Article Text
PDF
Abstract
Objective This study aimed to investigate the preferences regarding risks, benefits and other treatment attributes of patients with type 2 diabetes mellitus (T2DM) in China when selecting a second-line anti-hyperglycaemic medicine.
Methods A discrete choice experiment with hypothetical anti-hyperglycaemic medication profiles was performed using a face-to-face survey administered to patients with T2DM. The medication profile was described using seven attributes: treatment efficacy, hypoglycaemia risk, cardiovascular benefits, gastrointestinal (GI) adverse events, weight change, mode of administration and out-of-pocket cost. Participants chose between medication profiles by comparing attributes. Data were analysed using a mixed logit model with marginal willingness to pay (mWTP) and maximum acceptable risk (MAR) calculated. The preference heterogeneity within the sample was explored using a latent class model (LCM).
Results A total of 3327 respondents from five major geographical regions completed the survey. Treatment efficacy, hypoglycaemia risk, cardiovascular benefits and GI adverse events were major concerns among the seven attributes measured. Weight change and mode of administration were of lesser concern. Regarding mWTP, respondents would pay ¥236.1 (US$36.6) for an anti-hyperglycaemic medication with an efficacy of 2.5% points reduction in HbA1c, while they were willing to accept a weight gain of 3 kg only if they received a compensation of ¥56.7 (US$8.8). Respondents were willing to accept a relatively large increase in hypoglycaemia risk (MAR=15.9%) to improve treatment efficacy from intermediate (1.0% points) to high (1.5% points). LCM identified the following four unobserved subgroups: trypanophobia, cardiovascular-benefits-focused, safety-focused and efficacy-focused and cost-sensitive.
Conclusion Patients with T2DM prioritised free out-of-pocket costs, highest efficacy, no hypoglycaemia risk and cardiovascular benefits over weight change and mode of administration. There exists great preference heterogeneity among patients, which should be taken into account in healthcare decision-making processes.
- Health policy
- Health economics
- Diabetes
Data availability statement
Data are available on request. The data sets used and/or analysed during the current study are available from the corresponding author on reasonable request.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
The clinical decision to choose one anti-hyperglycaemic medication over another is complex and should be guided by a patient-centred approach, in which patient preferences should be included.
The preferences for anti-hyperglycaemic medications of Chinese patients with type 2 diabetes mellitus are still unknown.
WHAT THIS STUDY ADDS
Out-of-pocket cost, treatment efficacy, hypoglycaemia risk, cardiovascular benefits and gastrointestinal adverse events were major concerns among the seven attributes measured. Weight change and mode of administration were of lesser concern.
We identified the following four unobserved preferences heterogeneity subgroups: trypanophobia, cardiovascular-benefits-focused, safety-focused and efficacy-focused and cost-sensitive.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
This study provides information on which anti-hyperglycaemic medication characteristics are most important from a patient perspective and shows that patients demonstrate heterogeneity in their medication preferences.
Recognising that patients will make different trade-offs when choosing one treatment over another can improve the healthcare decision-making processes and promote a patient-centred diabetes care.
Introduction
The prevalence rate of diabetes in China increased from less than 1% in the 1980s to more than 12% in 2018.1 2 In 2021, China was reported to have the largest number of patients with diabetes and the second highest healthcare spending on diabetes and its complications worldwide.3–5 Previous research showed lower rates of awareness, treatment and control of diabetes in China compared with the USA.3 6 7 Type 2 diabetes mellitus (T2DM) is a complex chronic illness requiring continuous medical care including multifactorial risk-reduction strategies beyond glycaemic control.
