Journal Club for Analysis of Complex Datasets

1
Journal Club for Analysis of Complex Datasets

• Frost FJ, Petersen H, Tollestrup K, Skipper B.
  Influenza and COPD mortality protection as
  pleiotropic, dose-dependent effects of statins.
• Chest 2007 131 1006-1012.

2
Acknowledgement

• Carlos Grijalva, MD, MPH
• Department of Preventive Medicine

3
Background

• Data on antiinflammatory immunomodulatory
  effects of statins suggest they may reduce
  mortality risks from influenza COPD.
• 2 observational studies suggest statins may
  protect community-acquired pneumonia COPD.
• Mortenson EM, et al. Respir Res 2005 6 82.
• Mancini GB, et al. J Am Coll Cardiol 2006 47
  2554.

4
Purpose

• Objective compare influenza COPD mortality
  risks for low-dose moderate-dose statins users
  nonusers.
• Design matched cohort study 2 case-control
  studies.
• Data Lovelace Patient Database.
• Deidentified, longitudinal db of health-care
  encounter data comprising several HMOs.
• Time period 92Jan01 03Dec31.

5
Matched cohort

• E 90 cumulative days statin exposure prior to
  death or disenrollment (N19,058).
• Further classified as low daily dose (lt4 mg/d) or
  moderate dose (4 mg/d).
• Mean mg/d for 3-month 1-yr period after statin
  initiation.
• E- 3 HMO members w/o hx of statin therapy
  matched to each E based on gender, birth yr,
  HMO enrollment period (N57,174).
• Enrollment period matching permitted observation
  of events for both groups during closely
  comparable time periods.

6
Outcomes

• Pneumonia or influenza death.
• Unspecified pneumonia influenza deaths.
• COPD deaths.

7
Analysis

• Logistic proportional hazards reg.
• Adjusted for
• Duration of enrollment before initiation of
  statin therapy (phase 1).
• Phase 1 Charlson comorbidity index.
• of different meds taken during phase 1.
• Receiving 3 influenza vaccinations after statin
  therapy initiation (phase 2).

8
Case-control studies

• E had more CV comorbidities than E-.
• Adjustment for these could have inadvertently
  affected other outcomes.
• Pneumonia/influenza study
• D Hospital discharge of deceased pneumonia or
  influenza.
• D- Surviving patients w/ 1 in/outpatient visits
  for pneumonia or influenza.
• COPD study
• D Hospital discharge of deceased COPD.
• D- Surviving COPD patients w/ 2 in/outpatient
  visits.

9
Analysis

• Logistic reg.
• Adjusted for
• Sex.
• Birth year.
• Duration of phase 2 enrollment.

10
Results matched cohortlogistic regression
11
Results matched cohortproportional hazards
12
Results case-control study
13
Concerns

• Observational study.
• Adjusted for few covariates.
• Unmeasured confounders?
• Confounding by indication.
• unlikely explanation since it would affect both
  low- and moderate-dose statin users.
• Large cohort (Ngt75,000).
• Propensity scores?
• Sensitivity analysis?

14
Concerns (2)

• Motivation for case-control studies unclear.
• Results similar to cohort study is not a
  validation of results.
• Since the reductions were observed in both the
  cohort and case-control studies, it is unlikely
  they could be due to artifacts of either study
  design or the analysis.
• Cases controls sampled from same db as matched
  cohort.
• Case-control studies had 2xs events as matched
  cohort.

15
Concerns (3)

• Logistic reg analysis of survival data.
• Ignores censoring.
• Cannot account for time-dependent covariates.
• Immortal-time bias Ayumi!

16
(No Transcript)