Gender Matters: On Recurrence Risk of Cardiovascular Complications for Type 2 Diabetes Subjects

doi:10.21203/rs.3.rs-4902571/v1

Download PDF

Research Article

Gender Matters: On Recurrence Risk of Cardiovascular Complications for Type 2 Diabetes Subjects

https://doi.org/10.21203/rs.3.rs-4902571/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

Previous reports suggest that, despite men facing higher absolute risks of cardiovascular complications, women with type 2 diabetes (T2DM) experience greater relative risks of cardiovascular events. Studies prior to 2000 indicated a significant impact of cardiovascular disease (CVD) in women with diabetes, but recent research shows inconsistent results. The uncertainty persists on whether sex differences in the impact of diabetes on CVD and mortality exist, and if so, whether they depend on age or other cardiovascular risk factors.

Methods

This study analyzed data from Italian Diabetes Outpatient clinics (2005–2018) for more than 300,000 patients. We leveraged sequential data mining techniques such as the incremental risk score (lift score) to investigate whether T2DM is associated with a higher risk of recurrent macrovascular events in women compared to men. We used Cohen's d to compare risk factors in men and women and Chi-square test to compare the number of treatments prescribed.

Results

The incremental risk for women was significantly higher (20% on average) as compared to men for most of the analyzed sequence of major cardiovascular events. A deep analysis of risk factors and therapeutic treatment did not provide a clear explanation for this risk difference .

Conclusions

In this nationwide study, we observed that although men have higher absolute rates of first-occurrence of macro vascular complications, the relative risk of recurrent cardiovascular complications are higher in women across all ages. The study suggests that the treatment and care for cardiovascular complications in diabetic patients should reflect the differences in the risk for subsequent major cardiovascular events in men and women.

Electronic medical records

type 2 Diabetes

Gender inequality

Some previous reports suggest that, although men have higher absolute risks of cardiovascular complications, the relative risks of cardiovascular events are higher in women than in men in presence of a diagnosis of type 2 diabetes [1].

Several studies conducted prior to 2000 suggested a differential and greater impact of cardiovascular disease (CVD) in women as compared to men with diabetes. However, these findings may not reflect the actual picture of the disease risk given the constant changes in the prevalence and the management of diabetes, in particular of type 2 diabetes (T2DM), accounting for 85 to 95% of all diabetes cases [2].

Interestingly, recently conducted studies have found inconsistent results with regards to gender-differences in the relative rates of heart failure, stroke, coronary ischemic disease (CHD), and death from CHD in subjects with diabetes [3]. Thus, uncertainty persists relative to the impact of diabetes on CVD gender-wise, especially with increasing age. Nevertheless, a recent study using a Danish nationwide registry did find a higher absolute rate of CVD in T2DM men as compared with T2DM women, but a higher relative rate of first occurrence of CVD in women as compared to in men across all the ages [4].

It remains uncertain whether gender differences persist in the relative impact of diabetes on CVD and mortality, and whether these gender-differences depend on age is poorly investigated. Lastly, whether these gender-differences on the relative impact of diabetes are also present in recurrent events of cardiovascular complications remain largely unknown.

Most cross-sectional analyses of diabetes data fail to consider the longitudinal nature of the natural history of the disease and its complications and relationships between subsequent events over time.

Using conventional data mining techniques, we modeled patterns of recurrent macrovascular events in subjects with type 2 diabetes (T2DM) and ischemic heart disease, describing their relative frequency and their sequence in women and in men using a contemporary, real-world dataset.

Based on electronic health care data collect from 2005 to 2018 from a network of Italian Diabetes Outpatients clinics, we investigate whether T2DM is associated with a higher risk of a second major adverse macrovascular events in women as compared with men, and if so, whether potential gender-differences in the relative impact of diabetes depend on age or other cardiovascular risk factors. We focused only on macrovascular complications to evaluate those complications apparently sharing the same pathophysiological mechanisms and risk factors.

Data source

In Italy, a continuous improvement effort implemented by a network

of diabetes clinics has been in place since 2006 [6–8]. The initiative,

promoted by the Italian Association of Medical Diabetologists (AMD),

involves a wide network of 300 outpatient diabetes clinics operating within

the national healthcare system, and allows the routine monitoring of a

large set of indicators relative to over 500,000 patients with T2DM.

For this study, the reference population consisted of all patients with T2DM (according to WHO criteria) who were seen between 2005 to 2018.

The dataset was previously used to describe the patterns of patients’ care in Italy [6], to improve type 2 diabetes care [7], and to derive some prediction.

