Results
The study was initiated in November 2000 and recruitment was completed in June 2004. The follow-up of study outcomes was completed in December 2005.
The study was conducted in seven European countries where DRSP-containing OCs are marketed, i.e. Austria, Belgium, Denmark, France, Germany, the Netherlands, and the United Kingdom.
A total of 59,510 women (see Table 1) was enrolled by 1,113 active study centers.
Overall, 836 out of these 59,510 women (1.4%) had to be excluded because:
- they refused to sign the informed consent form;
- they were enrolled two or more times by one or more study centers;
- they continued to use their old OC (long-term users); or
- they did not start OC use. The remaining data from 58,674 were analyzed.
At baseline, 16,534 women were prescribed DRSP-containing OCs, 15,428 women LNG-containing OCs, 26,341 women ‘Other OCs’, and 371 women NOHC. Although the study was aimed primarily at users of oral contraceptives, the latter group was also analyzed separately to acquire some data that represents the baseline characteristics of women who switched to NOHC after study entry.
Overall, the 58,674 women were followed for 142,475 women-years (WY) of observation. Those women who stopped or interrupted OC use (e.g., because of planned pregnancy, no partner, or adverse events) were followed in the same way as continuing OC users.
Follow-up for periods of non-use of hormonal contraceptives accounts for 25,767 WY, and follow-up for periods of use accounts for 116,708 WY of exposure.
Further details are summarized in Table 1.
On average, users of DRSP-containing OCs had a slightly shorter exposure than the other OC cohorts because of an increasing proportion of recruited users of DRSP-containing OCs during the performance of the study (due to increasing market penetration of DRSP-containing OCs) as well as a higher proportion of women who stopped using any contraceptive method because of planned pregnancy, no partner, or age.
The switching rates from one cohort to another (e.g., switch from DRSP-containing OCs to LNG-containing OCs) were comparable: 20.8%, 22.3%, and 26.9% of the initial DRSP, LNG, and ‘Other OCs’ users, respectively, switched to another OC or NOHC brand.
A total of 11.5%, 15.8%, and 20.0% of the DRSP, LNG, and ‘Other OC’ users reported that they switched to another brand or stopped OC use because of adverse events or intolerance reactions.3.1 Baseline characteristics of the user cohorts
Table 2 depicts for each user cohort (i.e.; DRSP-containing OCs, LNG-containing OCs, ‘Other OCs’, and NOHC) the number of women with baseline information (N), the corresponding mean age with standard deviation, mean weight with standard deviation, and mean BMI with standard deviation. Mean age was similar for the three OC cohorts.
The proportions of women in the age groups of under 20, 20-29, 30-39, and over 40 years, namely 30%, 42%, 21%, and 7% respectively, reflected the typical age profile of OC users in Europe. The NOHC users were about 2.5 years older than the OC users on average.
Mean weight and mean BMI of all cohorts were in the same order of magnitude. However, the values for DRSP users were slightly higher than those for the other cohorts.
The distribution of risk factors and preexisting diseases was comparable among the OC user cohorts.
However, two exceptions were found:
- The percentage of obese (BMI > 30) women and
- the percentage of women with preexisting arrhythmia were higher in the DRSP-containing OC cohort compared to the ‘Other
OCs’ cohort and the LNG-containing OC cohort (rate ratios approx. 1.5).
Since an elevated BMI and preexisting arrhythmia are well-known risk factors for VTE and recurrent arrhythmia, respectively, the DRSP cohort had a slightly higher baseline risk for VTE and arrhythmia compared to the other two OC cohorts.
However, in absolute numbers the differences were small and the preferential prescribing pattern identified here could only slightly increase the incidence of VTE and arrhythmia for the DRSP cohort. Figure 1 shows some selected risk factors and preexisting diseases for the three OC cohorts.3.2 Loss to follow-up
A total of 58,674 study participants was followed up for 142,475 WY of observation. In sum, 1,401 out of 58,674 women, or 2.4%, were lost to follow-up.
