Impact of two consecutive rounds of National Volume-Based Procurement (NVBP) for insulin on utilisation and structure: a quasi-experimental study using interrupted time series analysis in Hubei, China

Abstract

Objectives

To examine the impacts of two consecutive rounds of the National Volume-Based Procurement (NVBP) programme for insulin on medication utilisation and structure in Hubei province, China.

Design

Insulin procurement records were extracted from the Hubei Provincial Drug Procurement Trading Platform from May 2020 to December 2025. Descriptive statistics and interrupted time series analysis with segmented regression were employed.

Setting

All medical institutions on the Hubei Provincial Drug Procurement Trading Platform.

Primary outcome measures

Defined daily doses (DDDs), defined daily dose cost (DDDc) and procurement amount.

Results

After two consecutive rounds of NVBP for insulin, average monthly DDDs increased by 2.33 million and 1.80 million DDD (first and second NVBP, respectively), with growth rates of 65.47% and 30.51% (both p<0.01). Interrupted time series analysis showed a stable monthly DDD trend after the first NVBP (trend: 13 230 DDD per month, p>0.05) and a significant increase after the second (trend: 78 280 DDD per month, p<0.01). DDDc declined immediately at both intervention points (¥−4.818 and ¥−0.174, respectively; both p<0.01), with no further monthly trend thereafter. Further subgroup analysis revealed that the growth rate of consumption for insulin analogues was faster than human insulin; the growth rate of consumption for domestic enterprises was faster than imported enterprises. Tertiary, secondary and primary medical institutions all achieved significant increases in procurement volume over two consecutive procurement periods.

Conclusions

This study suggests that two consecutive rounds of the NVBP for insulin have achieved sustained price reductions accompanied by steady growth in overall procurement volume. The policy has significantly improved insulin medication utilisation patterns, sustained optimisation of the market structure and balanced development of the pharmaceutical supply.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Provincial-level insulin procurement data from Hubei, covering nearly 60 million residents and a mixed urban–rural tiered medical system over 68 months.

• Interrupted time series analysis with segmented regression captured both immediate level changes and post-intervention trends across two consecutive National Volume-Based Procurement rounds.

• Outcome indicators were derived from procurement records, which reflect institutional ordering rather than actual patient insulin consumption.

• As a quasi-experimental, single-province design, the interrupted time series analysis cannot fully exclude time-varying confounders.

Introduction

The 11th edition of the International Diabetes Federation Diabetes Atlas highlights a persistent worsening trend in the global diabetes epidemic. By 2024, the number of adults aged 20–79 years with diabetes had reached 589 million, accounting for 11.1% of the total population in this age group. Among them, China has approximately 148 million adults aged 20–79 years with diabetes, accounting for one-quarter of the global total in this age group, making it the country with the largest diabetes population worldwide.¹ Insulin constitutes an indispensable cornerstone of diabetes pharmacotherapy. For people with type 1 diabetes, exogenous insulin is a life-sustaining necessity, with no alternative therapeutic agents capable of fully replacing it. For patients with type 2 diabetes, as pancreatic β-cell function progressively declines, more than half of patients eventually require exogenous insulin to maintain long-term glycaemic control, thereby delaying the development of serious complications such as kidney failure, blindness and cardiovascular disease.² In 2024, global health expenditures related to diabetes exceeded US$1 trillion for the first time, accounting for 8.9% of total global health expenditures³; China’s medical expenditures on diabetes reached US$168.9 billion, with insulin being the most widely used type of antidiabetic medication.⁴ The long-term need for medication to treat diabetes has become a substantial economic burden on global health systems.⁵

To mitigate the escalating medical expenditure for patients with diabetes, the National Healthcare Security Administration officially launched the National Volume-Based Procurement (NVBP) exclusively for insulin in November 2021.⁶ While ensuring a diverse range of clinical treatment options, the programme achieved an average price reduction of up to 48% through a volume-for-price mechanism.⁷ The selected products were officially implemented nationwide in all provinces in May 2022.⁷ Following the expiration of the first insulin NVBP contract, a second consecutive round was promptly initiated in 2024. By optimising bidding rules and stabilising supply expectations, the final winning two times per day price was reduced by an additional 3.8%, resulting in a steady declining price trend⁸ that continued to benefit tens of millions of people with diabetes. Previous research has indicated that Wuhan’s local insulin volume-based procurement pilot only temporarily relieved financial pressure on drug expenditure, but failed to effectively address the long-standing issue of a monopolistic market supply structure dominated by imported insulin suppliers.⁹ The procurement share of imported enterprises actually increased from 88.46% to 92.39% following the local pilot.⁹ Therefore, it remains to be further validated whether subsequent NVBP for insulin will be effective in breaking this market structure and fostering diversified competition on the supply side.

Although a small proportion of patients had not yet perceived a significant reduction in medical expenses attributable to the NVBP policy,¹⁰ a large amount of research indicated that the policy has led to a significant reduction in total healthcare spending on medications such as antihypertensives,¹¹ antidepressants¹² and antiviral drugs for hepatitis B,¹³ thereby further improving the affordability of these medications.¹⁴ Meanwhile, several studies based on prescription data from medical institutions have found that NVBP policy has a significant impact on the structure of healthcare expenditures for certain diseases.^{15 16} However, the studies reported above have primarily focused on chemical drugs, while there are relatively few reports on biologics such as insulin. Compared with chemical drugs, which are highly standardised, feature short production capacity expansion cycles and low marginal production costs, biologics are characterised by longer production capacity expansion cycles and far less flexible production and cold supply chains than chemical drugs.^{17 18} Therefore, statistical analyses of biological drug usage should not be limited to short-term cross-sectional data; accurate assessment of policy effects requires long-term, multidimensional analysis.