Established guidelines recommend that once initiated, metformin should be continued as long as it is tolerated and not contraindicated.8 However, given the progressive nature of the disease, within 3 years of receiving monotherapy, 50% of patients with T2DM are inadequately controlled and require add-on therapies.9 The American Diabetes Association highlights that the choice of medication added to metformin should be based on the clinical characteristics of the patients and their preferences.8 For patients with established atherosclerotic cardiovascular disease (ASCVD) or indicators of high ASCVD risk, heart failure (HF) or chronic kidney disease (CKD), a sodium-glucose cotransporter-2 inhibitor (SGLT2i) or glucagon-like peptide-1 receptor agonist (GLP-1 RA) with demonstrated cardiovascular disease (CVD) benefits is recommended. For patients without the above indicators, the choice of a second agent to add to metformin is not yet guided by empirical evidence comparing across multiple classes.8 Rather, drug choice is based on efficacy, avoidance of side effects (hypoglycaemia or weight change), cost and patient preferences. For example, if the main driver of the treatment choice is out-of-pocket cost, a sulphonylurea or thiazolidinedione (TZD) might be considered because these drugs are currently reimbursed in China; if the main concern is to avoid hypoglycaemia, GLP-1 RAs, SGLT2i, dipeptidyl peptidase Ⅳ inhibitor (DPP-4i) or TZD might be preferred.10
In accordance with the national guidelines for the prevention and control of diabetes in primary care in China, patient preference is an important consideration in the formulation of individualised treatment plans.11 The US Food and Drug Administration released a guide in 2016 that encourages the use of data on preferences of patients and caregivers in medical product development.12 Discrete-choice experiment (DCE) is an effective approach used since before 1990 to quantify the preferences of patients regarding healthcare and a wide range of related topics.13
Several DCEs have been conducted in the European and American diabetes population.14 However, nationally representative Chinese patient preferences have not been investigated. Conclusions from previous studies conducted outside China may not be generalisable to Chinese patients because of inconsistent results and different cultural contexts between China and high-income countries. Huang et al15 investigated the treatment preferences of patients with T2DM in China but did not consider their medication preferences and only enrolled participants in the hospital setting. However, a large proportion of patients with T2DM are in the primary healthcare setting, such as, those patients who are living in rural areas.2 16
Therefore, the present study used the DCE as a patient decision aid aiming to bridge research gaps by identifying the preferences regarding risks, benefits and other treatment attributes of patients with T2DM in China when selecting a second-line anti-hyperglycaemic medicine. Findings from this study may provide empirical evidence for clinicians or policymakers on how to choose an optimal therapy among multiple suitable alternative medications for patients with T2DM, thus improving medication adherence and promoting a patient-centred diabetes care.
Methods
Patient and public involvement
Patients were involved in the design, conduct, reporting or dissemination plans of our research.
Study sample
In China, there are regional variations in the prevalence of diabetes,16 and different geographical regions have different levels of economic development, which may have an impact on a patient’s medication choices. As such, we used a multistage stratified cluster-sampling procedure that considers geographical region and economic development status for data collection. In stage one, we selected six provinces and municipalities that represent the socioeconomic statuses and lifestyles of five major geographical regions in China. In stage two, we selected a provincial capital city (developed city) and one or two non-provincial capital cities (underdeveloped city) from each geographical region in China. In stage three, we randomly selected one or two hospitals and two or three primary care institutions from each city. Finally, we selected 50 hospitals or primary care institutions from 13 cities in 5 provinces and 1 municipality. Participants were eligible to complete the survey if they were Chinese, ≥18 years of age and diagnosed with T2DM by a healthcare professional. Individuals were asked to participate regardless of their medication history.
The guidelines proposed by Johnson and Orme suggested that the sample size can be calculated using the equation N>500×c/(t×a), where c indicates the number of analysis cells, t refers to the number of choice tasks and a is the number of alternatives. In the main-effects only design, c is equal to the largest number of levels among different attributes in the DCE.17 In the present study, the corresponding values for c, t and a are 6, 6 (exclude the duplicated choice set) and 2, respectively; therefore, N can be estimated as (500×6)/(6×2)=250. To conduct preference heterogeneity analyses, we set the sample size for each geographical region as no less than 250, and the total sample was no less than 1250.
