Epidemiology research using administrative database: Poisson regression
Source:    Publish Time: 2013-01-06 06:05   3743 Views   Size:  16px  14px  12px
Author: Xuanqian Xie Ho et al. published an article on the increased adverse events after stopping clopidogrel (Ho PM

Author: Xuanqian Xie

Ho et al. published an article on the increased adverse events after stopping clopidogrel (Ho PM, Peterson ED, Wang L, Magid DJ, Fihn SD, Larsen GC, Jesse RA, Rumsfeld JS. Incidence of death and acute myocardial infarction associated with stopping clopidogrel after acute coronary syndrome. JAMA. 2008; 299(5):532-9). I used an administrative database to examine whether their conclusions are valid in other datasets.

1. Enrollment of study patients

2. Instantaneous event risk of death and AMI over time when using/stopping clopidogrel

The meaning of time zero is different between figure 2 and figure 3, which are the day beginning using clopidogrel and the day stopping using clopidogrel, respectively. Therefore, I drew them in two graphs, rather than in one graph.

3. Statistical methods

 

According follow-up time of using and stopping clopidogrel, patients are distinguished into four groups, using clopidogrel from filling prescription day to 90 days (“use3” group), using clopidogrel from 91 days to 365 days (“use9” group), the first 90 days of stopping clopidogrel (“stop3” group) and the subsequent 275 days (91-365 days) of stopping clopidogrel (“stop9” group).  Our interest is whether the cluster adverse events is correlated with stopping clopidogrel, so our analyses focus on incidence rate ratios (IRR) of stop3 vs. stop9 and stop3 vs. use9. Poisson regression model was used to calculate crude/adjusted IRR, and incidence rate per 1000 patient-days.

4. Table 1: Baseline characteristics of study populations

Characteristic

Using clopidogrel

(period, 0-90 d) (N=18,255) “Use3” Group

Using clopidogrel

(period, 91-365 d) (N=3,535)  “Use9” group

Stopping clopidogrel (period, 0-90 d)

(N=17,913) “Stop3”group

Stopping clopidogrel (period, 91-365 d)

(N=15,435) “Stop9”group

Demographics

 

 

 

 

Age, year

 

 

 

 

    Mean (SD)

66.5(11.0)

66.4(11.1)

66.4(11.0)

66.4(11.0)

    Median (range)

68(16-99)

68(20-95)

68(16-99)

68(24-95)

Male sex, No. (%)

11,962(65.53)

2,341(66.22)

11,744(65.56)

10,111(65.51)

Comorbidities

No. (%)

 

 

 

 

hypertension†

9,545(52.29)

1,910(54.03)

9,353(52.21)

7,987(51.75)

Diabetes mellitus#

4,441(24.33)

870(24.61)

4,313(24.08)

3,619(23.45)

Heart failure   

3,767(20.64)

683(19.32)

3,641(20.33)

3,102(20.10)

Stroke

750(4.11)

149 (4.21)

721(4.03)

580(3.76)

Ischemic heart disease

16,414(89.92)

3,160(89.39)

16,099(89.87)

13,846(89.71)

Peripheral vascular disease

2,137(11.71)

384(10.86)

2,051(11.45)

1,696(10.99)

Renal diseases#

1,031(5.65)

202(5.07)

988(5.52)

817(5.29)

Glaucoma#

714(3.91)

169(4.78)

697(3.89)

594(3.85)

Medications

No. (%)

 

 

 

 

Diuretic drug

2,358(12.92)

680(19.24)

2,305(12.87)

1,891(12.25)

ACE

10,205(55.90)

2,465(69.73)

9,975 (55.69)

8,409(54.48)

Beta blocker

13,504(73.97)

2,948(83.39)

13,238(73.90)

11,347(73.51)

CCB

5,627(30.82)

1,253(35.45)

5,493(30.66)

4,615(29.90)

Stain

13,609(74.55)

3,186(90.13)

13,344(74.49)

11,321(73.35)

Aspirin

15,288(83.75)

3,217(91.00)

15,008(83.78)

12,915(83.67)

Other factors

 

 

 

 

Stent No. (%)

17,547(96.12)

3,400(96.18)

17,222(96.14)

14,853(96.23)

Hospitalization within1 month before stopping clopidogrel

1,902(10.42)

485(13.72)

1,683(9.40)

1,262(8.18)

Duration using clopidogrel post-hospitalization, days ‡

 

 

 

 

    Mean (SD)

84.6(131.8)

297.6(173.9)

83.2(129.3)

66.4(102.0)

    Median (range)

30(2-1,876)

300(77-1,876)

30(2-1,876)

30(2-1,530)

Duration of hospitalization within1 year before using clopidogrel, days Mean (SD)

10.3(11.0)

9.9(10.1)

10.2(11.0)

10.0(10.6)

Year receiving PCI treatment. No. (%)

 

 

 

 

1999

804(4.40)

40(1.13)

795(4.44)

738(4.78)

2000

3,363(18.42)

105(2.97)

3,309(18.47)

3,081(19.96)

2001

4,439(24.32)

230(6.51)

4,375(24.42)

4,021(26.05)

2002

4,919(26.95)

912(25.80)

4,826(26.94)

4,237(27.45)

2003

4,730(25.91)

2,248(63.59)

4,608(25.72)

3,358(21.76)

 

†: Co-morbidities are based on hospitalization records within one year prior to using clopidogrel.  

