Patients’ baseline and angiographic characteristics for both sexes are presented as percentages and absolute values with regard to categorical variables and compared by Pearson chi-squared test and odds ratios with 95%-confidence intervals. The distribution of continuous variables is characterised by median and quartiles and compared between genders by Wilcoxon rank-sum test. The stent diameter and the number of stents per procedure is summarized by mean and standard deviation. These descriptive statistics are based on the available cases. As patients admitted multiple times cannot be identified in the data base, we considered different interventions to be independent.

The proportion of DES compared to all implanted stents is shown for men and women in categories of relevant factors. The 95%-intervals of odds ratios adjusted standard errors were calculated using the Taylor linearization technique to allow for clustering. The use of DES in categories of age and indication for PCI is visualised in bar charts and tested for interaction by the Breslow-Day test.

In order to adjust the effect of gender on the choice of a drug eluting stent for other determinants, the variables whose distributions differed significantly between men and women on the one hand and DES and BMS on the other hand as well as the significant interaction of age and gender were included in a multivariable logistic model. As multiple stents implanted during the same session strongly tended to be of the same type, generalized estimating equations assuming an exchangeable working correlation structure were applied and robust standard errors calculated for the odds ratios. For explanatory variables with missing information of more than 1%, conditional means, calculated by a regression on age, gender and indication for PCI, were used.

All p-values are the results of two-tailed tests. P-values ≤ 0.05 were considered significant. The statistical calculations have been performed using the SAS system release 9.3 on a personal computer (SAS Institute, Cary, NC, USA).

Women were significantly older than men (71.9 vs. 66.7 years) and were more likely to suffer from diabetes (26.5% vs 20.2%). In contrast, men more often had previous CABG or PCI (13.0% vs. 8.2% and 34.7% vs. 28.2%), while there was no difference regarding renal failure (Table 1).

The presentation with STEMI, NSTEMI or stable CAD as well as cardiogenic shock and with or without signs of heart failure, showed statistically significantly different but numerically similar values between genders. The same holds true for the lesion characteristics, where we found more left anterior descending (LAD) lesions and fewer left circumflex (CX) lesions, stent re-stenosis and complex lesions in women than in men. The centre experience in terms of stent implantations performed per year was comparable for men and women.

Between 1^st quarter 2005 and 4^th quarter 2009, the use of DES increased from 16.0% to 43.9%. After a rapid increase from 2005 to early 2006, the implantation rate reached a plateau and decreased thereafter. Beginning with the 1^st quarter 2008, the rate of DES Implantation steadily increased until the end of the observation period. For all quarters of a year that have been analysed, women received lower rates of DES (p < 0.001; Fig. 1).

Compared to men, women received stents with lower diameters, with a mean ± standard deviation of 2.94 ± 0.48 mm vs. 3.04 ± 0.53 mm (p < 0.0001) than in men (median 3.0 (quartiles 2.5-3.0) mm vs. 3.0 (2.7-3.5) mm, Fig. 2a). This difference between women and men was the same for BMS (2.96 ± 0.48 vs. 3.07 ± 0.53 mm respectively (p < 0.0001) (Fig. 2b), and DES, with a mean of 2.89 ± 0.48 vs. 2.97 ± 0.51 mm (p < 0.0001) (Fig. 2c).

All indications for stent implantation show a lower DES use in women compared to men. While for stable coronary artery disease and NSTEMI there was significant difference of 4.0 and 2.5% (p < 0.001) respectively, the use of DES in STEMI showed a trend (p < 0.086) with an absolute 1% lower use of DES in women (Fig. 3). However, no significant interaction was detected (p = 0.11).