Attributes and levels development
A range of medications in oral and/or injectable form are available to control blood glucose. This variety of treatment options naturally provides a diversity of clinical efficacy, administration mode and adverse event profiles. An initial list of attributes was obtained by conducting a review of the published literature.8 10 14 18 Each initial attribute was then subjected to multiple rounds of prioritisation following three principles: relevance to the research question, relevance to the decision context and whether attributes are related to one another. This procedure yielded 13 attributes. Subsequently, an expert stakeholder panel consisting of five endocrinologists and three administrators in diabetes care was convened to refine the attribute list and identify possible gaps. In this stage, we combined the attributes ‘nausea’, ‘vomit’ and ‘diarrhea’ into ‘gastrointestinal (GI) adverse events’ and defined the cost as ‘monthly out-of-pocket cost’. We then conducted a best-worst scaling (BWS) experiment involving 362 patients with T2DM. This experiment is an effective priority determination method that can provide quantitatively robust evidence for the DCE attribute selection and helped us to refine the 11 attributes (online supplemental table 1) generated from the literature review and the first-stage expert consultation. The BWS results provided a clear view of the priority of the 11 attributes. Details of the BWS results can be found in the study by Liu et al.19 These attributes were further refined through focus-group discussions with six endocrinologists in Shanghai, China.
Supplemental material
Based on the literature review, expert consultation, BWS results and focus-group interview, seven attributes were included in the final survey. The levels of treatment efficacy,8 10 20 21 weight change,10 22 hypoglycaemic risk8 10 23 24 and GI adverse events8 10 25–27 were mainly determined by randomised controlled trials, systematic reviews and clinical guidelines focused on the efficacy and safety of medication treatments for patients with T2DM. We also reviewed the package inserts of anti-hyperglycaemic medicines that were available to Chinese patients. To reduce the cognitive burden of participants, the levels of the ‘cardiovascular benefits’ attribute were qualitatively identified with ‘yes’ and ‘no’ to represent whether the anti-hyperglycaemic medications can provide cardiovascular benefits or not.8 10 The levels of the ‘monthly out-of-pocket cost’ attribute were determined by analysing the health insurance data obtained from one tertiary hospital and experts interview. Considering the Chinese context, ¥0 represented the nearly free anti-hyperglycaemic medications, such as sulphonylureas and biguanides, whereas ¥600 represented the expensive anti-hyperglycaemic medications, such as insulin therapy and some GLP-1 RA (table 1).
Attributes and levels in the discrete-choice experiment survey
Experimental design
The final set of DCE questions included in the survey was generated with a commonly used D-optimal algorithm by SAS to construct a fractional factorial experimental design. The resulting design included 48 unique DCE questions, which were assembled into 8 blocks of 6 DCE questions each. The final experimental design was evaluated for level balance, minimal overlap and orthogonality. The aim of this type of experimental design is to strike a balance between asking enough DCE questions of each respondent to estimate the model parameters for each attribute level while not overburdening the respondent with too many questions. Each respondent was randomly assigned to answer one of the blocks of six questions. Each DCE question offered respondents a choice between two hypothetical medication profiles, and patients were asked which medication they would prefer to choose. Within each block, a strictly dominant option (alternative A is preferred over alternative B) was included to evaluate the internal validity of the data. After the DCE questions, the surveys contained additional questions about demographic information, such as duration of diabetes and current medication use. The design was developed following good research practice guidelines published by the International Society for Pharmacoeconomics and Outcomes Research (ISPOR).28
Data collection
To ensure the quality of data collection, we collected the data in a face-to-face manner. In the tertiary hospitals, the questionnaire was administered to patients by clinicians and the research team. In the primary care institutions, the face-to-face survey was conducted in a conference or waiting room by the research team. The instructions of the DCE and the rationale of the survey were explained in detail by one or two investigators who received specific training from the research team.
Cognitive pre-tests were conducted to ensure the face validity of the survey instrument and comprehensibility of the survey method. We conducted iterative pre-tests on paper (n=5) and, after revisions, conducted additional pre-tests in an online format (n=7). Respondents were provided pre-test questions and were given the option to verbally share comments during the pre-test. The patients completed the questionnaire themselves anonymously after they provided informed written consent. Respondents who had difficulty filling in the questionnaire—for example, those who had poor eyesight—were offered the opportunity to complete the questionnaire verbally. The completion of the questionnaire took about 10–20 min, and each participant was reimbursed for their time. All completed questionnaires were directly returned to the investigators.
