HbA1c Variability and Cardiovascular Events in Patients with Prostate Cancer Receiving Androgen Deprivation Therapy

Take Home Message In patients with prostate cancer receiving androgen deprivation therapy (ADT), visit-to-visit HbA1c variability (VVHV) increased after ADT initiation and higher VVHV was associated with higher cardiovascular risks. VVHV is a potential tool for cardiovascular risk stratification in these patients.


Introduction
Prostate cancer (PCa) was the third most common cancer globally in 2020, with 1.4 million incident cases and accounting for over 375 000 deaths [1]. Androgen deprivation therapy (ADT) is one of the key therapies for PCa, in which testosterone activity is suppressed pharmacologically and/or surgically [2,3]. ADT is recommended alone or in combination with other therapeutic modalities for diseases of intermediate or higher risk [3,4]. Despite its established oncological efficacy, studies have shown associations between ADT and increased risks of diabetes mellitus (DM), cardiovascular mortality, myocardial infarction (MI), and stroke [5][6][7], and among diabetic patients, worsened diabetic control and a higher risk of diabetic complications [8][9][10]. Currently, risk factors and prognosticators for adverse cardiovascular events among patients with PCa receiving ADT remain actively investigated.
Studies of the glycaemic effects of ADT focused on glycaemic markers, such as HbA1c, as point estimates at fixed time points [8][9][10]. Emerging evidence suggested that visitto-visit HbA1c variability (VVHV) has incremental prognostic value atop point estimates that neglect longitudinal variations in HbA1c levels. Higher VVHV, reflecting more fluctuating HbA1c levels between hospital or clinic visits (ie, less stable glycaemic control), has been associated with increased risks of mortality and adverse cardiovascular events in patients with and without DM [11][12][13][14][15]. Nonetheless, the effect of ADT on VVHV, as well as the prognostic value of VVHV in patients with PCa receiving ADT, is unexplored. Given the adverse cardiometabolic effects of ADT and the prognostic value of VVHV, this study aimed to test the hypothesis that ADT adversely affects VVHV, and that VVHV is prognostic of cardiovascular outcomes in patients with PCa receiving ADT.

Patients and methods
This retrospective cohort study was approved by Joint Chinese University of Hong Kong  All p values were two sided, with p < 0.05 considered statistically significant.

Results
Altogether, 13 537 patients were identified, of whom 2198 had at least three HbA1c results available within 3 yr after ADT initiation. After applying the exclusion criteria, 1065 patients were included (median age 74.4 yr [interquartile range 68.3-79.5 yr]; Fig. 1), of whom 850 (79.8%) had DM. Characteristics of the included patients are summarised in Table 1. Within the 3 yr after ADT initiation, the patients had a median of five (four to six) available HbA1c measurements. The median CV of HbA1c was 0.081 (0.046-0.135), and the median ARV was 0.57% (0.31-1.03%). The 25th percentile, median, and 75th percentile values were also used  for defining the cut-off values for categorising the VVHV markers into quartiles.
A subgroup analysis by prior diagnosis of DM (Supplementary Table 4), prior use of antidiabetic medication(s) (Supplementary Table 5), and type of ADT (Supplementary Table 6) found that there were generally no differences between subgroups in the change in VVHV after ADT initiation, except for the percentage change in CV of HbA1c, which was significantly smaller in those who used antidiabetic medication(s) than in those who did not use such medication(s) at baseline (p = 0.025). There was also a numerical trend for a smaller percentage change in the ARV of HbA1c in those who used antidiabetic medication (s), which approached, but did not reach, statistical significance (p = 0.072).

3.2.
Prognostic value of VVHV  Table 2 and Fig. 3B) of HbA1c had a significantly higher risk of MACEs than those in the lowest quartile. However, among patients who had at least three HbA1c results within the 3 yr prior to ADT initiation, neither perunit changes nor percentage changes in VVHV were significantly associated with the risk of MACEs (Supplementary  Table 7).

Subgroup analysis of the prognostic value of VVHV
In the subgroup analysis by prior diagnosis of DM (Supplementary Table 8 Table 11) showed consistent results, where patients in the highest quartile of both CV and ARV had significantly shorter restricted mean survival time than those in the lowest quartile.

Discussion
In this study, we showed that ADT may increase VVHV and that higher VVHV, but not changes in VVHV, was associated with a higher risk of MACEs among patients with PCa receiving ADT. To the best of the authors' knowledge, this is the first study that explored the effects of ADT on VVHV, as well as the prognostic value of VVHV in the context of ADT.
We found that VVHV increased after ADT, consistent with prior findings of ADT being associated with poor glycaemic control [6][7][8][9][10]. Previous studies about glycaemic control in patients receiving ADT focused on time pointspecific HbA1c or fasting glucose levels, which capture only a snapshot of a patient's glycaemic metabolism and ignore temporal variations in glycaemic indices. VVHV adds a longitudinal element to the assessment of glycaemic control, with higher VVHV indicating lower glycaemic stability. Numerous measures of VVHV exist [12,14,25]. Here, we chose CV and ARV as measures of VVHV. CV is one of the most common measures of VVHV, as its definition (SD divided by mean) inherently considers the effects of mean HbA1c on VVHV. Meanwhile, ARV focuses on differences between consecutive measurements and has been found to be superior to SD in terms of prognostic significance [24]. Originally devised for blood pressure measurements, ARV has been adopted for VVHV [12] as well as for visitto-visit fasting glucose variability [26,27]. We showed that both CV and ARV of HbA1c increased after ADT initiation, providing robust evidence that ADT adversely affects glycaemic stability.
Our finding that higher VVHV was prognostic of MACEs agrees with prior studies of VVHV in other populations [11][12][13][14][15]. Notably, our results indicated that a threshold effect may exist in the relationship between VVHV and the risk of MACEs, as only the highest quartile, but not the second or third quartiles, was associated with an increased risk of MACEs compared with the lowest quartile. Clinically, this may necessitate determination of an upper limit of normal VVHV, rather than aiming to minimise VVHV; further, larger studies are required. Additionally, our subgroup analysis found no significant interaction between VVHV and prior diagnosis of DM, use of antidiabetic medications, or the type of ADT, suggesting that VVHV is prognostic regardless of these factors. Although associations were insignificant in several subgroups, the statistical insignificance was probably due to the small sample sizes. Nevertheless, competing risk regression found significant association only between HbA1c ARV and the risk of MACEs. Kim and colleagues [12] have made similar observations in patients with type 2 DM, possibly indicating that ARV is a more robust measure of VVHV and prognosticator.