For the women under age 50 years and between 50 and 60 years, there was a trend (p = 0.056 and p = 0.066) to a higher use of DES in women compared to men, with a difference of 2.7 and 2.6%, respectively. In the age group between 60 and 70 years, there is a non significant difference of 0.7%, while women between 70 and 80 years and over 80 years receive fewer DES compared to men, 2.8% (p < 0.001) and 4.6% (p < 0.001) respectively. The age group over 70 years comprises over 42% of the stent implantations that were analysed (Fig. 4). The interaction was highly significant (p < 0.0001).

In both genders, heart failure and renal disease are related to the predominant use of BMS instead of DES (Table 2), while diabetes and previous PCI or CABG were associated with a higher use of DES. Concerning the treated lesions, in-stent restenosis and complex anatomy as well as left main stem or LAD locations correlated with greater DES use.

In the multivariable model, diabetes was a strong predictor of DES use (OR 1.39, p < 0.001), while STEMI (OR 0.34, p < 0.001), cardiogenic shock (OR 0.56, p < 0.001), NSTEMI-ACS (OR 0.61, p < 0.001), stent diameter (OR 0.67 for every mm increase in stent diameter, p < 0.001) and age (OR 0.80 for every 10 years increase in age) and female gender (above the age of 75) were all associated with a lower usage of DES. As a significant interaction between age and gender has been detected (p < 0.0001), the effect estimates of age are reported separately for men and women, and those of female gender for distinct age values (Table 3). When we fitted the regression model separately in the subgroups of stable CAD, NSTE-ACS and STEMI, the adjusted effect of female gender on the use of DES was strongest in patients with stable CAD (OR 0.88 (0.82-0.94) at age 80 years), weaker in NSTE-ACS (OR 0.90 (0.80-0.99)) and insignificant in STEMI (OR 1.00 (0.85-1.18)).

Medical therapy including platelet inhibition and anticoagulation in patients aged 70 years and older did not show a clinically relevant difference between women and men. However, in-hospital mortality and MACCE were significantly higher in women. Most notably, the difference was driven by a significant difference in stable CAD (Table 4). Non-MACCE related access site complications, predominantly bleeding, were more common in women than in men; the difference was significant for all indications (Table 4).

Correlated with a smaller body surface area [15], women have smaller diameter coronary arteries than men, which explains the inferior results in revascularisation procedures, either PCI [5] or CABG [6]. In the ALKK (Arbeitsgemeinschaft Leitende Kardiologische Krankenhausärzte)-PCI registry we found that stents used in women were smaller than those used in men, either for BMS and DES (Fig. 2a, b, c). These data indirectly confirm, that women have smaller epicardial vessels. For both genders, DES were more frequently used in smaller vessels, which reflects the fact that our data were derived from 2005 to 2009, before large data on the use of DES in larger vessels were available [16, 17]. The tendency to use DES in smaller vessels suggests women may receive a predominance of DES compared to men. However, univariate analysis showed that women received a lower percentage of DES compared to men between 2005 and 2009 (Fig. 1). Further analysis in a multivariable logistic model revealed that the lower likelihood for women to receive a DES is observed only in women above the age of 75 year.

The finding of less frequent DES use in women were also evident irrespective of different indications for PCI, like stable angina, NSTEMI and STEMI. However, while the difference in the two former were statistically significant, there was only a trend towards a lower usage of DES in the latter (Fig. 3). The higher frequency of DES implantation in stable disease compared to ACS reflects data progression suggesting superiority of DES even in STEMI-ACS [18]. This benefit of DES use in ACS is confined to reduced repeat target revascularisation, rather than lower mortality [19].

The underuse of DES is an unexpected finding with different possible explanations:

First, the lower rates of DES in older women could be a chance finding. However, the large number of stent implantations and the high significance (p < 0,001) render this explanation unlikely. Furthermore, the adjusted effects show a higher usage of DES in diabetes, whereas ACS and cardiogenic shock were correlated with a lesser use of DES (Table 1). These results are all quite expected and confirm the plausibility of the database.