Statistical analysis
We fitted the DCE data to a conditional logit model (CLM) and mixed logit model (MIXL) using dummy coding. The optimal model was chosen by examining the Bayesian information criterion (BIC) and Akaike information criterion (AIC). The signs of coefficients of the model indicated whether the corresponding attributes had a positive or negative effect on utility, and the sizes of the coefficients indicated the attributes’ relative importance based on total relative utility. The cost attribute was assumed to be continuous. Thus, the marginal willingness to pay (mWTP) was calculated by assessing the ratio of the preference for other attributes to the preference for out-of-pocket cost.
There is a sensitive trade-off between glucose control and hypoglycaemic risk in patients with T2DM in medication.8 Thus, in our study, in addition to preference weights for attributes, the risk tolerance, that is, the hypoglycaemic risk accepted by patients with T2DM to achieve an incremental glucose control benefit, was quantified using the maximum acceptable risk (MAR) introduced by Johnson et al.29 MAR expresses the risk patients are willing to take to gain a certain amount of benefit. The MAR is theoretically higher if the benefits produced by the intervention increase. Quantitative estimates of MAR can provide useful information for several areas of product development and marketing, such as development strategies for new pharmaceuticals, regulatory approval and risk management.
We conducted a latent class analysis (LCA) to assess the presence of structural differences in the ranking of preferences within the sample (ie, preference heterogeneity). This method tests the presence of unobserved (latent) classes within the sample based on a set of variables that define each observation—in this case, patient preferences for anti-hyperglycaemic medication attributes. Within each class, preference weights were estimated using a CLM.30 A user-written Stata module,31 32 lclogit, was used to conduct this analysis. Demographics for each class were tabulated using by class: tabulate in Stata after model estimation. We fitted the data to one to six classes and used the BIC and AIC to choose the model with the best goodness-of-fit. All statistical analyses were conducted using Stata statistical software.
Results
A total of 3919 respondents met the eligibility criteria and consented to participate. Of those who consented to participate, 3533 completed the survey (90.2%). Some respondents were excluded for failing the internal validity test (n=206), leaving 3327 respondents for the final analysis ; there were no statistically significant differences in demographic characteristics between those who failed versus those who passed the validity test (online supplemental table 2). Among the participants, 50.8% were women, 63.9% lived in urban areas, 50.0% had a body mass index (BMI) ≥24 and 55.8% were more than 60 years old, which is consistent with the estimated prevalence of diabetes among Chinese adults, making the sample broadly nationally representative.2
Approximately 65.2% of respondents had been diagnosed with diabetes for at least 5 years, and 65.1% took at least two medications for diabetes. In addition, 18.0%, 11.8% and 8.6% of respondents were diagnosed with a CVD, CKD and HF, respectively. Most of the respondents indicated having a clear understanding of the harmful effects of diabetes on their health. More than half of the respondents (65.3%) considered that the six DCE tasks were of average difficulty (neither easy nor difficulty) to complete. The number of respondents in each block had a good balance (11.6% to 13.4%).
Preference weight
The AIC and BIC values suggested that the MIXL outperformed the CLM (online supplemental table 3), and the preferences results were not substantially different between the two models. Table 2 shows the MIXL results of the participants who passed the internal validity test (39 924 observations from 3327 patients). A sensitivity analysis involving the 206 participants who failed the internal validity test was conducted (online supplemental table 4), although these changes did not substantially affect the findings. Except for a weight change of −2 kg, all levels within each attribute were statistically different from one another at the 1‰ level, suggesting that the selected attributes were all significant predictors of patient medication choice. The participants particularly cared about treatment efficacy, strongly favouring a 2.5% point reduction in glycated haemoglobin A1c (HbA1c) (β=1.378; p<0.001). They also exhibited strong preferences for a medication without hypoglycaemia risk over one with a 30% risk (β=1.271; p<0.001), and for a medication with cardiovascular benefit over one without (β=1.172; p<0.001). The attribute GI side effects (β=1.138; p<0.001) was in the middle rank and had significant impact.
