1. Self-Controlled Studies Case Crossover Case Series
Extreme Cohort Stratification
with and without Sampling
September 2013
Alec Walker
With thanks to Sonia Hernandez-Díaz

2. What “Triggering” Means
Ref. Maclure M. Am J Epidemiol 1991:133:

3. What “Triggering” Means
Hypothesized Curves of Incidence Rate in Temporal Relation to Exposure
True
Approximated
Exposure
Baseline Incidence
Risk Periods
Ref. Maclure M. Am J Epidemiol 1991:133:

4. Comparative vs. Crossover Designs
Treatments are compared between individuals. Baseline comparability of groups is sought through random allocation (RCT) or through careful subject selection and covariate control.
Randomized trial
Cohort Study
Today we need to consider another type of trial as the counterpart for our observational studies, the cross-over design.
“Self-controlled” analyses, in which treatments are compared within individuals, are common in intervention trials, but are also possible in purely observational studies. The design greatly enhances the baseline similarity of comparison groups.
Randomized trial Cross-over design
Cohort Study with Changing Exposure

5. Cohorts with Changing Exposure
Black: Unexposed person-time Red: Exposed person-time Dot: Case event
Case-crossover studies are matched cohort studies that take personal matching to its limit:
The individual forms the unit of stratification, and the comparison is between different exposure windows within individuals.
 End of fu no event
 Event
Time 

6. You Can Match Analyses on Person
Case-crossover studies are matched cohort studies that take personal matching to its limit:
The individual forms the unit of stratification, and the comparison is between different exposure windows within individuals. 6

7. Persons with no event drop out
Case-crossover studies are matched cohort studies that take personal matching to its limit:
The individual forms the unit of stratification, and the comparison is between different exposure windows within individuals. 7

8. Sampling Frames for Controls
Choose past days from the same person
Case-crossover studies are matched cohort studies that take personal matching to its limit:
The individual forms the unit of stratification, and the comparison is between different exposure windows within individuals. 8

9. Compare the Exposure at Case Occurrence to Sampled Exposure(s) in the Past
Case-crossover studies are matched cohort studies that take personal matching to its limit:
The individual forms the unit of stratification, and the comparison is between different exposure windows within individuals. 9

10. Compare the Exposure at Case Occurrence to Sampled Exposure(s) in the Past
Control window
Case window
Case window=effect period

11. Celecoxib, Naproxen and GI Bleeding
versus
Naproxen versus No Treatment
MD-perceived risk of peptic ulcer & bleeding (PUB)
True risk of PUB
PUB
Hospital
Admission

12. Confounding by Contraindication
Celecoxib
versus
Naproxen
versus No Treatment
MD-perceived risk of peptic ulcer & bleeding (PUB)
The physician’s belief in the patient’s risk for peptic ulcer and bleeding cannot be measured directly.
True risk of PUB
PUB
Hospital
Admission

13. Self-Matching by Case-Crossover
Case-crossover studies stratify the analysis on individuals, looking at exposure history leading up to an event. X X X X
Key
Celecoxib
Naproxen
No Therapy
Event
Individuals with no event drop out of the analysis. X

14. Case-Crossover Matching
Matching on person and examining time on and time off therapy has the effect of removing the confounding effect of time-invariant (or time-slowly-variant) covariates.
Celecoxib
versus
Naproxen
No Treatment
MD-perceived risk of peptic ulcer & bleeding (PUB)
True risk of PUB
PUB
Hospital
Admission

15. Case-Crossover Mechanics
Use only for events that may be triggered by a discrete exposure
For each case, identify the presence or absence of the exposure during the postulated trigger period
Identify the “expected” exposure in the trigger period from
Usual frequency of exposure over a long time
Documented occurrence of exposure during one or more comparable times in the past
Summarize Observed-to-Expected ratio across all cases

16. Case and Control Windows
Case window: period preceding the event (GI bleeding) during which the exposure (e.g. celecoxib) may have altered the risk
Control window(s): periods of the same length as, and not overlapping with, the case window that provide an estimate of the expected frequency of exposure for each case.
The core study technique is to identify cases, then ascertain exposure status in the case window and at earlier points in time – the control windows. 16