4.1.
Clinical relevance and future directions Clinically, our findings reinforced the potential utility of VVHV as a tool for cardiovascular risk stratification. Little has been done in terms of cardiovascular risk stratification for patients with PCa receiving ADT. VVHV may be a simple marker that can be explored for such purposes. More generally, our findings should raise clinicians' awareness of the importance of VVHV and prevent fluctuations in HbA1c from being dismissed as random or measurement errors. Nonetheless, much work remains before VVHV may be adopted for clinical use. Having observed a possible threshold effect, normal values of VVHV need to be established for patients with PCa receiving ADT. Additionally, drivers of VVHV remain unclear: whilst medication adherence may be an intuitive driver, studies have demonstrated associations between VVHV, inflammation, and oxidative stress [28,29]. Given the intimate relationships between inflammation and both cancer and cardiovascular diseases [30,31], the association between VVHV and the risk of MACEs may vary depending on a patient's inflammatory state. Lastly, whilst it is enticing to suggest VVHV to be a treatment target of glycaemic control, it remains unclear how interventions, both pharmacological and nonpharmacological, influence VVHV. Findings from our subgroup analysis suggested that the usage of antidiabetic medication(s) may be associated with smaller changes in VVHV as compared with nonusage. However, it remained unclear whether such an association was independent of confounders, and such findings should be viewed as hypothesis generating only. These areas require further investigation before VVHV may be utilised clinically.

Strengths and limitations
Utilising data from a population-based database, this study included as many patients as pragmatically possible from Hong Kong, increasing the representativeness of our findings. Additionally, we demonstrated robust associations between VVHV and the risk of MACEs in multiple subgroup and sensitivity analyses, reinforcing the validity of our findings. Nonetheless, this study has some limitations. First, many patients had fewer than three HbA1c levels recorded within the 3 yr after ADT initiation, with only 1065 of the 13 537 patients (7.9%) who fulfilled the inclusion criteria analysed, limiting the generalisability of our findings and necessitating larger studies to validate our findings. Furthermore, this study selected patients with high cardiometabolic risks, as they were more likely to receive frequent HbA1c testing than those with low metabolic risks. Therefore, it is unclear whether VVHV would be as prognostic in patients with lower metabolic risks. Similarly, our selection criteria for patients with at least 6 mo of ADT limited generalisability of our findings to patients receiving shorter durations of ADT. Future studies should therefore further explore the effects that shorter courses of ADT may have on VVHV, as well as the prognostic value of VVHV in these patients. Nonetheless, our findings were consistent with prior findings, including those in the general population [25], meaning that VVHV is likely prognostic in patients with lower metabolic risks as well. Additionally, the observational nature of this study predisposed to residual and unmeasured confounders. Specifically, some studies have suggested that gonadotropinreleasing hormone agonists and antagonists may differ in the risk of MACEs [32], although this has remained highly controversial following the publication of the PRONOUNCE trial, the first randomised controlled trial specifically designed to compare the cardiovascular safety of gonadotropin-releasing hormone agonists and antagonists, which found no significant difference in the risk of MACEs between these agents [33]. Moreover, cancer staging, histology, disease risk profile, and individual indications for specific treatment regimens were not available. Neverthe-less, we have considered many important cardiovascular risk factors for multivariable adjustment. Lastly, due to the deidentified nature of the database used (CDARS), the data could not be adjudicated individually, and miscoding of diagnoses and outcomes was possible. Nonetheless, all data inputs were performed by the patients' treating clinicians who were independent of the authors, and none of the authors had the rights to alter recorded data. Previous studies of CDARS have also demonstrated good data completeness and coding accuracy [17].

Conclusions
In patients with PCa receiving ADT, VVHV increased after ADT initiation. Higher VVHV was associated with an increased risk of MACEs, independent of prior diagnosis of DM, use of antidiabetic medication(s), and type of ADT. Further studies are required to validate our findings and to further explore VVHV as a potential tool for cardiovascular risk stratification in patients with PCa receiving ADT.
Author contributions: Gary Tse had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Chan.
Analysis and interpretation of data: Chan, Dee, K. Ng, Tse.
Drafting of the manuscript: Chan.  Data sharing: All data underlying this study are available on reasonable request to the corresponding authors.