Data pre-processing

The original dataset was submitted to conceptualizing, cleaning, and other data-preparation tasks [5], which are foundational to the application of subsequent artificial intelligence and machine learning approaches, with the ultimate goal of providing predictions to better support diagnostic algorithms and therapeutic choices for individuals with diabetes.

Occurrences classification

For the purposes of the present study, cardiovascular events were classified as follows (see supplementary for ICD9-CM codes):

Coronary Heart Events (CHE) includes ischemic heart disease and angina, previous heart ischemic events, ischemic heart failure or procedures such as coronary angioplasty and coronary heart bypass.

Cerebrovascular events (CVE) included ischemic stroke or hemorrhagic stroke transient ischemic attack, cerebral endovascular revascularization procedures. The definition of ischemic stroke included diagnoses of ischemic stroke and unspecified stroke, as the majority of unspecified strokes are of ischemic origin.

Peripheral artery disease events (PAD) included clinical diagnosis of peripheral artery disease as well as related procedures such as lower extremities angioplasty: or lower extremities bypass.

Population

Electronic Medical Records (EMRs) present a promising opportunity for analyzing large amounts of data, but challenges arise from data quality issues and inherent biases, hindering the extraction of meaningful patterns and potentially exacerbating healthcare disparities. We overcame these problems by selecting a well-defined subsample of subjects which provided reliable data within reasonable confidence. We selected subjects with available data on clinical follow-up between 2005 and 2018, having a record of: Glycated Hemoglobin (HbA1c), weight and blood pressure values (BP). Since our analysis is focused on the incremental risk of a second comorbidity within 3 and 5 years, we selected only subjects that had a clinical follow up of at least 5 consecutive years, with 1 year gap of tolerance. We will refer to these subjects as active patients, which total 320,267. Of these, 320,199 had at least one biochemical record, 307,392 with at least one drug for controlling glycemia and 294,631 with at least one therapy for controlling risk factors and finally, 85,703 with at least one CVD (68,150 with one, 16,731 with two, 822 with three). To only rely on the order of CVD diagnoses that happened within the data collection timeframe, we defined for each patient a first_visit, which is the first data entry with a follow-up record (see above), representing the first time the patient was visited at an AMD outpatient clinic. Any CVD event happened before that was set to first_visit date. As a result, if two CVD happened before a patient visited an AMD outpatient clinic, we would not infer any time order between the two and consider them happened at the same time. We found that at the time of their first_visit, 26,431 patients had one CVD, 2,747 had two and 77 had three.

Statistics

To quantitatively compare the relative increase of a CVD (d2), given a different previous CVD (d1), in male vs female, we used the probability lift ratio (Lift). The Lift is a common quantity used in sequential rule mining, which quantifies the incremental risk of a complication given a previous event. It is defined as:

Where ℙ(d2) is the probability of getting the complication d2, regardless of any previous complication, whereas ℙ(d2|d1) is the probability of getting the complications d2, after or together with having d1. Therefore, if the Lift score is > 1, then the risk for d2 increases after getting d1; if it is < 1 it decreases; if it is = 1 it stays the same, namely the likelihood of getting a second d2 is independent of having the first d1. We rely on this statistic to minimize assumptions on the disease risk model, providing a simple but effective observation.

Data from Electronic Medical Records (EMR) indeed, are hard to model due to incomplete information. We thus limited our analysis to a sequential rule model, due to the inconsistency of the sampling time frequency (i.e., patient visiting regularity the center). Therefore, rather than analyzing the data considering time as a continuous variable, we only considered the order of the events looking at the record only as a sequence of the events (sequential rule mining). We computed an empirical p-value (n = 10000 permutations) to estimate the statistical significance of the Lift changes between men and women. In each permutation, we randomly assigned the gender to each patient, preserving the ratio between men and women, and computed the lift difference.

Estimating number of cardiovascular and glucose lowering prescribed therapies.

To compare the level of treatment given to men and women, we studied the number of different drugs that have been prescribed to patients. In the AMD dataset prescriptions follow the Anatomical Therapeutic Chemical Classification System (ATC). Since ATC may include multiple compounds within the same ATC code, we counted for each prescription the number of different chemical compounds associated with that code.

We decided to count only the first occurrence of the prescription in the records of the patient. Given the sparse and incomplete nature of the EMR, we assumed this would be the best choice, to limit the impact of missing prescriptions in the EMR not because the therapy is not prescribed but because the data has not been collected. Furthermore, as a clinical practice, prescribed treatments for patients with T2DM are often incremental, that is the new therapy is given together with the old ones. This also explains why many times a prescribed therapy is absent from the EMR, because after being prescribed the first time, it is assumed to be confirmed at future visits. Finally, counting only the first occurrence of a prescription is also an estimation of the heterogeneity of the treatment provided to men and women.