17. Estimating the Relative Risk
For dichotomous exposures
Form the matched 2x2 table 
Place each case according to exposures in the case and control windows
Mantel-Haenszel odds ratio for matched sets
Reduces to ratio of counts in discordant exposure windows ( f10 / f01 ) when there is one control
Concordant case-control windows are uninformative
Otherwise
Conditional logistic regression is used when
There are several different exposures or exposure levels
There are concurrent time-varying confounders
Case Window
Control Window Exposed
Exposed
Yes No
f10
f01
Dichotomous exposures and two risk periods can be handled as described in the preceding paragraph.
Polytomous, or continuous exposure measures and multiple risk periods require a conditional logistic regression analysis.
Risk periods of varying length can be analyzed using the techniques of stratified analysis of person-time data.
The matched analysis may also be stratified by fixed personal characteristics to look for factors that modify the effect of the exposure on the risk of outcome.

18. Pharmacoepidemiology and Drug Safety 2011; 20: 763–771
Risk of hospitalization for upper GI adverse events associated with nonsteroidal anti‐inflammatory drugs…
Chia‐Hsuin Chang, Hsi‐Chieh Chen, Jou‐Wei Lin, Chuei‐Wen Kuo, Wen‐Yi Shau and Mei‐Shu Lai
Purpose This study aimed to evaluate the risks of upper gastrointestinal (GI) adverse events across a variety of oral and parenteral coxibs and nonselective nonsteroidal anti‐inflammatory drugs (nsNSAIDs) in the general population of Taiwan.
Methods In a case‐crossover study, all patients aged ≥20 years who were hospitalized for upper GI adverse events (peptic ulcer and bleeding; gastritis and duodenitis) in were identified ... For each patient, the case period was defined as 1–30 days and the control period as 31–60 days before the date of hospitalization. Outpatient pharmacy prescription database was searched for individual NSAID use during the case and control periods. A conditional logistic regression model was applied ...
Results A total of patients hospitalized for upper GI adverse events were included. The adjusted OR was 1.52 (95%CI: 1.27–1.82) for celecoxib and (95%CI: 2.44–2.69) for oral nsNSAIDs…
Conclusion Use of celecoxib and all nsNSAIDs studied was associated with a greater risk of upper GI toxicity as compared with nonuse…
Pharmacoepidemiology and Drug Safety 2011; 20: 763–771

19. *Conditional logistic regression adjusted for important potential time‐varying confounding variables including selective serotonin reuptake inhibitors, other antidepressants, calcium channel blockers, nitrates, systemic corticosteroids, low‐dose aspirin, proton pump inhibitors, histamine 2 receptor blockers, and sucralfate.

20. Pharmacoepidemiology and Drug Safety 2011; 20: 763–771
Case Window
Control Window Exposed
Exposed
Yes No
f10
f01
Pharmacoepidemiology and Drug Safety 2011; 20: 763–771

21. Pharmacoepidemiology and Drug Safety 2011; 20: 763–771
Case Window
Control Window Exposed
Exposed
Yes No
413
f01
Pharmacoepidemiology and Drug Safety 2011; 20: 763–771

22. Pharmacoepidemiology and Drug Safety 2011; 20: 763–771
Case Window
Control Window Exposed
Exposed
Yes No
413
232
Pharmacoepidemiology and Drug Safety 2011; 20: 763–771

23. Pharmacoepidemiology and Drug Safety 2011; 20: 763–771
Case Window
Control Window Exposed
Exposed
Yes No
413
232
RRcrude = 413/232
= 1.78
Pharmacoepidemiology and Drug Safety 2011; 20: 763–771

24. *Conditional logistic regression adjusted for important potential time‐varying confounding variables including selective serotonin reuptake inhibitors, other antidepressants, calcium channel blockers, nitrates, systemic corticosteroids, low‐dose aspirin, proton pump inhibitors, histamine 2 receptor blockers, and sucralfate.