The percentage of participants lost at each of the semiannual follow-ups was less than 0.5%. A comparison of the loss to follow-up rate among cohorts showed only very minor differences.
The loss to follow-up per cohort was 2.4%, 2.7%, and 2.2% for the DRSP, LNG, and ‘Other OCs’ cohorts, respectively. On account of the low and symmetrical loss to follow-up rates in this study, a relevant bias associated with loss to follow-up can be largely excluded.3.3 Serious Adverse Events (SAEs)
Overall, 5,310 SAEs (i.e., adverse events that result in death, a life-threatening experience, inpatient hospitalization, persistent or significant disability/incapacity, or require medical/surgical intervention to prevent one of said outcomes) were reported by the study participants.
A total of 965 SAEs (337/104 WY) were reported in the DRSP cohort, 1,059 SAEs (337/104 WY) in the LNG cohort, 1,815 SAEs (345/104 WY) in the ‘Other OCs’ cohort, and 150 SAEs (346/104 WY) in the NOHC cohort.
These overall SAE rates are similar. Rate ratios between the cohorts are close to unity (e.g., DRSP/LNG = 1.00, and DRSP/’Other OCs’ = 0.98). The study participants who stopped all use of hormonal contraception reported 1,331 SAEs. This corresponds to a reporting rate of 517/104 WY and is substantially higher than the reporting rate for the OC cohorts (e.g., no use/DRSP = 1.53).
Closer analysis, however, shows that these differences are primarily a matter of SAEs in connection with pregnancy, delivery, or puerperium (data not shown).
Figure 2 shows the SAEs by organ system, categorized according to ICD-10 codes. A comparison between the OC cohorts shows no notable differences.
The DRSP/LNG and DRSP/‘Other OCs’ rate ratios for the individual disease categories vary from 0.65 (diseases of the eye) to 1.17 (infectious diseases).
A direct comparison of the 14 disease categories for the three OC cohorts shows that the pattern found in this study can be easily attributed to chance and provides no indication of a heightened SAE risk for users of DRSP-containing OCs --- in comparison to other OCs --- in any of the 14 disease categories.
The incidence of malignant neoplasms in the DRSP cohort was similar or lower than in the other two OC cohorts (3.5, 3.8, and 4.8 events per 10,000 WY for the DRSP, LNG, and ‘Other OCs’ cohorts, respectively).
As is typical for this age group, the most common malignant neoplasms diagnosed during the study were cancer of the breast as well as of the cervix and corpus uteri. Here again, the incidence for the DRSP cohort was similar or slightly lower than for the other cohorts (see section 4, ‘Discussion’).3.4 Fatal outcomes
A total of 32 deaths occurred during 142,475 WY of observation, corresponding to a rate of 2.2/104 WY. Mortality was lowest in the DRSP cohort: 1.4, 2.5, 1.7, 2.5, and 3.9 cases per 10,000 WY were reported for the DRSP, LNG, ‘Other OCs’, NOHC, and ‘No Use’ cohorts, respectively. Differences between the treatment groups were not statistically significant. The results of the Cox regression analysis are shown in Table 3.
The Advisory Council analyzed all cases of death with special attention to the causal relation between OC/NOHC use and fatal outcomes. The study medication was blinded for this analysis.
A total of 5 women died due to cardiovascular diseases which is the disease category of highest interest for this study. Two of these women --- one woman used a DRSP-containing OC and one woman used an ‘Other OC’ --- died following the rupture of an aortic and a cerebral arterial aneurysm, respectively.
In these two cases, the Advisory Council considered a causal relationship to be unlikely. The other three cases were reported in the LNG cohort; one was related to a VTE and two to an acute myocardial infarction.
According to the Advisory Council’s assessment, these three cases were possibly related to the treatment.
In all other cases, the Advisory Council considered a causal relationship to be at least unlikely.3.5 VTE
A total of 118 VTE was observed in the study. A similar incidence was found in all OC/NOHC cohorts: DRSP cohort 26 cases and 9.1 VTEs per 104 WY, LNG cohort 25 cases and 8.0 VTEs per 104 WY, ‘Other OCs’ cohort 52 cases and 9.9 VTEs per 104 WY, and NOHC cohort 3 cases and 7.4 VTEs per 104 WY (see Table 2).