Mixed logit model for respondent preferences
Other attributes had less influence on respondent choice than did the four described above. The respondent choice was not greatly influenced by administration mode or weight change, although certain levels of each of the two attributes had positive or negative effects on respondent choice. For example, weight gain had a negative effect on the medication preferences of patients with T2DM, but to a lesser extent. The relative preference weight for a weight change of +3 kg was only less than one-third of the 2.5% point reduction in HbA1c. The ‘out-of-pocket cost’ attribute, treated as continuous, was −0.0058. Thus, every 100 additional yuan spent on the anti-hyperglycaemic medication decreased the patient’s utility by 0.58.
MWTP
The mWTP analysis quantified the monetary equivalence that the respondents were willing to sacrifice for the desired level of an anti-hyperglycaemic medication attribute compared with the reference level. mWTP estimates are depicted in figure 1. Respondents were willing to pay ¥215.2 (US$33.4) for an anti-hyperglycaemic medication with an efficacy of 1.5% point reduction in HbA1c over one with an efficacy of 0.5% point reduction in HbA1c. However, respondents were willing to pay only another ¥20.9 (US$ 3.2) to obtain an additional 1.05% point reduction in HbA1c. They were also willing to pay ¥217.7 (US$33.8) for a medication with 0% hypoglycaemia risk over one with a 30% risk, ¥200.7 (US$31.1) for a medication with cardiovascular benefits over one without such benefits and ¥194.9 (US$30.2) for a medication without GI risk over one with GI risk. Respondents were willing to pay ¥−56.7 (US$−8.8) for a weight change of +3 kg over their current weight. In other words, they were willing to accept a weight gain of 3 kg when compensated with ¥56.7.
Marginal willingness to pay. Based on a currency exchange rate of ¥6.449 to US$1.00 in 2021. GI adverse events, gastrointestinal adverse events; HbA1c, glycated haemoglobin A1c; mWTP, marginal willingness to pay.
MAR
The MAR results are illustrated in figure 2. For a specified improvement in treatment efficacy (Δefficacy), MAR is the highest level of hypoglycaemia risk patients would tolerate in return for a reduction in HbA1c offered by the anti-hyperglycaemic medication. Patients were willing to accept a relatively large increase in hypoglycaemia risk (MAR=15.9%) to improve treatment efficacy from intermediate (1.0% point) to high reduction (1.5% point), closely followed (MAR=13.8%) by an improvement from poor (0.5% point) to intermediate reduction (1.0% point); however, they would only accept a small increase in risk (MAR=2.8%) from high (1.5% point) to highest reduction (2.5% point).
Benefit-risk trade-off curve and maximum acceptable risk. HbA1c, glycated haemoglobin A1c.
Preference heterogeneity
The LCA results showed that four classes were the best fit for these data (table 3 and online supplemental table 5). The first class ‘trypanophobia’, which included 359 respondents (10.8%), strongly preferred oral administration. Those 1065 respondents (32.0%) in the second class ‘cost sensitive’ were most sensitive to out-of-pocket costs. The third class ‘cardiovascular-benefits-focused’, which was composed of 739 respondents (22.2%), was most concerned with cardiovascular benefits. The fourth class ‘safety-and efficacy-focused’ was the largest, composed of 1164 respondents (35.0%) who were primarily concerned with treatment efficacy, hypoglycaemia risk and GI adverse events. When assessing demographic characteristics, those without injection therapy experience were more likely to be grouped in class 1 than those with such experience. Respondents from urban areas or with higher education levels were less likely to belong to class 2. Class 3 was more likely to include urban respondents and those with a BMI ≥24. Older inpatients with higher education levels had a higher probability to be classified into class 4 than their counterparts.
Latent class model for preferences heterogeneity (n=3327)
Discussion
To the best of our knowledge, this study is the first to investigate the preferences of patients with T2DM regarding second-line anti-hyperglycaemic medications by performing a DCE with a nationally representative sample in China. For the clinician or decision-maker, our study provides robust evidence on the relative importance to patients of factors that likely influence their treatment decision-making related to medications for T2DM. T2DM is a particularly preference-sensitive condition because of the multitude of available treatments, which have varied benefits, risks and burdens.33 In this study, we estimated the relative importance of preference weight, associated mWTP, MAR and preference heterogeneity. The results showed that out-of-pocket cost, treatment efficacy, hypoglycaemia risk, cardiovascular benefits and GI adverse events were the most important attributes in patient decision-making, whereas administration mode and weight change were the least important. These preferences except for weight gain differed largely by key demographic characteristics.
