the study was to investigate the risk of hospitalization for hip fracture and pneumonia in elderly subjects exposed to antipsychotic medication using the self-controlled case-series design to control for unmeasured confounding. Methods: The source of data for this study was the Australian Government Department of Veterans’ Affairs Health Care Claims Database. A self-controlled case-series design was used to measure the excess risk of hospitalization for hip fracture and pneumonia after antipsychotic exposure compared with no-exposure over the 4-year period from 2005 to 2008. Results: There was a significantly increased risk of hip fracture 1 week after exposure to typical antipsychotics, and the risk remained significantly raised with >12 weeks of continuous exposure (incidence rate ratio [IRR] 2.19; 95% CI 1.62, 2.95). The risk of hip fracture was highest in the first week after initiation of atypical antipsychotics (IRR 2.17; 95% CI 1.54, 3.06). The risk then declined with longer-term treatment; however, it remained significantly raised with >12 weeks of continuous exposure (IRR 1.43; 95% CI 1.23, 1.66). The risk of hospitalization for pneumonia was raised in all post-exposure periods for both typical and atypical antipsychotics. Conclusions: Antipsychotic use in the elderly is associated with an increased risk of hospitalization for hip fracture and pneumonia. Given the increased risks of morbidity and mortality associated with these outcomes, practitioners should consider these additional risks when prescribing antipsychotics to treat behavioural symptoms of dementia in the elderly.

Choice of observational study design impacts on measurement of antipsychotic risks in the elderly: a systematic review.

BackgroundAntipsychotics are frequently and increasingly prescribed to treat the behavioural symptoms associated with dementia despite their modest efficacy. Evidence regarding the potential adverse events of antipsychotics is limited and little is known about the longer-term safety of these medicines in the elderly. The aim of this review was to determine the impact of the choice of observational study design and methods used to control for confounding on the measurement of antipsychotic risks in elderly patients.MethodsWe searched PUBMED and the Cochrane controlled trials register for double-blind randomised controlled trials (RCTs), meta-analyses and published observational studies of antipsychotics.ResultsForty four studies were identified for the endpoints; death, cerebrovascular events, hip fracture and pneumonia. RCTs found a 20% to 30% increased risk of death, or an absolute increase of 1extra death per 100 patients with atypical antipsychotics compared to non-use. Cohort and instrumental variable analyses estimated between 2 to 7 extra deaths per 100 patients with conventional compared to atypical antipsychotics. RCTs found a 2 to 3 times increased risk of all cerebrovascular events with atypical antipsychotics compared to placebo and no association with serious stroke that required hospitalisation. Observational studies using cohort and self-controlled case-series designs reported similar results; no association where the endpoint was stroke causing hospitalisation and a doubling of risk when minor stroke was included. No RCTs were available for the outcome of hip fracture or pneumonia. Observational studies reported a 20% to 40% increased risk of hip fracture with both antipsychotic classes compared to non-use. The risk of pneumonia was a 2 to 3 times greater with both classes compared to non-use while a self-controlled case-series study estimated a 60% increased risk. Conventional antipsychotics were associated with a 50% greater hip fracture risk than atypical antipsychotics, while the risk of pneumonia was similar between the classes.ConclusionsChoice of observational study design is critical in studying the adverse effects of antispychotics. Cohort and instrumental variable analyses gave more consistent results to clinical studies for mortality outcomes as have self-controlled case-series for the risk of cerebrovascular events and stroke. Observational evidence has highlighted the potential for antipsychotics to be associated with serious adverse events that were not reported in RCTs including hip fracture and pneumonia. Good quality observational studies are required, that employ appropriate study designs that are robust towards unmeasured confounding, to confirm the potential excess risk of hip fracture and pneumonia with antipsychotics.

Risk of hospitalization for stroke associated with antipsychotic use in the elderly: a self-controlled case series.

AbstractBackground Antipsychotics are commonly used in the elderly despite a lack of safety data from randomized trials, particularly for the typical antipsychotics. Observational studies have investigated the association between antipsychotics and stroke but results vary, which may be due to lack of control for unmeasured confounding. Objective To estimate the risk of hospitalization for stroke in elderly users of antipsychotics. Study Design and Setting Using the Australian Government Department of Veterans’ Affairs administrative claims dataset we utilized a self-controlled case series design to risk-adjust for potential unmeasured confounding. Risk periods prior to antipsychotic initiation were also included to search for evidence of confounding by indication. Unexposed patients were included to adjust for the increasing incidence of hospitalization for stroke with age. Results There were 10 638 patients aged ≥65 years with at least one hospitalization for stroke identified during the 4-year period from 1 January 2003 to 31 December 2006. Of these, 514 patients were initiated on typical antipsychotics and 564 patients were initiated on atypical antipsychotics. Hospitalization for stroke was increased in the first week after initiation of a typical antipsychotic (incidence rate ratio [IRR] 2.3; 95% CI 1.3, 3.8). There was no evidence of an increased risk of hospitalization for stroke after initiation of atypical antipsychotics. The risk of hospitalization for stroke progressively increased in the weeks leading up to first-time antipsychotic treatment. However, while the risk of hospitalization for stroke in the week prior to initiating antipsychotic therapy was significantly increased for patients initiated on typical antipsychotics (IRR 7.2; 95% CI 5.3, 9.8), patients initiated on atypical antipsychotics had no excess risk in the same period (IRR 1.2; 95% CI 0.7, 2.3). Conclusions The results of this study are consistent with randomized controlled trial evidence indicating that there is no increased risk of serious cerebrovascular events requiring hospitalization in patients taking atypical antipsychotics. No randomized controlled trial evidence is available on the risk of hospitalization for stroke with use of typical antipsychotics in the elderly. This study found a small but significantly increased risk of hospitalization for stroke immediately following the initiation of typical antipsychotics. Antipsychotics are likely to be initiated after hospitalization for stroke. This practice is likely to reflect the prescribing of antipsychotics during hospital admission for post-stroke complications such as delirium; however, the long-term effects of this practice are unknown.

