### Harm Scenario

You have been called to a meeting with the Health Authority Chief Executive to discuss concerns about long term effects of the oral contraceptive pill, particularly following a recent national pill scare. You formulate the question, in women using oral contraceptives: “What is the risk of death?”

Following the local librarian’s advice, you search MEDLINE using the following text words (cohort$and oral contraceptive$ and relative risk \$).tw for the best quality cohort study you can find that specifically addresses the question of safety of the pill. Amongst a large number of much smaller studies using poorer quality designs, one recent study stands out in both size and duration ( BMJ 1999;318:96-100 ) which you download from the British Medical Journal’s free website: http://www.bmj.com/.

• Are the results of this harm study valid?
• Are the results of this harm study important?
• Should these valid, important results of this study about a potentially harmful treatment change the treatment of your population?

### Completed Harm Worksheet for Evidence-Based Purchasing

#### Citation

Beral V, Hermon C, Kay C, Hannaford P, Darby S, Reeves G. Mortality associated with oral contraceptive use: 25 year follow up of cohort of 46,000 women from Royal College of General Practitioners’ oral contraception study.
BMJ 1999;318:96-100

#### Are the results of this harm study valid?

Were there clearly defined groups of patients, similar in all important ways other than exposure to the treatment or other cause?
Yes. Although the article does not contain a table comparing characteristics of women in the two groups, the results controlled for possible confounding factors: age, social class, parity and smoking (p.100).
Were treatment exposures and clinical outcomes measured the same ways in both groups (e.g., was the assessment of outcomes either objective (e.g., death) or blinded to exposure)?
Yes. In both groups, exposure to oral contraception was measured from GPs’ prescribing records every six months, and death from all causes was measured from English and Scottish central registers (p.96).
Was the follow-up of study patients complete and long enough?
Yes. Although only about 75% of the original participants were followed up, the other 25% left only because their GPs had left the study before original participants had been flagged – not because of systematic differences of the patients. Subsequent checking showed that those who left the study had similar mortality rates to the 75% who remained.
##### Do the results meet “causation criteria”? (also known as the Bradford Hill criteria)?
Is it clear that the exposure preceded the onset of the outcome?
Yes.
Kind of. The study did not control for different strengths or types of oral contraception (most were the combined pill containing 50lg of oestrogen (p.99), but it did take into account duration of exposure (Tables 4 & 5).
Is there positive evidence from a “dechallenge-rechallenge” study?
No.
Is the association consistent from study to study?
Don’t know (would need to look at other studies).
Does the association make biological sense?
Yes.

#### Are the valid results from this harm study important?

(death from all causes*)
Totals
Present (case) Absent (case)
Exposed to the Treatment Yes (Cohort) 0.0018
a
0.9982
b
a + b
No (Cohort) 0.0019
c
0.9981
d
c + d
Totals a + c b + d a + b + c + d
* note that you could use this table to analyse risks of death from specific causes using the data in Tables 4 & 5. These figures are taken from the “ever vs. never” used columns in table 1, with the person-year denominators (517,519 vs 335,998) in the ever vs never groups on page 97.
** note that the death rates uses person-years as the denominator, rather than people. That is, if you take the pill for one year, you can expect a death rate from all causes of 0.18% = 945 deaths/517,519 person years [page 97]
*** In this study, relative risk = RR = 0.0018/0.0019 = 0.95, and when adjusted for several baseline differences, RR to 1.1 (95% CI 0.9-1.1, p value of 0.7) i.e. there is no significant difference in death between the two groups.

There are many different figures that you could analyse from this large study. For example, you could pull out the risks from a specific form of death. One possible learning point to draw out is the use of the more accurate measure of exposure, person-years, as a denominator, rather than just people.

### Oral contraceptive pills has no lasting effect on mortality

The oral contraceptive pill does not seem to have lasting effects on mortality, particularly after 10 years’ cessation.

#### Citation

Beral V, Hermon C, Kay C, Hannaford P, Darby S, Reeves G. Mortality associated with oral contraceptive use: 25 year follow up of cohort of 46,000 women from Royal College of General Practitioners’ oral contraception study.
BMJ 1999;318:96-100

What is the risk of death in women who are taking oral contraceptives?

#### Search Terms

MEDLINE search using mortality, oral contraception, exp cohort studies.

#### The Study

25 year prospective cohort study, starting May 1968, of 46,000 women from 1400 General Practices throughout Britain (75% follow up due to GPs dropping out before routine flagging occurred, not due to systematic differences in women leaving the study). GPs provided routine information about prescribing oral contraceptives, pregnancies, new illness and deaths. Deaths were also recorded from the English and Scottish central registers and coded according to ICD-8, up to 31/12/93. Exposure was measured in person-years, and years from cessation were noted. The analysis controlled for social class, age, parity and smoking.

#### The Evidence

(death from all causes*)
Totals
Present (case) Absent (case)
Exposed to the Treatment Yes (Cohort) 0.0018
a
0.9982
b
a + b
No (Cohort) 0.0019
c
0.9981
d
c + d
Totals a + c b + d a + b + c + d
* note that you could use this table to analyse risks of death from specific causes using the data in Tables 4 & 5. These figures are taken from the “ever vs. never” used columns in table 1, with the person-year denominators (517,519 vs 335,998) in the ever vs never groups on page 97.
** note that the death rates uses person-years as the denominator, rather than people. That is, if you take the pill for one year, you can expect a death rate from all causes of 0.18% = 945 deaths/517,519 person years [page 97]
*** In this study, relative risk = RR = 0.0018/0.0019 = 0.95, and when adjusted for several baseline differences, RR to 1.1 (95% CI 0.9-1.1, p value of 0.7) i.e. there is no significant difference in death between the two groups.