The confidence intervals for the OC cohorts were relatively narrow (~ -33% and +44% of the point estimate), while the point estimate for the NOHC cohort showed a wide confidence interval due to the low exposure (-84% and +191% of the point estimate).
For 26 of the 118 VTE cases (22%), a pulmonary embolism (PE) was observed (DRSP cohort 7 cases, LNG cohort 7 cases, ‘Other OCs’ cohort 11 cases, NOHC cohort 1 case). The incidences for PE were almost identical at 2.1 to 2.5 events per 104 WY.
For 4 of the 118 VTE cases (3%), a non-septic venous sinus thrombosis (VST) was diagnosed (no case in the DRSP, LNG, and NOHC cohorts, 3 cases in the ‘Other OCs’ cohort, and 1 case in the 'no use' cohort).
The EURAS study was powered to show non-inferiority of DRSP-containing OCs regarding VTE risk in comparison to LNG-containing OCs and ‘Other OCs’. A Cox regression analysis was carried out in accordance with the final analysis plan. The following pre-defined confounder variables were included in the Cox regression model: age, BMI, duration of use, and VTE history. The results are shown in Table 3.
The adjusted hazard ratio (HRadj.) for DRSP vs. LNG was 1.05 with a confidence interval of 0.61 to 1.81. Thus a twofold higher risk of VTE during DRSP use compared to LNG use was excluded, and non-inferiority according to the study objectives was demonstrated. Because users of DRSP-containing OCs had a slightly higher baseline risk for VTE compared to the LNG population, adjustment for the pre-defined confounder variables resulted in a reduction in the HRadj., although this reduction was minimal (i.e., 0.09).
A comparison of DRSP vs. ‘Other OCs’ yielded an HRadj. of 0.77 with an upper confidence limit of 1.26.
Here, too, the higher baseline VTE risk for the DRSP cohort meant that adjustment led only to a slight reduction in the HR (-0.15). Combining the LNG cohort with the ‘Other OCs’ cohort and then comparing them with the DRSP cohort yielded an HRadj. of 0.87 with an upper confidence limit of 1.37.
This means that compared to all other OCs, an approximately 1.5-fold increase in VTE risk for DRSP users can be excluded.
A comparison of the DRSP and LNG cohorts means that a monophasic OC that contains 30mcg EE is compared with a quite non-homogenous group of LNG products (monophasic vs. sequential regimens with 15 to 50mcg EE). A total of 83% of the LNG users used a monophasic regimen at study entry (30% used a product with less than 30mcg EE, 51% a product with exactly 30mcg EE, and 2% a product with more than 30mcg EE).
The 30mcg EE/LNG products represented 53% (16,649 WY) of the LNG exposure. This exposure was sufficient to make a direct comparison of DRSP vs. LNG on the basis of 30mcg EE.
This eliminates the influences of the EE dose and the regimen, so the progestins DRSP and LNG can be directly compared.
A total of 17 VTEs was observed in the LNG/30mcg EE sub-cohort and 26 VTEs in the DRSP cohort. This corresponds to an incidence of 10.2/104 WY for the LNG sub-cohort and 9.1/104 WY for the DRSP cohort.
The Cox regression analysis yielded a crude HR for DRSP of 0.89 (95% CI: 0.48-1.63) and an adjusted HR of 0.82 (95% CI: 0.45-1.51). Thus a comparison of these homogenous groups also showed no indication that DRSP is associated with a higher VTE risk than LNG.
Rather, the narrow confidence limits suggest that the two progestins are associated with a similar VTE risk.
In the validation process for self-reported VTE, 17 events were identified that were probably not VTEs. Because these cases were also unanimously classified by the blinded adjudicators as not being VTEs, the risk of misclassification seems low. In order to assess possible error, however, a sensitivity analysis was performed, in which potential VTEs were combined with confirmed VTEs. The Cox regression analyses for DRSP vs. LNG, ‘Other OCs’ and LNG/’Other OCs’ combined yielded adjusted HRs of 1.04, 0.80, and 0.88, respectively. In summary, the inclusion of potential VTE did not lead to any modification of the conclusions regarding VTE risk for DRSP-containing OCs.