Our study revealed the importance of out-of-pocket costs in patient preferences. Other studies from countries with universal health coverage, such as Canada and America, also found similar results.34 35 Financial barriers remain a major source of health disparities, and costs should be a focus of treatment goals.36 According to a National Health Interview Survey, one-half of adults with diabetes reported financial stress; among them, most of those who reported not taking medications as prescribed due to cost-related barriers to medication use never shared this with their physician.37 38 Findings from the present study suggest that, in order to reduce the risk of cost-related non-adherence, the costs of care should be strongly considered by physicians in China when developing treatment plans.
Patients weighted the efficacy for a reduction in HbA1c as more important than other risks and benefits measured. These results are in line with a recent systematic review of DCE studies of patient preference in diabetes, where control of blood glucose (reflected in HbA1c) was identified as the most important attribute relating to patient preference.39 In our study, patients were willing to pay ¥215.2 (US$ 33.4) to decrease HbA1c by 1.5% point compared with 0.5% point. However, they were only willing to pay another ¥20.9 (US$ 3.4) to decrease HbA1c from 2.5% point to 1.5% point. It suggested that patients in our study had a clear awareness of the efficacy attribute, as they perceived that higher efficacy was not necessarily better. In clinical practice, when making medication recommendations for patients, physicians must make a careful trade-off between various medicines in terms of glucose control and avoiding hypoglycaemia because when optimal glucose levels are achieved, the risk of short-term complications, such as hypoglycaemia, is expected to increase.40 41 Our study confirmed this again, as we found that treatment efficacy and hypoglycaemia risk had a great impact on patient medication preferences and patients made a sensitive trade-off between the two attributes.
Other preference studies separated the hypoglycaemia attribute into mild and severe, which may increase the patient’s trade-off burden when making a choice.35 42 By contrast, the present study used ‘risk of moderate or mild hypoglycemia events within 6 months’ to define the hypoglycaemia attribute because severe hypoglycaemia did not seem to occur more often with any particular monotherapy or combination therapy.25 Older patients were more likely to be grouped in the safety-focused and efficacy-focused class. The strength of the preference for glucose control reflects the challenges (eg, multiple coexisting chronic illnesses and poor treatment adherence) faced by older patients to achieve such control.
Patients, particularly those with a BMI ≥24, had a high preference for cardiovascular benefits in our study. Although the use of anti-hyperglycaemic medications is generally not associated with increased cardiovascular risk, patients with diabetes, especially those with comorbid obesity, inherently have an increased risk of CVD.43 Clinical trials showed that GLP-1 RAs have beneficial effects on cardiovascular health, but high costs are an important barrier to GLP-RA use. In the present study, the respondents were willing to pay only ¥200.7 (US$ 31.1) to take medication with cardiovascular benefits over one without, which is far below the medical cost of GLP-RA in China.44 Thus, the government should take steps to have more GLP-RA medicines covered by the public health insurance.
Although GLP-RA has demonstrated various benefits for patients, it has the highest GI adverse events risk among anti-hyperglycaemic medications.11 45 46 In the present study, older patients had relatively high concern for GI adverse events. Thus, in clinical practice, when recommending GLP-1 RA for older patients, a careful trade-off should be made between the benefits and potential GI adverse event risks to improve their medication adherence.
Overall, patients preferred the oral versus injection mode of administration. Several previous studies also found similar results.47 48 The preference heterogeneity results indicated that the patients who had never had an injection preferred oral administration over injection therapy. Thus, a lack of understanding of the characteristics of injection therapy may be a potential barrier to insulin use. Another study also suggested that diabetes education can considerably influence the importance assigned to administration mode among patients with T2DM.48
Patient preferences were least influenced by weight change. This is consistent with a recently published double-blind, randomised three-way crossover trial study examining patient preferences for therapy in T2DM.49 Although worry about weight gain is common among patients with T2DM, the present study suggests that Chinese patients were generally not willing to trade efficacy or safety to avoid weight gain.