Zero Inflated Poisson Model for number of prescribed therapies comparison. To compare the number of therapies prescribed to men and women we counted the number of therapies prescribed to each patient in four different time intervals: since the diagnosis of diabetes to the first occurrence of a macrovascular complication (T0); from the first occurrence of a macrovascular complication to the second event or one year mark, whichever occurred first (d1); from the first to the second occurrence of a macrovascular complication (d1_d2); and from the diagnosis of diabetes to the second occurrence of a macrovascular complication (before d2). The incremental analysis and the time intervals are meant to provide insight on the therapeutic interventions prescribed by clinicians before the first complication, right after the first complication (within 1 year from the event), and up to the second complication, both in terms of incremental therapy and as a total therapy given before the second complication. We then leveraged the Zero Inflated Poisson (ZIP) model to address the significance of the difference in the number of treatments between men and women. The ZIP model allowed us to estimate the impact of covariates to the percentage of zeros and the value of the target variable, when > 0, separately. We modeled the number of prescriptions (target variable) using as covariates the gender, the age of the patient and the number of years since the diabetes onset at the time of the prescription. We used a significance threshold of 0.05.

Ethics

All participating diabetes centers obtained the authorization of local Ethics Committees.

We leverage a dataset of over 300,000 individuals with T2DM diabetes, a subset obtained using criteria to guarantee completeness imposed on the AMD dataset, with an average of 167 recorded characteristics (treatments, prescriptions, clinical tests and diagnosis) per patient. To maximize our confidence in the data collected, we selected individuals having at least a follow-up of 5 consecutive years, 142,655 women and 177,612 men. Population characteristics at the time of T2DM diagnosis and at the first occurrence of a macrovascular event, are shown in Table 1.

Active Patients characteristics in AMD dataset

Table 1

The table shows the mean and standard deviation (*SD)* of the values for all the patients (*Overall), Female* and *Male*. The count column provides the number of observations in the dataset. The values for “Diabetes Duration” represent the number of years between the diabetes diagnosis and the given complication. CVE: Cerebrovascular events; PAD: Peripheral artery disease events; CHE: Coronary Heart Events.
		Active Patients n = 320267	Female n = 142655 (44.5%)	Male n = 177612 (55.5%)	P-Value Female-Male difference
Age at diagnosis of diabetes, mean (SD)		57.5 (11.5)	58.3 (11.8)	56.8 (11.2)	< 0.001
CHE	n (%)	54587 (17.0)	15813 (11.1)	38774 (21.8)	< 0.001
	Age diagnosis, mean (SD)	69.5 (8.9)	71.8 (8.6)	68.5 (8.8)	< 0.001
	Diabetes Duration (years), mean (SD)	12.0 (10.0)	13.7 (10.6)	11.3 (9.6)	< 0.001
PAD	n (%)	42792 (13.4)	14521 (10.2)	28271 (15.9)	< 0.001
	Age diagnosis, mean (SD)	71.4 (8.5)	73.3 (8.6)	70.5 (8.4)	< 0.001
	Diabetes Duration (years), mean (SD)	14.1 (10.1)	15.3 (10.4)	13.4 (9.9)	< 0.001
CVE	n (%)	6699 (2.1)	2425 (1.7)	4274 (2.4)	< 0.001
	Age diagnosis, mean (SD)	71.6 (8.8)	73.4 (8.9)	70.6 (8.6)	< 0.001
	Diabetes Duration (years), mean (SD)	12.9 (10.0)	13.7 (10.3)	12.4 (9.7)	< 0.001

First event CVD complication risk

Men were at higher risk of developing any of the three CVD taken into consideration. Specifically, we observed a higher percentage of men having any of these complications as compared to women, such as 21.8% versus 11.1% for CHD (p < 0.001), 15.9% versus 10.2% for PAD (p < 0.001) and 2.4% versus 1.7% for CVE (p < 0.001). We considered complications both at baseline and during follow-up after the diagnosis of diabetes.

Furthermore, men tended to develop such complications at a younger age (3 years younger) and closer to the T2DM diagnoses (2 years).