The same is true for the sub-analyses of starters (first-ever OC users) and switchers. None of the 13 VTEs in starters occurred in the DRSP cohort, and Cox regression analyses for DRSP vs. LNG, ‘Other OCs’, and LNG/’Other OCs’ combined yielded adjusted HRs of 1.10, 0.90, and 0.98 respectively.
In the 'no use' cohort, 12 VTE were observed. This corresponds to an incidence of 4.7 VTE/104 WY. In the combined OC/NOHC group, incidence was around 90% higher, at 9.1 VTE/104 WY. The 12 VTE in the 'no use' cohort include 7 VTE that occurred in the course of pregnancy.
This corresponds to an incidence of around 19.4 VTE/104 WY. Here one should consider that many of the women who stopped study participation did so because they became pregnant.
Thus the early phase of pregnancy which has a relatively low risk of VTE is overrepresented vis-à-vis late pregnancy and the first weeks following delivery, which are associated with a very high VTE risk.
As such, the observed incidence probably represents an underestimation rather than an overestimation of VTE incidence in pregnancy. Among women in the 'no use' cohort who were not pregnant, a total of 5 VTE were observed, for an incidence of 2.3 VTE/104 WY.
The results for the ‘no use’ cohort, however, are not necessarily representative for non-users who never have used OCs (cf. sections 2.6 and 4). Therefore, they must be considered as tentative.3.6 Arterial Thromboembolism (ATE)
A total of 25 ATE was observed in the study: 11 AMIs, 13 strokes, and 1 complete thrombosis of the Arteria hepatica propria. Transient ischemic attacks (TIAs) were not included among the strokes. But alternative evaluations were also performed for which the TIAs were included. These results are also briefly reported below.
In the DRSP cohort 2, LNG cohort 9, ‘Other OCs’ cohort 9, NOHC cohort 2, and 'no use' cohort 3 ATE cases were observed. This corresponds to ATE incidences of 0.7 ATE/104 WY for the DRSP cohort, and of 2.9, 1.7, 4.9, and 1.2 for the LNG, ‘Other OCs’, NOHC, und 'no use' cohorts, respectively.
In view of the rather narrow confidence intervals in the large OC cohorts, the relatively low incidence in the DRSP cohort is notable. A Cox regression analysis was carried out and the following pre-defined confounder variables were included in the Cox regression model: age, BMI, smoking, and hypertension. The results are shown in Table 3.
The adjusted hazard ratios for the comparisons DRSP vs. LNG, ‘Other OCs’, and LNG/’Other OCs’ combined were well below 1, at 0.25, 0.34, and 0.30 respectively.
However, no statistically significant advantage for DRSP was shown for any of the three comparisons. Including the 5 TIAs (DRSP cohort: 0; LNG cohort: 2; ‘Other OCs’ cohort: 3), the Cox regression yielded upper confidence limits just below 1, at 0.91 and 0.97 respectively for LNG and LNG/’Other OCs’ combined.3.7 All Thromboembolic Events (TE)
Venous and arterial thromboembolic events have different pathophysiological backgrounds. Nevertheless, these two event groups are often combined as thromboembolic events.
In the interest of rendering these results comparable with those from other studies, Cox regression analysis of the combined VTE and ATE results was carried out.
The adjusted hazard ratios for the comparisons DRSP vs. LNG, ‘Other OCs’, and LNG/’Other OCs’ combined were 0.85, 0.69, and 0.76, respectively. Statistically significant results were not found (see Table 3). The same is true when TIAs were included.
A total of 3 TEs had a fatal outcome. All three TEs occurred in the LNG cohort. In two cases the fatal outcome was related to an acute myocardial infarction, and in one case to a VTE.