Another explanation could be an unknown confounder accounting for the findings. Concerns exist regarding DES (and hence dual antiplatelet therapy) use where there is the need for oral anticoagulation, (such as after implantation of a mechanical heart valve or as a result of repeated thrombo-embolic disease). A commonly encountered scenario is dual antiplatelet therapy (DAPT) in addition to oral anticoagulation for patients with atrial fibrillation. The burden of atrial fibrillation is unlikely to explain the disparity in DES use in women over 75 compared to men over 75 however because although information about long term anticoagulation or about atrial fibrillation are not available in our database, the Framingham heart study [20] suggests atrial fibrillation is 1.5 times more common in men over 75 compared to women over 75 so this in itself is unlikely to explain lower DES use in women of this age. Operators may attribute a higher risk of bleeding to elderly women, which is based on objective data on higher peri-procedural bleeding complications [21] and a higher prevalence of anemia [22], that are also predictors of long term mortality [23], as well as a subjective perception of frailty in elderly women.

Therefore, even if peri-procedural bleeding complications do not differ depending on the stent used, the awareness of a higher liability for bleeding could have encouraged interventionalists to rather use BMS instead of DES, whenever there is a suspected risk of bleeding complications, which is typically encountered in elderly women. Interestingly, the tendency to use BMS instead of DES is confined to the type of stent, since anti-platelet therapy and anticoagulation do not differ (Table 4). Especially the use of GPIIb/IIIa inhibitors, which are known for a higher rate of periprocedural bleeding complications [24] (Table 4), was similar in both genders.

Actually, our study (Table 4) confirms a substantially higher periprocedural risk of access site complication, bleeding, MACCE and death in women [15, 25, 26]. Given the higher probability of target lesion revascularisations for in stent restenoses with

BMS [27], elderly women are likely exposed to a higher overall risk due to repeat revascularisation procedures.

Paradoxically, the intention to prevent bleeding complications in women by the use of BMS instead of DES, could actually increase morbidity and mortality.

There could be doubts regarding the efficacy of DES in women, (as women are thought to have less complex coronary lesions [28] which could be treated equally with BMS or DES), particularly as DES are more expensive than BMS. Indeed, Our data shows, a lower percentage of complex lesions and in-stent restenosis in women comared to men. While there is a lack of improvement for mortality and MI, adequately powered RCTs impressively demonstrated a reduction of target vessel revascularisation with DES compared to BMS, which were similar for both sexes [28], even in the elderly [29]. These results were confirmed by registry data on the use of DES, that show that these findings proved to be valid in daily practice [30].

The data from the PCI registry show, that women receive lower percentage of DES compared to men; this difference is significant only for the age groups over 70 years (Fig. 4). However, the age group between 70 and 80 years is the largest in the registry and overall, people older than 70 years account for more than 40% of the sample. Given the high number of PCIs analysed, these data are considered to be clinically relevant, showing that women do not receive best available treatment for coronary artery disease.

The duration of DAPT was not included in our database, however, according to guidelines, patients with STEMI and NSTEMI received DAPT for 12 month, independent of stent type, while patients with stable CAD received DAPT for one or six month for BMS and DES, respectively. Actually, we found the biggest difference in stent usage in the group with stable angina. Given there are more bleedings, this could explain the tendency to prefer BMS in elderly woman in the era of 1st generation DES.

However, latest data for 2nd generation DES show good results for short DAPT, favouring a DAPT for only three month [31] in patients with concerns of bleeding. This shortening of DAPT should further reduce the difference in DES use between men and women.

Our large sample, which is representative for current PCI-procedures in Germany provides data on the use of DES and BMS in women and men in different revascularisation settings. However, our study has some limitations.

First, whilst we adjusted for baseline differences, we cannot fully eliminate them. As a result, it is possible that unmeasured confounders (especially atrial fibrillation and oral anticoagulation therapy requirement) exist and may have contributed to the differences observed. Second, the analysis that was performed is retrospective and is, therefore, dependent on the data already collected.