The Robust Inference for the Cox Proportional Hazards Model

Lin, Danyu; Wei, L. J.

doi:10.1080/01621459.1989.10478874

Cited by 1,977 publications

(828 citation statements)

References 17 publications

Supporting

Mentioning

814

Contrasting

Unclassified

Order By: Relevance

“…When the proportional hazards assumption was violated, we adjusted the covariance matrix using Lin and Wei's robust sandwich estimates. 9 We adjusted Cox proportional hazards models for age, sex, race/ethnicity, education, income, and the baseline measure of the independent variable. When models were not stratified by treatment arm, we included treatment arm as a covariate.…”

Section: Discussionmentioning

confidence: 99%

Early Response to Preventive Strategies in the Diabetes Prevention Program

Maruthur

Delahanty

et al. 2013

J GEN INTERN MED

View full text Add to dashboard Cite

BACKGROUND:Recommendations for diabetes prevention in patients with prediabetes include lifestyle modification and metformin. However, the significance of early weight loss and glucose measurements when monitoring response to these proven interventions is unknown. OBJECTIVE: To quantify the relationship between early measures of weight and glucose and subsequent diabetes in patients undergoing diabetes prevention interventions. DESIGN: Analysis of results from a randomized controlled trial in 27 academic medical centers in the United States. PARTICIPANTS/INTERVENTIONS: 3,041 adults with hyperglycemia randomized to lifestyle (n=1,018), metformin (n=1,036), or placebo (n=987) with complete follow-up in The Diabetes Prevention Program. MAIN MEASURES: Independent variables were weight loss at 6 and 12 months; fasting glucose (FG) at 6 months; hemoglobin A1c (HbA1c) at 6 months; and post-load glucose at 12 months. The main outcome was time to diabetes diagnosis. KEY RESULTS: After 6 months, 604 participants developed diabetes in the lifestyle (n=140), metformin (n=206), and placebo (n=258) arms over 2.7 years. In the lifestyle arm, 6-month weight loss predicted decreased diabetes risk in a graded fashion: adjusted HR (95 % CI) 0.65 (0.35-1.22), 0.62 (0.33-1.18), 0.46 (0.24-0.87), 0.34 (0.18-0.64), and 0.15 (0.07-0.30) for 0-<3 %, 3-<5 %, 5-<7 %, 7-<10 %, and ≥10 % weight loss, respectively (reference: weight gain). Attainment of optimal 6-month FG and HbA1c and 12-month post-load glucose predicted >60 % lower diabetes risk across arms. We found a significant interaction between 6-month weight loss and FG in the lifestyle arm (P=0.038). CONCLUSION: Weight and glucose at 6 and 12 months strongly predict lower subsequent diabetes risk with a lifestyle intervention; lower FG predicts lower risk even with substantial weight loss. Early reduction in glycemia is a stronger predictor of future diabetes risk than weight loss for metformin. We offer the first evidence to guide clinicians in making interval management decisions for high-risk patients undertaking measures to prevent diabetes.

show abstract

Section: Discussionmentioning

confidence: 99%

Early Response to Preventive Strategies in the Diabetes Prevention Program

Maruthur

Delahanty

et al. 2013

J GEN INTERN MED

View full text Add to dashboard Cite

show abstract

“…We used the Efron (Efron 1977) method to handle tied cases with a robust standard error estimator (Lin and Wei 1989).…”

Section: The Modelmentioning

confidence: 99%

Is Household Wealth Sustainable? An Examination of Asset Poverty Reentry After an Exit

Leonard

2013

J Fam Econ Iss

View full text Add to dashboard Cite

This paper analyzed the influence of financial behaviors on the duration out of asset poverty while controlling for households' life cycle and demographic characteristics. We found evidence for the existence of structural barriers to asset acquisition. Asset accumulation at or above levels equal to nine-months worth of income at the incomepoverty level was important for improving a household's odds of permanently escaping asset poverty, but a linear relationship between asset accumulation and the likelihood of returning to asset poverty did not emerge. Moreover, minimizing debt and diversifying the asset portfolio to include more productive assets were positively related to maintaining assets; but households should also consider the risks associated with portfolio allocations.

show abstract

“…The data given here are based on the stepwise method. The qualitative result was not dependent on the method chosen (stepwise method, forward selection, and backward elimination methods were used) [4,8]. 2-sided statistical significance was set at p < 0.05.…”

Section: Resultsmentioning

confidence: 99%

“…Regression analysis of survival data was performed based on the Cox proportional hazards model and using the conditional logit regression with the SAS PHREG procedure. Stepwise forward selection and backward elimination methods were used [4,5]. 2-sided statistical significance was set at 0.05.…”

Section: Methodsmentioning

confidence: 99%

Characteristics, Treatment and Prognostic Factors of Patients with Gynaecological Malignancies Treated in a Palliative Care Unit at a University Hospital

et al. 2013

View full text Add to dashboard Cite

Background: Limited clinical data have been published on patients suffering from advanced gynaecological malignancies treated in palliative care units, and little is known about prognostic factors. Methods: In a retrospective study, the data of 225 patients with breast, ovarian and cervical cancer treated in the palliative care unit of a university hospital between 1998 and 2009 were assembled. Clinical aspects and baseline symptoms, laboratory parameters, the clinical course, and outcome were evaluated. Results: 225 patients (497 cases; cancer diagnoses: breast 79%, ovarian 13%, and cervix 8%) were included in the analysis. The main symptoms were weakness/fatigue (71%), pain (65%), anorexia/nausea (62%), and dyspnea (46%). Pain control was achieved in 85% of all cases, satisfying control of other symptoms in 80%. The median overall survival (OS) was 59 days. 53% of the patients died at the palliative care unit. In the Cox proportional hazards model, 8 parameters indicated an unfavourable outcome: anorexia/nausea, disordered mental status, elevated lactate dehydrogenase, γ-glutamyltransferase, leukocyte count, hypoalbuminaemia, anaemia and hypercalcaemia. Based on these parameters 3 risk groups were defined: low risk (0-2 factors), intermediate risk (3-5 factors), and high risk (6-8 factors). Median survival for high-risk group was 13 days, for intermediate group 61 days, and for low-risk patients 554 days (p < 0.0001). Conclusion: Weakness/fatigue, pain and anorexia were the main symptoms leading to the hospitalisation of patients with gynaecological malignancies. Symptom and pain control was accomplished in 80% of cases. 8 parameters were identified as indicating a poor outcome, and patients showing at least 6 or more of these factors had a very limited prognosis. Although studied retrospectively, these results may be helpful for individual treatment decisions in patients with advanced gynaecological malignancies. Prospective data and the introduction of documentation systems could help to gain more powerful knowledge about the quality of palliative care.

show abstract

The Robust Inference for the Cox Proportional Hazards Model

Cited by 1,977 publications

References 17 publications

Early Response to Preventive Strategies in the Diabetes Prevention Program

Early Response to Preventive Strategies in the Diabetes Prevention Program

Is Household Wealth Sustainable? An Examination of Asset Poverty Reentry After an Exit

Characteristics, Treatment and Prognostic Factors of Patients with Gynaecological Malignancies Treated in a Palliative Care Unit at a University Hospital

Contact Info

Product

Resources

About