Incremental risk in CVD risk as a second complication

Focusing on the risk of multiple complications, we modeled each patient’s health record as a sequence of AMD event codes at irregular intervals. To detect differences between men and women in the likelihood for a second complication, we computed the incremental risk (Lift) for all the possible combinations of complication sequences. As shown in Fig. 1, four of the six combinations (PAD→ CHE; CHE→PAD; CHE→CVE; CVE→CHE) yielded higher risk for women than for men for every age group and for any follow-up period. The two combinations containing CVE and PAD events were higher for men than for women, however, this result is not statistically significant, being the lift difference very small. Interestingly, this result might be a consequence of the significantly lower rate of events of these comorbidities. As described in the methods we computed an empirical p-value the evaluate the significance of these differences (n = 10000 permutations), Supplementary Table 1 shows that all the statistically significant differences yield a higher risk for women than for men. Considering all the combinations and the different number of years of follow up, there is an average increment of the risk in women compared to man of 20%.

Risk factors changes between men and women at the time of the first and second complication

Having observed a gender differential risk for the second CVD complication in diabetic patients, we analyzed possible correlations with known risk factors. Table 2 shows the differences in the percentages of men and women whose parameters are above or below thresholds at the time of CHE as a first complication and CVE or PVD as a second one. The tables for the other complication combinations are listed in the supplementary materials (Supplementary table 2 and 3). At the time of the first cardiovascular event, women were older and had longer diabetes duration; moreover, a significant higher proportion of female subjects showed higher levels of traditional risk factors such as BMI, systolic blood pressure, poor glycemic control (HbA1c), higher levels of LDL-C and renal failure or impaired glomerular filtration rate as compared to the male subjects.

Table 2

Age and duration of diabetes represent the average years (± standard deviation) at the time of the event. n the total number of patients with available information about the risk factor (within one year of occurrence of the described event). Risk factors rows represent the % of patients that are considered at risk, the first column of the table contains the thresholds used. Cohen d and the p-value of the difference was calculated using the variable considered as a continuous value. Average Number of prescribed treatments
	First event					Subsequent event					Subsequent event
	Coronary Heart Events					Peripheral artery disease events					Cerebrovascular events
Risk factor	F (n = 15799)	M (n = 38734)	rate diff.	Cohen d#	P	F (n = 2730)	M (n = 8220)	rate diff.	Cohen d#	p	F (n = 393)	M (n = 1038)	rate diff.	Cohen d#	p
Age at diagnosis	71.82y (± 8.6) (n = 15799)	68.54y (± 8.78) (n = 38734)	-3.28	-0.37	< 0.001	74.31y (± 7.91) (n = 2730)	71.15y (± 7.92) (n = 8220)	-3.15	-0.4	< 0.001	75.0y (± 8.73) (n = 393)	72.17y (± 8.05) (n = 1038)	-2.83	-0.35	< 0.001
Duration of diabetes	13.68y (± 10.58) (n = 15799)	11.25y (± 9.6) (n = 38734)	-2.43	-0.24	< 0.001	17.4y (± 10.93) (n = 2730)	14.58y (± 10.11) (n = 8220)	-2.82	-0.28	0.0	16.24y (± 10.88) (n = 393)	14.43y (± 9.97) (n = 1038)	-1.81	-0.18	< 0.001
SBP > = 140 mmHg	52.65% (n = 15676) [51.9–53.4]	47.49% (n = 38544) [47.0–48.0]	5.16%	-0.13	< 0.001	58.77% (n = 2721) [56.9–60.6]	53.03% (n = 8198) [51.9–54.1]	5.74%	-0.13	< 0.001	60.31% (n = 393) [55.5–65.1]	52.27% (n = 1035) [49.2–55.3]	8.03%	-0.11	0.057
DBP > = 90 mmHg	13.03% (n = 15678) [12.5–13.6]	13.39% (n = 38539) [13.1–13.7]	-0.36%	0.02	0.009	13.74% (n = 2721) [12.5–15.0]	14.88% (n = 8203) [14.1–15.7]	-1.14%	0.01	0.664	16.79% (n = 393) [13.1–20.5]	16.33% (n = 1035) [14.1–18.6]	0.47%	0.04	0.534
BMI > = 30 Kg/m²	44.08% (n = 15112) [43.3–44.9]	36.9% (n = 37415) [36.4–37.4]	7.18%	-0.17	< 0.001	43.5% (n = 2632) [41.6–45.4]	35.36% (n = 7918) [34.3–36.4]	8.14%	-0.17	< 0.001	38.89% (n = 378) [34.0-43.8]	34.06% (n = 1004) [31.1–37.0]	4.83%	-0.15	0.011
HbA1c > 7%	63.28% (n = 15732) [62.5–64.0]	56.31% (n = 38576) [55.8–56.8]	6.97%	-0.14	< 0.001	68.87% (n = 2721) [67.1–70.6]	62.03% (n = 8194) [61.0-6e3.1]	6.84%	-0.16	< 0.001	62.76% (n = 392) [58.0-67.5]	62.67% (n = 1034) [59.7–65.6]	0.09%	-0.04	0.46
eGFR < 60 mL/min/m² (%)	48.07% (n = 8336) [47.0-49.1]	36.11% (n = 18935) [35.4–36.8]	11.96%	0.31	< 0.001	54.49% (n = 1347) [51.8–57.2]	39.72% (n = 3784) [38.2–41.3]	14.77%	0.35	< 0.001	51.44% (n = 208) [44.7–58.2]	43.3% (n = 522) [39.0-47.5]	8.15%	0.18	0.028
LDL-C > = 100 mg/dL	26.96% (n = 4440) [25.7–28.3]	16.81% (n = 12929) [16.2–17.5]	10.14%	-0.29	< 0.001	34.96% (n = 635) [31.3–38.7]	20.82% (n = 2450) [19.2–22.4]	14.14%	-0.33	< 0.001	32.11% (n = 109) [23.3–40.9]	21.81% (n = 321) [17.3–26.3]	10.3%	-0.21	0.059
HDL-C < 40 mg/dL	20.25% (n = 14329) [19.6–20.9]	38.2% (n = 35553) [37.7–38.7]	-17.95%	-0.52	< 0.001	20.02% (n = 2452) [18.4–21.6]	36.32% (n = 7445) [35.2–37.4]	-16.3%	-0.45	< 0.001	21.1% (n = 346) [16.8–25.4]	35.03% (n = 925) [32.0-38.1]	-13.93%	-0.52	< 0.001