Conclusions regarding an increased potential risk for LNG users cannot be drawn on account of the small numbers.3.8 Arrhythmia
So-called ‘arrhythmic events’ are rather frequently reported by young women, especially when directly queried about their occurrence. Only 19% (336 out of 1,779 reports) of these subjectively perceived “arrhythmias” were confirmed by ECG. Of the confirmed events, 71% were recurrences of conditions that had occurred before study entry and only 29% represented new conditions.
Out of 99 confirmed new conditions (e.g., hyperthyroidism with tachycardia, ventricular and supraventricular extrasystoles, WPW syndrome, tachycardia absoluta, sick sinus syndrome), 56 received medical treatment (drug therapy or pacemaker implantation).
Arrhythmic events that could be suggestive of an increased serum potassium level (e.g., because of the antimineralocorticoid activity of DRSP) were not observed. Overall, new arrhythmias that require medical treatment were a rare event in all 5 cohorts, with an incidence of 2.8 to 5.1 events/104 WY.
Numerical comparisons of the cohorts with each other show a higher incidence of confirmed events in the DRSP cohort (rate ratio DRSP/LNG 1.27).
However, the frequency of arrhythmic events was already significantly higher at baseline for the DRSP cohort (cf. section 3.1). As such, it is no surprise that differences between the cohorts disappear when the comparison is restricted to new conditions or new conditions requiring treatment. For these categories, in fact, incidences in the 'no use' cohort were at least as high as they were in the DRSP cohort.
The incidence rate ratios for new conditions requiring treatment, namely 0.55 (DRSP/LNG) and 0.87 (DRSP/’Other OCs’), are lower than unity but Cox regression analysis yielded no statistically significant differences between the OC cohorts: DRSP vs. LNG: HR = 0.52 (95% CI: 0.22-1.22); DRSP vs. ‘Other OCs’: HR = 0.77 (95% CI: 0.34-1.76).
4. Discussion
As a class, estrogen/progestin combinations --- OCs as well as hormone replacement preparations --- increase the risk of VTE [14].
This heightened risk has been associated with both estrogens and progestins [1,2,3,4]. Given the serious clinical impacts that VTEs can have, the question of whether a newly introduced progestin is associated with a higher risk of VTE is by no means academic. But the heated scientific debate over whether third-generation progestins are associated with a higher risk of VTE than other progestins has also shown that it is not a simple methodological task to distinguish between confounding, bias, and true heightened risk [15,16,17,18,19].
In many cases, it is difficult to satisfactorily clarify the impact of e.g. preferential prescribing to risk patients, referral bias of risk patients to specialized centers, duration of use, or starter/switcher/long-term user status.
When the EURAS study was designed, therefore, a high priority was placed on the following:
- documenting risk factors, past medical history, family history of VTE, and previous OC use as completely as possible;
- excluding long-term users of the same preparations and differentiating between starter and switcher status;
- documenting all VTEs --- regardless of whether diagnosis and/or treatment were in-patient or out-patient;
- recruiting women prospectively and basing follow-up on direct contact between the investigator team and patients;
- validating all patient-reported VTE (and other outcomes of interest) via the attending physician;
- avoiding erroneous and/or inconsistent classification of VTE by blinded adjudication at the end of the study; and
- ensuring a low loss to follow-up rate by means of a multifaceted, four-level, follow-up procedure.
The low loss to follow-up rate of 2.4% is particularly noteworthy. In theory, a disproportionately high percentage of SAEs could have occurred in those patients who were lost to follow-up, because SAEs could be the reason for the break in contact with the investigators. This could considerably bias the study results.
For all cohort studies (including double-blind, randomized clinical trials), therefore, the loss to follow-up should be kept as low as possible. This is especially difficult for studies of OC users.
A large share of these (young) women participate in such studies during a period of their lives that is marked by frequent major changes in circumstances (e.g., moving to another city as part of the transition from school to university, completing their education, marriage, or change of partner).
An advantage of the EURAS study design, however, is that it addresses precisely this point. Because the investigator team had direct contact with the women, contact was not lost if the women changed their gynecologists, for example (e.g., due to change of residence or dissatisfaction with treatment).