25. Another Option for Case-Crossover Design: Usual frequency
Ref. Mueller JE. Am J Cardiol 2000;86(suppl):14F–18

26. Another Option for Case-Crossover Design: Usual frequency
Less frequently done because of the difficulty of directly ascertaining usual frequency of exposure.
Ref. Mueller JE. Am J Cardiol 2000;86(suppl):14F–18

27. Case-Series Analyses Originally developed for analysis of vaccination.
Same data set-up as case-crossover
Rationale: Individually stratified cohort analysis
Only persons with both exposure and event contribute person-time
For display, person-time is summed over the contributing individuals collectively.
For analysis, the matching is taken into consideration.
The person-time in the people with events may not have the same distribution as in the general population in non-vaccine studies.
Example: Murphy et al. N Engl J Med 2001;344: (next slide)

29. Exposure-time categories for case-series analysis of individuals in motor vehicle collisons,UK, 1986–2004.
Exposure-time categories for case-series analysis of individuals in the United Kingdom, 1986– = unexposed time; 2 = 4-week preexposure period, up to and including the date of the first prescription; 3 = first 4 weeks following the first prescription; 4 = remainder of the course of treatment; 5 = first 12 weeks after the end of the course of treatment; 6 = second 12 weeks after the end of the course of treatment. THIN, The Health Improvement Network.
1 = unexposed time; 2 = 4-week preexposure period, up to and including the date of the first prescription; 3 = first 4 weeks following the first prescription; 4 = remainder of the course of treatment; 5 = first 12 weeks after the end of the course of treatment; 6 = second 12 weeks after the end of the course of treatment.
Gibson J E et al. Am. J. Epidemiol. 2009;aje.kwn364
American Journal of Epidemiology © The Author Published by the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please

30. 1 = unexposed time; 2 = 4-week preexposure period, up to and including the date of the first prescription; 3 = first 4 weeks following the first prescription; 4 = remainder of the course of treatment; 5 = first 12 weeks after the end of the course of treatment; 6 = second 12 weeks after the end of the course of treatment.
Gibson J E et al. Am. J. Epidemiol. 2009;aje.kwn364

31. Exposure-time categories for case-series analysis of individuals in motor vehicle collisons,UK, 1986–2004.
Exposure-time categories for case-series analysis of individuals in the United Kingdom, 1986– = unexposed time; 2 = 4-week preexposure period, up to and including the date of the first prescription; 3 = first 4 weeks following the first prescription; 4 = remainder of the course of treatment; 5 = first 12 weeks after the end of the course of treatment; 6 = second 12 weeks after the end of the course of treatment. THIN, The Health Improvement Network.
There are too many theories of cause-and-effect for the pre-exposure periods to be readily interpretable. (Did the effect possibly cause or prevent the exposure?)
Gibson J E et al. Am. J. Epidemiol. 2009;aje.kwn364
American Journal of Epidemiology © The Author Published by the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please

32. Similar concerns apply to the earlier reference period.
Exposure-time categories for case-series analysis of individuals in motor vehicle collisons,UK, 1986–2004.
Exposure-time categories for case-series analysis of individuals in the United Kingdom, 1986– = unexposed time; 2 = 4-week preexposure period, up to and including the date of the first prescription; 3 = first 4 weeks following the first prescription; 4 = remainder of the course of treatment; 5 = first 12 weeks after the end of the course of treatment; 6 = second 12 weeks after the end of the course of treatment. THIN, The Health Improvement Network.
Similar concerns apply to the earlier reference period.
Gibson J E et al. Am. J. Epidemiol. 2009;aje.kwn364
American Journal of Epidemiology © The Author Published by the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please

33. Key Points
Case-crossover studies compare exposure at the time of an event to times in the past
Case-series studies compare event frequencies at different exposure windows in exposed persons with events
Very efficient, not always feasible
Works for easily ascertained, intermittent exposures whose associated risks rise and fall quickly
Requires that the onset time of the outcome event can be pinned down with a precision that is substantially tighter than the width of the exposure window
Cautions on interpreting pre-exposure time in case-series studies

34. Thank You!

35. Case-crossover time categories a study of motor vehicle collisions in the UK, 1986–2004.
Summary of case-crossover time categories for individuals in the United Kingdom, 1986–2004. R = at-risk period (1–28 days before motor vehicle crash (MVC)); 1 = control period 1 (29–56 days before MVC); 2 = control period 2 (57–84 days before MVC); 3 = control period 3 (85–112 days before MVC); 4 = control period 4 (113–140 days before MVC); 5 = control period 5 (141–168 days before MVC). THIN, The Health Improvement Network.
Gibson J E et al. Am. J. Epidemiol. 2009;aje.kwn364
American Journal of Epidemiology © The Author Published by the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please