Treatment cost has become an important factor in health care decision making, especially when there are many medications with similar clinical efficacies to choose from . It is critical that we analyze the cost difference from a real-world perspective to provide economic evidence to support optimal clinical treatment decisions.
The advantages of premixed insulin over long-acting insulin in terms of long-term treatment cost have been demonstrated through an economic modeling methodology . Our study further proves it from a real-world perspective by following two retrospective cohorts of insulin-naïve patients during the one-year period from their first use of insulin until discontinuation in order to study their respective treatment costs. We conducted this analysis with a comprehensive viewpoint from a cost perspective and considered all antidiabetic medications, which better reflects what happens in real-world diagnostics and treatment environments.
Our study results suggest that, in China, long-acting insulin users pay more (mean difference 4.17 RMB) for diabetes-related medication each day than do premixed users. The gap becomes even wider for those on higher insulin dosages. This is likely attributable to the facts that (1) the unit price is higher for long-acting insulin and (2) Chinese long-acting insulin users used more OADs as combination therapies to control postprandial blood sugar. This is consistent with our study’s finding that patients from the long-acting group used more OAD classes during both the three-month baseline period before the index insulin prescription and six months after insulin initiation compared with the premixed group.
Patients on premixed insulin injected higher doses compared with long-acting insulin users. This relates to component differences between the two kinds of insulin. Premixed delivers both long-acting (protaminated portion) and prandial (insulin lispro/insulin aspart) insulin in one injection at mealtimes . Although the daily dose needed to achieve the extended effect (protaminated portion) for premixed users is comparable to that for their long-acting counterparts, its improved HbA1C reduction and better control of postprandial blood sugar contributes to the rapid-acting component of premixed insulin, which accounts for approximately 30% of the total daily dosage .
Our study findings also indicate that patients in the long-acting group used more OAD classes six months after insulin initiation compared with the premixed group, which may partially explain the higher per-day cost and lower insulin dosage for long-acting users; however, due to data limitations, records for patients prescribed OADs in community health care centers could not be used in our study. Therefore, true OAD usage is underestimated.
Qayyum et al.  conducted a systematic review of a series of published comparison studies on long-acting and premixed insulin and found that premixed analogs showed better performance in postprandial blood sugar and HbA1C reduction, whereas long-acting analogs succeeded in reducing fasting plasma glucose levels.
Glycemic control for Chinese patients with T2DM is suboptimal, because insulin therapy is initiated rather late. They tend to have early pancreatic beta-cell dysfunction and early reduction of the insulin secretion peak, leading to more severe postprandial hyperglycemia, which is made even more serious by the traditional Chinese high-carbohydrate diet. Therefore, the lifestyle of Chinese patients could be a major factor contributing to the widespread use of premixed insulin in China .
The calculated health care costs and resource use are underestimated. First, diabetes-related resource use (e.g., outpatient visits, inpatient visits) is defined as occurring when the patient’s first diagnosis is diabetes. However, in real clinical practice, patients with diabetes could still be prescribed an antidiabetic medication even if their first diagnosis is another disease; for example, a patient might be given hypertension as a first diagnosis and diabetes as a second diagnosis in an outpatient setting, leading to the prescription of both anti-diabetic and anti-hypertensive medications. Second, lab test expenses were not available in our database, so the true cost of health care is unknown. Third, the medication costs calculated in this study only include those associated with the study hospitals. Expenses incurred in other hospitals or pharmacies are not included, which may have resulted in an underestimation of the costs.
A broadly recognized challenge with the analysis of real-world data is the need to correct for sample selection bias [14, 15]. Generalized linear models adjusting for the propensity score, including baseline demographics, number of OAD classes, costs, and health care resource utilization, were used to improve baseline comparability between the two groups; however, missing clinical endpoints (such as HbA1C) may compromise the ability to adjust for selection bias.
Given the limitations of the current database, we do not have enough data to be able to compare the two schemes in terms of clinical effectiveness. A future study based on more data would be able to proceed further within the economic evaluation framework, leading to a more comprehensive view of the results of the comparison .