Nevertheless, when looking at the second complications, such as CVE, the differences in most of the considered risk factors remained the same or narrowed, such as for example BMI, HbA1c values, glomerular filtration rate; in the case of PVD, as second event, the differences in risk factors increased in the case of diabetes duration, BMI, HbA1c, LDL-C and impaired glomerular filtration rate, but at levels not fully elucidating the increased relative risk.

Treatment changes between men and women at the time of the first and second complication

Finally, we looked at the differential treatment that has been recorded between men and women. We studied patients having CHE→PAD events, the CVD events combination with the highest incremental risk. Table 3 illustrates the average number of treatments both as Cardiovascular therapy (ATC C10-lipid lowering drugs, ATC C02- drugs for blood pressure control, ATC B01-antithrombotic drugs) and as glucose-lowering drugs (ATC A10-drug used in diabetes). While glucose-lowering therapy prescriptions were equally increased in males and females (who started from a higher number of drugs at the beginning, coherently with the worse disease control recorded at baseline), for cardiovascular therapy, the increase seems to be slower, ending with almost the same average number of treatments for men and women. Supplementary table 4 shows the average number of prescribed therapies for all the sequences of CVE events.

Table 3

Average Number of prescribed treatments at different time intervals for patients with CHE→PAD events. T0 →CHE goes from the diagnosis of diabetes to CHE; CHE column is an interval of 1 year after CHE, if an event of PAD occurs before 1 year after CHE then the CHE→PAD interval is considered; CHE→PAD column is the time interval between these two events; T0→PAD goes from the diagnosis of diabetes to the PAD event. Bold numbers are statistically significant differences between men and women in the zero inflated Poisson model, as described in the methods section. The list of Cardiovascular and Glucose-lowering therapies is provided in the supplementary information.
	Sex	T₀ →CHE	CHE	CHE→PAD	T₀ → PAD
Cardiovascular Therapy	F	0.71	0.93	1.35	2.03
Cardiovascular Therapy	M	0.63	1.04	1.45	2.05
Glucose-lowering therapy	F	1.44	1.06	1.78	3.19
Glucose-lowering therapy	M	1.09	1.01	1.68	2.73

The role of gender or sex difference in the development of diabetes complications is debated and a matter of clinical interest, since it may impact the treatment strategies as well as public health policies.

As the Framingham study showed several years ago, diabetic women have a cardiovascular risk 3.5 fold higher than non-diabetic women, while the increase in men is 2.1 fold [11]. However, the baseline cardiovascular risk in men is still higher. Our assessment, using the aforementioned described constrained subcollection from the AMD data, shows that when a first event is recorded, the odds of developing subsequent macrovascular events is consistently higher in women than in men, across almost all age groups, number of years of observation, and specific event sequences.

This is in line with other reports looking at gender differences in subjects with type 2 diabetes and cardiovascular outcomes. Interestingly, a recent meta-analysis systematically comparing adverse cardiovascular outcomes (mainly cardiac and cerebrovascular) in male versus female patients with T2DM following percutaneous coronary intervention (PCI) and including 19,304 patients found that women were more susceptible to short and long-term cardiovascular complications compared to male patients [12].