#: On the basis of either hospitalization records or medications are used.

‡: Medications are used during receiving clopidogrel treatment.

 

5. Table 2.  Incidence of AMI and death

 

No. at risk

No. of events

Incidence rate per 1000 patient-days (95% CI)

Using clopidogrel (period, 0-90 d)

18,255

262

0.3291(0.2310-0.4688)

Using clopidogrel (period, 91-365 d)

3,535

63

0.1015(0.0493-0.2089)

Stopping clopidogrel (period, 0-90 d)

17,913

393

0.2657(0.1990-0.3547)

Stopping clopidogrel (period, 91-365 d)

15,435

556

0.1434(0.1125-0.1829)

 

 

 

6. Table 3. Crude/adjusted incidence rate ratios

 

Crude incidence rate ratios (95% CI)

Adjusted incidence rate ratios (95% CI)

Exposure variable

 

 

stop3 vs. use9

2.6176(1.2990-5.2744)

3.2497(1.5265-6.9184)

stop3 vs. stop9

1.8522(1.3180-2.6028)

1.7767(1.2864-2.4538)

stop9 vs. use9

1.4132(0.7115-2.8071)

1.8291(0.8586-3.8964)

 

 

 

Demographic Characteristics

 

 

Sex,  Male vs. Female

1.0107(0.7050-1.4488)

1.2641(0.9082-1.7594)

Age

 

 

< 55

0.6543 (0.3496-1.2244)

0.7860(0.4402-1.4035)

55-64

0.8254(0.4963-1.3726)

0.9083(0.5686-1.4511)

65-74

1.0000

1.0000

75-84

1.8464(1.2367-2.7567)

1.5838(1.0895-2.3023)

> 85

2.6249(1.2199-5.6481)

2.0737(1.0124-4.2475)

 

 

 

Comorbid conditions

 

 

Hypertension YES vs. NO

1.3824(0.9821-1.9460)

--

Diabetes YES vs. NO

2.0120(1.4156-2.8595)

1.5970(1.1499-2.2180)

Congestive heart failure  YES vs. NO

2.4564(1.7375-3.4726)

1.5736(1.1186-2.2136)

Cerebrovascular disease YES vs. NO

1.8232(0.9154-3.6312)

--

Ischemic heart disease YES vs. NO

1.2430(0.6732-2.2949)

--

Peripheral vascular disease YES vs. NO

2.3444(1.5492-3.5479)

1.4676(0.9963-2.1619)

Renal disease YES vs. NO

3.5480(2.2443-5.6090)

1.7510(1.1188-2.7404)

 

 

 

Medication use

 

 

ACE YES vs. NO

1.1852(0.8376-1.6771)

--

Beta-Blocker  YES vs. NO

0.7976(0.5495-1.1577)

--

CCB  YES vs. NO

1.3898(0.9762-1.9788)

--

Stain YES vs. NO

0.7141(0.4951-1.0299)

--

Aspirin YES vs. NO

0.8556(0.5478-1.3366)

--

 

 

 

Other risk factors

 

 

Hospitalization within1 month before stopping clopidogrel  (YES vs. NO)

3.8187 (2.6317-5.5411)

2.8213(1.9624-4.0563)

stent (YES vs. NO)

0.8432(0.3693-1.9254)

 

Days using clopidogrel  =< 28 vs. > 90

1.1621(0.7191-1.8779)

0.9033(0.5529-1.4756)

Days using clopidogrel   29-90 vs. > 90

0.9294(0.58931.4657)

0.7886(0.4920-1.2638)

Days staying in hospital: >10  vs. < 10

2.2510(1.6156-3.1362)

1.3969(0.9953-1.9605)

CI= confidence interval.

Statistically significant associations are shown in bold (p value < 0.05).

--: no adjusted incidence rate ratios, because this variable was excluded from the final regression model.

 

7. Sensitivity analysis

a. Only include the all cause mortalities events.

b. Include all AMI events, both fatal and non-fatal (exclude mortalities from other causes).

c. For some observations, there are some gaps (a few days without using clopidogrel) within periods of patients using clopidogrel. Currently, we defined once patients without using clopidogrel 14 days as stopping clopidogrel in base analysis. We apply 7 days and 21 days in sensitivity analysis.

d. We randomly choose 5% extend 7 days of using clopidogrel and 5% decrease 7days of using clopidogrel (to keep the overall balance of dataset). 

Similar as those, 5% extend 14 days and 5% decrease 14 days, 5% extend 28 days and 5% decrease 28 days, 10% extend 7 days and 10% decrease 7 days, 10% extend 14 days and 10% decrease 14 days, and 10% extend 28 days and 10% decrease 28 days.

e. Only include patients using clopidogrel 28 days or more.

 

Results in sensitivity analysis are similar as those in Base model. Therefore, our results are robust.