To our knowledge, there is no large-scale OC cohort study described in the literature with a follow-up extending to 5 years that attained lower loss to follow-up rates. In addition, it is important to compare the loss to follow-up rates among the user cohorts.
An asymmetrical loss to follow-up could indicate that the results of one cohort might show a greater degree of bias. However, a comparison of the cohorts showed that differences among them are minor (2.4, 2.7, and 2.2 for the DRSP, LNG, and ‘Other OCs’ cohorts, respectively).
On account of the low and symmetrical loss to follow-up rates in this study, a relevant bias associated with loss to follow-up can be largely excluded. Overall, it may be assumed that the study results are methodologically valid.
The results regarding the primary cardiovascular outcome of interest --- namely, the VTE hazard ratio between users of DRSP-containing OCs and users of LNG-containing OCs --- show that the VTE risks associated with the use of DRSP and LNG are similar.
This is supported without exception by all the sub-analyses (e.g., comparison of preparations with 30 mcg EE, inclusion of potential VTE, starter/switcher status). Furthermore, the small difference between the crude and adjusted HR (< 0.1) is an indication of the robustness of the data presented.
Adjusting for the slightly higher overall VTE baseline risk for DRSP users did not lead to clinically meaningful differences in the results. In fact, if the differences in baseline risk are ignored and the risk analysis is based solely on the crude HR (HRcrude=1.14, 95% CI: 0.66-1.97), a twofold higher VTE risk for DRSP users could still be excluded.
It is noteworthy that at approx. 9 VTE/10,000 WY, the VTE incidence in this study is higher than the 2-4 VTE/10,000 WY given in e.g. the public assessment report of the European Agency for the Evaluation of Medicinal Products Committee for Proprietary Medicinal Products (CPMP) [20].
This report also assumes that the incidence of VTE in non-users is between 0.5 and 1 VTE/10,000. The difference between the incidence found in the EURAS study for OC users and the incidence given in the public assessment report for non-users is thus substantial.
However, a systematic review of the literature (see the article by Heinemann and Dinger in this issue of Contraception) shows that the published figures fluctuate greatly, and that the incidence of VTE in women of reproductive age is likely to be in the range of 5-10 per 104 WY.
When considering these data, secular changes should also be considered. The introduction of duplex sonography and the availability of a sensitive screening method --- the D-dimer test --- have significantly reduced the amount of time and resources needed for diagnostic clarification of unspecific symptoms that could be signs of a DVT, as well as the inconvenience for patients.
Because the clinical symptoms of venous thromboembolism cover the spectrum from a complete absence or unspecific, slight symptoms to dramatic, acute, life-threatening symptoms, it is to be expected that improved diagnostic means will correlate with a higher incidence of VTE diagnoses [21,22,23]. It is therefore of interest to compare the VTE incidence for non-users in this study.
At 2.3 VTE/10,000 WY for non-users who were not pregnant, the incidence is considerably higher than that given in the public assessment report cited above. However, it should be noted that those women who stopped OC use might conceivably have done so due to factors that may influence the incidence of AEs (e.g., diagnosis of risk factors, prophylactic treatment).
Therefore, their results must be considered as tentative.
In order to acquire more reliable data, the investigators carried out an additional study in Germany, the country in which most of the study participants were recruited.
A total of 48,961 women of reproductive age was questioned about their use of OCs, pregnancy, and the occurrence of VTE during the last 12 months. The reported VTEs were validated using the same methodology as in the EURAS study (cf. section 2.4).
The women questioned were taken from a representative sample of the German population. The representative quality of the sample has been substantiated in three publications [24,25,26]. Overall, 15,811 out of 48,961 women were OC users. At 8.9 VTE/10,000 WY (95% CI: 4.8-14.8), the VTE incidence in this group was nearly exactly that found in the EURAS study.
This indicates a high level of comparability between the studies. The VTE incidence was 9.3 VTE/10,000 WY (95% CI: 6.3-13.3) for non-pregnant non-users, and 29.1 VTE/10,000 WY (95% CI: 6.0-82.8) for pregnant non-users. When comparing the data, however, it should be noted that the prevalence of OC use in the younger age groups is higher than that in older age groups.