Our data show that prior to the first cardiovascular event, women have a higher burden of risk factors (e.g., higher cholesterol level, higher BMI). Interestingly, most of the reports comparing women and men with T2DM and established cardiovascular disease in subsequent cardiovascular outcomes, implicated higher baseline comorbidities rate as responsible for higher adverse outcomes in women [13]. In line with these studies, a patient-level pooled analysis of randomized controlled trials showed that among women who underwent PCI with drug eluting stents (DES), chronic kidney disease was associated with a main risk for MACE [14].

In our data, impaired glomerular filtration rate seems to be the main comorbidity worsening in women as compared to men as well as BMI. Women need to attain higher levels of BMI to reach the same levels of visceral and ectopic fat required to become insulin resistant and so to develop diabetes and related complications.

Age, duration of diabetes mellitus, the use of other cardiovascular and oral glucose lowering medications as well as the use of insulin therapy could also have influenced the outcomes between male and female patients with diabetes mellitus. Unfortunately, the medication profile was limitedly reported in these studies.

Women in the AMD cohort tend to get any CVD at an older age. Thus, age could be thought as a confounding risk for the increased risk of a second CVD. However, observing the same lift difference within groups of patients with the same age suggests that the differential gender risk is not influenced by age.

Therapeutic inertia in diabetes refers to the failure of healthcare providers to intensify treatment when patients with diabetes fail to achieve glycemic control or experience complications [15–17]. This lack of treatment escalation can lead to suboptimal outcomes.

Factors contributing to therapeutic inertia include clinical inertia (hesitancy in initiating or intensifying treatment), patient-related factors such as poor adherence to treatment, and healthcare system barriers. Our results suggest highlighting the role of gender disparities in these factors. For example, gender differences in the use of cardioprotective drugs are also implicated as a factor that could adversely affect metabolic and cardiovascular risk factor profiles of women. In our analysis, at baseline, the number of prescribed drugs, either glucose-lowering or cardiovascular agents, seem to be higher in females as compared to males; nevertheless, the number of prescribed cardiovascular drugs was consistently lower for women as compared to men for the following cardiovascular event.

These findings are aligned with data from CVOTs reporting gender disparities in cardiovascular risk management, showing a low percentage of women using cardiovascular protective medications such as ACE inhibitors, statins and aspirin [18].

Another possible cause of women’s additional higher relative risk of second complications is that they could be less aware of their risk of CVD, or they could less likely adhere to treatment once they are at high risk of CVD. An American study found adherence to diabetes medications to be slightly lower amongst women than men [10]. Education and awareness campaigns can help both healthcare providers and patients to better understand the importance of timely treatment escalation.

It would be useful for the future to also understand these aspects. Since most of the evidence-based treatment approval comes from RCT who mostly enrolled men, there is an urgent need to verify if treatments are only tailored to men's clinical needs.

Limitations:

The primary limitation to this study is that our results rely on diagnoses of myocardial infarction, heart failure, ischemic stroke, and PAD reported in the medical charts at the hospitals and not on research data. However, potential minor misclassifications are non-systematic and not gender-biased, and thus, do not influence the overall validity of the analysis.

Furthermore, we cannot rule out that the true cardiovascular events were under or overestimated, which may have biased the results; however, CVD events are in line with those described for the Italian population [19], and generally, our results are concordant with previous studies specifically designed to look at diabetes gender differences on CVD [20]. Secondly, we did not have information on lifestyle and genetic risk factors and adherence to therapy that are potential confounding factors. Lastly, the Italian population is vastly Caucasian, and we did not include immigrating individuals. Thus, our results are not generalizable to non-Caucasian populations.

Strengths

Extracting data from electronic medical records (EMRs) offers a potent research avenue, showcasing notable strengths compared to randomized controlled trials. Such an approach is not new; our team has experience in doing so for over 3 decades [21–23]. The EMR-based data we used, in particular, capitalize on real word patient information, providing us the unique opportunity to view patients' health journeys as compared to other cross-sectional data. This approach allows for the analysis of a large dataset thus enhancing the generalizability of findings.

Unlike the controlled settings of available RCTs or studies specifically designed for specific outcomes, this research reflects the nuances of routine clinical practice, capturing a wealth of information on patient outcomes, treatment patterns and clinician attitudes. However, it is essential to acknowledge limitations such as data quality variations. Nonetheless, the fact that our findings confirm and reflect those obtained in experimental settings make EMR-based studies a valuable complement to the traditional clinical research paradigms.