Limitations
This study has some limitations. First, the attributes and their levels in the DCE survey must be formulated in such a way that respondents clearly and concisely understand their meaning. Although the levels of the treatment efficacy attribute in terms of HbA1c reduction was presented to participants using both qualitative (highest, high, intermediate and poor) and quantitative (2.5%, 1.5%, 1.0% and 0.5%) descriptions in this study, the understanding of HbA1c reduction may be problematic for some patients. Future studies are encouraged to include attributes that can be easily interpreted by the patients, such as, loss of life years.
Second, considering that DCEs are not appropriate when too many attributes are considered, we combined nausea, vomit and diarrhoea into ‘GI adverse events’ to prevent participant fatigue. However, we could not determine the specific preference weight between those factors. In addition, given that the burden of GI adverse events varies largely based on its severity, adopting the risk of overall GI events rather than of severe events could be another limitation of this study.
Third, we did not leave respondents an opt-out option. A forced choice design may cause some biases, but this design was consistent with our study setting. Because of the progressive nature of T2DM, medication therapy is necessary to maintain euglycaemia over time.
Fourth, an inherent limitation of the DCE design is that it requires participants to make hypothetical rather than observed choices. Stated preferences on hypothetical choices may not be reflective of reality, where patients make decisions with emotional, financial and clinical consequences.50 Nevertheless, we attempted to minimise differences between hypothetical and real-world decision-making environments by choosing attributes and levels that are consistent with real clinical decisions. In addition, to ensure the quality of the preferences evidence, we followed recommended guidelines proposed by ISPOR for DCE study design28 and analysis.30 Cognitive pre-tests were also conducted to ensure the face validity of the survey instrument and comprehensibility of the survey method.
Fifth, we only collected the number of medications currently taken by the patients for diabetes. Collecting further details of medications currently taken by patients may provide deep insights regarding their medication preferences, as current medications might impact the patients’ preferences for the hypothetical medicines presented in this DCE.
Finally, there may also be a limitation for the sampling procedure as we did not adopt random sampling. However, to make the preference results more representative, our study used a multistage stratified cluster-sampling procedure that considers geographical regions and economic development status for data collection. In addition, the key demographic characteristics are consistent with the estimated prevalence of diabetes among Chinese adults, making the sample broadly nationally representative.
Conclusions
The decision to choose one medication over another is complex and should be guided by a patient-centred approach. Considerations include the medication effects on cardiovascular and renal comorbidities, efficacy, hypoglycaemia risk, impact on weight, out-of-pocket cost, risk for side effects and patient preferences. This study suggests which attributes are most important from a patient perspective and shows that patients demonstrate heterogeneity regarding medication preferences, the quantitative trade-offs further illustrate the relative importance of these attributes. Hence, the results of this study are informative and robust and can be used to guide healthcare decision-making processes for patients with T2DM.
Data availability statement
Data are available on request. The data sets used and/or analysed during the current study are available from the corresponding author on reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
The ethics approval was obtained from the ethics review board of the School of Public Health at Fudan University (Reference No. IRB# 2021-07-0911), and the research adhered to the tenets of the Declaration of Helsinki. All patients provided their informed consent prior to their inclusion in the study.
Acknowledgments
The authors would like to thank all survey participants.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Handling editor Seye Abimbola
Twitter @Lei_Si_PhD
Contributors SL, YW and YC contributed to the study design. SL, JL, XK, YR, SB, LL, FL, YY, YW and YC were responsible for data collection. SL prepared and analysed the data. SL wrote the first draft of the paper. YW, LS, JL, YR, SB, LL, FL, YY, QP and YC revised the manuscript and provided critically important intellectual content. YC is the guarantor responsible for the overall content. All authors read and approved the final manuscript.
Funding This work was funded by the National Natural Science Foundation of China (Grant No. 72074047).
Competing interests None declared.
Patient and public involvement Patients and/or the public were involved in the design, or conduct, or reporting, or dissemination plans of this research. Refer to the Methods section for further details.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.