If the data is adjusted for the age profile of the EURAS population, the incidence for non-pregnant non-users is 4.4 VTE/10,000 WY (95% CI: 2.4-7.3). This shows that VTE incidence for non-users is considerably higher than often assumed. It also suggests that the VTE risk for OC users is roughly twice as high as for non-pregnant non-users.
In relative terms, this difference is generally smaller than previously assumed. The absolute attributable risk, however, is higher.
When evaluating the impact on public health, however, the comparison of OC users with non-pregnant non-users is only of secondary importance. Women who use OCs do so to prevent unwanted pregnancies. The crucial question concerns the risk that these women would face if despite continuing to engage in sexual intercourse, they stopped every form of contraception or switched from OC use to other contraceptive methods.
The VTE rate associated with unwanted pregnancy and abortion is much higher than the VTE rate associated with OC use, and other contraceptive alternatives are less reliable (e.g., condoms) [27,28,29] or associated with a safety profile which is not necessarily better than that of OCs (e.g., IUD are associated with the risk of uterine perforations) [30].Closer examination of the ‘no use’ cohort in the EURAS study shows that nearly half (45.5%) of all SAEs are linked to pregnancy, delivery, and puerperium. As a result, total SAE incidence for ‘no use’ is considerably higher than that for OC use (rate ratio approx. 1.5).
This further supports the frequently postulated positive impact of OCs on public health. Lower SAE incidences for OC users are often explained by a healthy user effect, i.e., that OCs are prescribed primarily for women without (major) health problems. This explanation does not hold in this form for the EURAS study population, because all the women had received an OC prescription at the time of study entry.
Overall, the EURAS study substantiates the safety of OC use for healthy women who need reversible and reliable contraception and who have no risk factors for VTE.
The results for arterial thromboembolic (ATE) events show lower incidences for DRSP-containing OCs than for the 'other OC' cohort. In view of the antimineralocorticoid activity of DRSP, it is conceivable that DRSP has a positive impact on the occurrence of ATE.
However, given the young, predominantly healthy study population and the relatively short exposure, it would be surprising to see a protective effect manifested in this study. Furthermore, because superiority of DRSP over other progestins with respect to ATE was not formulated a priori as a hypothesis, and in view of the multiple analyses, this result can only be classified here as an interesting indication for future studies in populations that have a rather high baseline risk for ATE (e.g., postmenopausal women).
Another important observation concerns the mortality rate in the EURAS cohort: The overall mortality in this study was low (32 deaths during 142,475 WY of observation, corresponding to a rate of 2.2/104 WY).
The Danish health registers [31], which provide very comprehensive and reliable health-related data on the entire Danish population (including pregnancies, deliveries, and abortions), indicate that a general population matched to the age profile of the EURAS population has a mortality of 4.3/104 WY.
The corresponding data for Germany are 3.9/104 WY [32]. The rather low mortality in this study is in line with the well-known low mortality in OC users.
This could be partly explained by the mortality associated with pregnancy, delivery, and abortion in non-users. However, it may be concluded that the EURAS data provide no indication of higher mortality for OC users compared to non-users.
Incidence rates for malignant neoplasms were slightly lower for the DRSP cohort than for the other cohorts.
However, the following must be considered:
- the incidence of malignant neoplasms in a young population of OC users is very low;
- the statistical power of the study is therefore not sufficient to investigate differences with respect to individual types of cancer;
- given the multi-year lag time between tumor induction and tumor diagnosis, the follow-up period is too short to provide any
diagnosis of the majority of tumors potentially induced by the study medication.
However, hormone use could potentially promote the growth of pre-existing tumors. But if so, this study would have been able to reveal a marked promotion of tumor growth triggered by DRSP-containing OCs --- both, in terms of absolute incidence as well as in comparison to other progestins.
The analysis of cancer incidence, therefore, does not suggest that DRSP has a promotional effect on preexisting cancer.