In this nationwide study, we observed that although men have higher absolute rates of first occurrence of macro vascular complications, the relative risk of recurrent cardiovascular complications are higher in women across all ages. Further studies are warranted to more clearly elucidate the mechanisms responsible for the substantial gender-differences in diabetes-related second occurrence of cardiovascular complications.

Ethics approval and consent to participate

Informed consent was obtained from all subjects and centers involved in this study. The study was conducted in accordance with the Declaration of Helsinki, and all participating diabetes centers obtained the authorization of local Ethics Committees.

Informed Consent Statement according to Italian Law 211/2003, no consent is required for epidemiological analysis regarding anonymous data.

Consent for publication: Not applicable

Availability of data and materials

The data that support the findings of this study are available from AMD and AMD Foundation, but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Code book, and analytic code will be made available upon request pending application and approval of all authors.

Competing interests

All authors have no conflict of interest to declare related to the contents of this study.

Funding: Not applicable

Authors' contributions

M.G., O.F., G.G., M.M. researched and analyzed the data, wrote the first draft of the manuscript, and edited the manuscript. F.C., M.L., R.R., A.N. developed algorithms for data curation. M.G., S.L., F.S. and O.F. Developed the methodology for the analysis. All authors contributed to the discussion and edited the draft. A.N. critically revised the manuscript and is the guarantor of this work and, as such, had full access to all the data in the study.

All authors have read and agreed to the published version of the manuscript.

Acknowledgements

The authors thank all of the centers participating in the AMD Annals Initiative (a complete list is provided in the Supplementary Data)