Different competing risks models applied to data from the Australian Orthopaedic Association National Joint Replacement Registry

Purpose Here we describe some available statistical models and illustrate their use for analysis of arthroplasty registry data in the presence of the competing risk of death, when the influence of covariates on the revision rate may be different to the influence on the probability (that is, risk) of the occurrence of revision. Patients and methods Records of 12,525 patients aged 75–84 years who had received hemiarthroplasty for fractured neck of femur were obtained from the Australian Orthopaedic Association National Joint Replacement Registry. The covariates whose effects we investigated were: age, sex, type of prosthesis, and type of fixation (cementless or cemented). Extensions of competing risk regression models were implemented, allowing the effects of some covariates to vary with time. Results The revision rate was significantly higher for patients with unipolar than bipolar prostheses (HR = 1.38, 95% CI: 1.01–1.89) or with monoblock than bipolar prostheses (HR = 1.45, 95% CI: 1.08–1.94). It was significantly higher for the younger age group (75–79 years) than for the older one (80–84 years) (HR = 1.28, 95% CI: 1.05–1.56) and higher for males than for females (HR = 1.37, 95% CI: 1.09–1.71). The probability of revision, after correction for the competing risk of death, was only significantly higher for unipolar prostheses than for bipolar prostheses, and higher for the younger age group. The effect of fixation type varied with time; initially, there was a higher probability of revision for cementless prostheses than for cemented prostheses, which disappeared after approximately 1.5 years. Interpretation When accounting for the competing risk of death, the covariates type of prosthesis and sex influenced the rate of revision differently to the probability of revision. We advocate the use of appropriate analysis tools in the presence of competing risks and when covariates have time-dependent effects.

Relative Risk Estimation in Randomized Controlled Trials: A Comparison of Methods for Independent Observations

The relative risk is a clinically important measure of the effect of treatment on binary outcomes in randomized controlled trials (RCTs). An adjusted relative risk can be estimated using log binomial regression; however, convergence problems are common with this model. While alternative methods have been proposed for estimating relative risks, comparisons between methods have been limited, particularly in the context of RCTs. We compare ten different methods for estimating relative risks under a variety of scenarios relevant to RCTs with independent observations. Results of a large simulation study show that some methods may fail to overcome the convergence problems of log binomial regression, while others may substantially overestimate the treatment effect or produce inaccurate confidence intervals. Further, conclusions about the effectiveness of treatment may differ depending on the method used. We give recommendations for choosing a method for estimating relative risks in the context of RCTs with independent observations.

Antipsychotics and the risk of death in the elderly: an instrumental variable analysis using two preference based instruments.

Observational studies have investigated the comparative safety of antipsychotics with varying results. Instrumental variable analysis has been suggested as a possible alternative to conventional analyses when there is concern about the effect of unmeasured confounding in observational studies. Using the example of the risk of death with typical compared to atypical antipsychotics, we aimed to explore the performance of two different instruments. We used the doctor prescribing preference instrument, which has been used in previous studies, to investigate further the assumptions of this instrument in the Australian population. We also propose an alternative instrument, nursing home facility preference.

Comparison of pain models to detect opioid-induced hyperalgesia

Objective Chronic opioid therapy may be associated with hyperalgesia. Our objective was to determine if opioid-induced hyperalgesia detection sensitivity is dependent on the stimulus used to detect it. Methods This open design study compared the detection of hyperalgesia in opioid-dependent subjects (n = 16) and healthy control subjects (n = 16) using the following pain stimuli: cold pain, electrical stimulation, mechanical pressure, and ischemic pain. The opioid-dependent subjects were maintained on either methadone (n = 8) or buprenorphine (n = 8) for at least 3 months. None of the controls was dependent on opioids or other drugs of abuse. Results The opioid-dependent subjects were markedly more sensitive than controls to the cold pain test. Compared with the control group, the hazard ratio for ceasing the test due to intolerable pain was 7.7 (95% confidence interval [CI] 2.6–23.3) in the buprenorphine group and 4.5 (95% CI 1.7–15.6) in the methadone group, with similar data for the cold pain threshold. Of the remaining tests, there were differences only for the electrical pain threshold between treatment groups, with the geometric mean threshold in the buprenorphine group being 1.5 (95% CI 1.1–1.9)-fold higher (ie, less sensitive) than that of the controls; the geometric mean for the methadone group was 1.3 (95% CI 1.04–1.7)-fold higher than that of the controls. There were no significant differences between buprenorphine and methadone patients in test responses. Women were more sensitive to the cold pain (hazard ratio for tolerance, 3.1 [95% CI 1.4–7.3]) and ischemic tests (hazard ratio for tolerance, 2.7 [95% CI 1.2–6.1]). There were significant correlations between cold and ischemic tolerances (r = 0.50; P = 0.003) and between electrical and mechanical pain tolerances (r = 0.52; P = 0.002). Conclusion These findings indicate that cold pain is the most suitable of the methods tested to detect opioid-induced hyperalgesia. This is consistent with its sensitivity to detect opioid analgesia.