Angoulvant D, Ducluzeau PH, Renoult-Pierre P, Fauchier G, Herbert J, Semaan C, Bodin A, Bisson A, Fauchier L. Impact of gender on relative rates of cardiovascular events in patients with diabetes. Diabetes Metab. 2021;47(5):101226.
Sun H, Saeedi P, Karuranga S, Pinkepank M, Ogurtsova K, Duncan BB, Stein C, Basit A, Chan JC, Mbanya JC, Pavkov ME. IDF Diabetes Atlas: Global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045. Diabetes Res Clin Pract. 2022;183:109119.
Khan M, Hashim MJ, King JK, Govender RD, Mustafa H, Al J. Al Kaabi. J. Epidemiology of type 2 diabetes–global burden of disease and forecasted trends. J Epidemiol Glob Health. 2020;10(1):107–11.
Sattar N. Type 2 diabetes-related sex differences in cardiovascular risk: reasons, ramifications, and clinical realities. Eur Heart J. 2020;41:1354–6. 10.1093/eurheartj/ehz914.
Maranghi M, Anagnostopoulos A, Cannistraci I, Chatzigiannakis I, Croce F, Di Teodoro G, Gentile M, Grani G, Lenzerini M, Leonardi S, Mastropietro A. AI-based Data Preparation and Data Analytics in Healthcare: The Case of Diabetes. arXiv preprint arXiv:2206.06182. 2022 Jun 13.
Cucinotta D, Nicolucci A, Giandalia A, Lucisano G, Manicardi V, Mannino D, Rossi MC, Russo GT, Di Bartolo P. Temporal trends in intensification of glucose-lowering therapy for type 2 diabetes in Italy: Data from the AMD Annals initiative and their impact on clinical inertia. Diabetes Res Clin Pract. 2021;181:109096.
Russo G, Di Bartolo P, Candido R, Lucisano G, Manicardi V, Giandalia A, Nicolucci A, Rocca A, Rossi MC, Di Cianni G, AMD Annals Study Group. The AMD ANNALS: A continuous initiative for the improvement of type 2 diabetes care. Diabetes Res Clin Pract. 2023;199:110672.
Russo G, Di Bartolo P, Candido R, Lucisano G, Manicardi V, Giandalia A, Nicolucci A, Rocca A, Rossi MC, Di Cianni G, AMD Annals Study Group. The AMD ANNALS: A continuous initiative for the improvement of type 2 diabetes care. Diabetes Res Clin Pract. 2023;199:110672.
Viazzi F, Russo GT, Ceriello A, Fioretto P, Giorda C, De Cosmo S, Pontremoli R. Natural history and risk factors for diabetic kidney disease in patients with T2D: lessons from the AMD-annals. J Nephrol. 2019;32:517–25.
Kirkman MS, Rowan-Martin MT, Levin R, Fonseca VA, Schmittdiel JA, Herman WH, Aubert RE. Determinants of adherence to diabetes medications: findings from a large pharmacy claims database. Diabetes Care. 2015;38(4):604–9.
Franconi F, Campesi I, Occhioni S, Tonolo G. Sex-gender differences in diabetes vascular complications and treatment. Endocrine, Metabolic & Immune Disorders-Drug Targets (Formerly Current Drug Targets-Immune. Endocr Metabolic Disorders). 2012;12(2):179–96.
Bundhun PK, Pursun M, Huang F. Are women with type 2 diabetes mellitus more susceptible to cardiovascular complications following coronary angioplasty? a meta-analysis. BMC Cardiovasc Disord. 2017;17:1–4.
Roffi M, Radovanovic D, Erne P, Urban P, Windecker S, Eberli FR, AMIS Plus Investigator. Gender-related mortality trends among diabetic patients with ST-segment elevation myocardial infarction: insights from a nationwide registry 1997–2010. Eur Heart Journal: Acute Cardiovasc Care. 2013;2(4):342–9.
Baber U, Giustino G, Sartori S, Aquino M, Stefanini GG, Steg PG, Windecker S, Leon MB, Wijns W, Serruys PW, Valgimigli M. Effect of chronic kidney disease in women undergoing percutaneous coronary intervention with drug-eluting stents: a patient-level pooled analysis of randomized controlled trials. JACC: Cardiovasc Interventions. 2016;9(1):28–38.
Cucinotta D, Nicolucci A, Giandalia A, Lucisano G, Manicardi V, Mannino D, Rossi MC, Russo GT, Di Bartolo P. Temporal trends in intensification of glucose-lowering therapy for type 2 diabetes in Italy: Data from the AMD Annals initiative and their impact on clinical inertia. Diabetes Res Clin Pract. 2021;181:109096.
Russo G, Di Bartolo P, Candido R, Lucisano G, Manicardi V, Giandalia A, Nicolucci A, Rocca A, Rossi MC, Di Cianni G, AMD Annals Study Group. The AMD ANNALS: A continuous initiative for the improvement of type 2 diabetes care. Diabetes Res Clin Pract. 2023;199:110672.
Khunti K, Ceriello A, Cos X, De Block C. Achievement of guideline targets for blood pressure, lipid, and glycaemic control in type 2 diabetes: a meta-analysis. Diabetes Res Clin Pract. 2018;137:137–48.
Clemens KK, Woodward M, Neal B, Zinman B. Sex disparities in cardiovascular outcome trials of populations with diabetes: a systematic review and meta-analysis. Diabetes Care. 2020;43(5):1157–63.
Solini A, Penno G, Bonora E, Fondelli C, Orsi E, Arosio M, Trevisan R, Vedovato M, Cignarelli M, Andreozzi F, Nicolucci A. Diverging association of reduced glomerular filtration rate and albuminuria with coronary and noncoronary events in patients with type 2 diabetes: the renal insufficiency and cardiovascular events (RIACE) Italian multicenter study. Diabetes Care. 2012;35(1):143–9.
Penno G, Solini A, Bonora E, Fondelli C, Orsi E, Zerbini G, Trevisan R, Vedovato M, Gruden G, Laviola L, Nicolucci A. Renal Insufficiency And Cardiovascular Events (RIACE) study, group. Gender differences in cardiovascular disease risk factors, treatments and complications in patients with type 2 diabetes: the RIACE Italian multicentre study. J Intern Med. 2013;274(2):176–91.
Yao HR, Cao N, Russell K, Chang DC, Frieder O, Fineman JT. Self-supervised Representation Learning on Electronic Health Records with Graph Kernel Infomax. ACM Trans Comput Healthc. 2024;5(2):1–28.
Allen L, Frieder O. Exploiting database technology in the medical arena. IEEE Eng Med Biol Mag. 1992;11(1):42–9.
Jackson HA, Cashy J, Frieder O, Schaeffer AJ. Data mining derived treatment algorithms from the electronic medical record improve theoretical empirical therapy for outpatient urinary tract infections. J Urol. 2011;186(6):2257–62.
Changes in. risk factors at the time of the first and the second events.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Gender Matters: On Recurrence Risk of Cardiovascular Complications for Type 2 Diabetes Subjects

Status:

Version 1

Abstract

Background

Methods

Results

Conclusions

Figures

Background

Methods

Data source

Data pre-processing

Occurrences classification

Population

Statistics

Ethics

Results

Active Patients characteristics in AMD dataset

First event CVD complication risk

Incremental risk in CVD risk as a second complication

Treatment changes between men and women at the time of the first and second complication

Discussion

Limitations:

Strengths

Conclusions

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1