To date, research on the effectiveness of the NVBP policy for insulin remains extremely scarce; published studies fall into only two categories: one consists of small-sample descriptive statistical analyses based on procurement data from a single medical institution^{19 20}, while the other comprises short-term policy evaluations using interrupted time series analysis based on data from provincial procurement platforms.²¹ However, these published studies all have certain limitations. On the one hand, marked regional disparities exist in insulin utilisation patterns,²² and single-institution, small-sample studies yield findings with limited generalisability. On the other hand, the sole available provincial-level study only covers the short 2-year period following policy implementation. This fails to cover the second consecutive round of insulin NVBP and is thus unable to track long-term changes after policy implementation. Against this background, this study aimed to systematically evaluate the short-term effects and long-term trends of NVBP for insulin. Using provincial-level procurement data covering medical institutions of various tiers and regions, combined with an extended observation period of 68 months spanning multiple procurement periods, we provide a comprehensive and robust evaluation of policy effects.

Methods

Study design

This study employed descriptive statistical analysis and interrupted time series (ITS) to examine changes before and after two consecutive rounds of NVBP for insulin. The ITS is a quasi-experimental method that is applied at the population level to study the longitudinal effects of the evaluation of policy interventions.²³ In late May 2022, the first NVBP for insulin was implemented in Hubei province, China, marking the first intervention point²⁴; in late August 2024, the second NVBP for insulin was implemented in Hubei province, marking the second intervention point.²⁵ Therefore, this study employs an ITS analysis method with two intervention points to evaluate the short-term effects and long-term trends behind the implementation of the policy.

Setting and data sources

This study was conducted in Hubei province. Wuhan, the provincial capital, was the first city in China to pilot insulin volume-based procurement.⁹ Following the NVBP, the Hubei Provincial Drug Procurement Trading Platform continued to provide province-wide data, enabling this extended evaluation. Hubei’s mixed urban and rural hierarchic medical system and large population (nearly 60 million residents) also make it broadly representative of central China.

The data for this study was obtained from the Hubei Provincial Drug Procurement Trading Platform, through which all drug procurement transactions by medical institutions, pharmaceutical companies and distribution companies across the province are conducted. This study included monthly insulin procurement transaction records from May 2020 to December 2025. The transaction records contained key variables such as procurement date, drug name, drug price, dosage form, package size, manufacturer, procurement volume, procurement amount, name of the medical institution and medical institution grade. The period from May 2020 to May 2022 constitutes the pre-implementation study series; the period from June 2022 to August 2024 constitutes the first-implementation study series; and the period from September 2024 to December 2025 constitutes the second-implementation study series. This study did not require ethical approval because it did not employ or present any human data that could potentially be identified.

Data collation and outcome indicators

To ensure the quality of the data, the two researchers (JT and SL) separately classified and logically proofread the extracted raw data. All raw procurement records were categorised by insulin type (human insulin, insulin analogue), enterprise ownership (domestic, imported) and institution grade (tertiary, secondary, primary) and compiled into an analytical dataset. To identify potential institutional stockpiling, we defined an outlier as any month in which total DDDs exceeded the period-specific median by more than three times the IQR. No month met this criterion across the three procurement periods (0 of 68 observations flagged). All monthly records were therefore retained. The collated data are shown in online supplemental table S1.

The primary outcome indicators established for this study are defined daily doses (DDDs), defined daily dose cost (DDDc) and procurement amount. According to the WHO WHO Anatomical Therapeutic Chemical (ATC)/DDD index, the DDD for all insulins is set at 40 IU.²⁶ DDDs reflect the intensity of a drug’s clinical usage, while the DDDc represents the average daily cost burden on patients using the drug. These two metrics are core standardised indicators used in drug utilisation studies. The calculation formulas are as follows:

DDDs=purchase quantity×unit potency×number of packages/DDD

DDDc=total purchase amount/DDDs

Statistical analysis

The outcome indicators for each procurement period are expressed as monthly mean values. For data that follow a normal distribution, comparisons were performed using a t-test²⁷; for other data, comparisons were performed using the Mann-Whitney test.²⁸

Segmented regression, the statistical method used to implement the ITS design, was applied to estimate the level and trend changes of each outcome variable after the NVBP policy. The following models have been adopted:²⁹

$${\displaystyle {\displaystyle \begin{array}{rl}{Y}_t=& {\beta }_0+{\beta }_1\times tim{e}_t+{\beta }_2\times interventio{n}_1+{\beta }_3\times afterinterventio{n}_1\\ & +{\beta }_4\times interventio{n}_2+{\beta }_5\times afterinterventio{n}_2+{\epsilon }_t\end{array}}}$$

$Y_t$ is the outcome variable (DDDs, DDDc, purchase amount). ${\beta }_0$ estimates the baseline level of the outcome variable at the start of observation; ${\beta }_1$ is the slope, representing the trend prior to the intervention; ${\beta }_2$ represents the instantaneous change in the outcome variable at the first intervention time point; $({\beta }_1+{\beta }_3)$ represents the segmented trend (slope) following the first intervention; ${\beta }_4$ represents the instantaneous change in the outcome variable at the second intervention time point; ${(\beta }_1+{\beta }_3+{\beta }_5)$ represents the segmented trend value (slope) after the second intervention; ε denotes the error term, capturing unobserved time-varying confounding factors not explained by covariates. In the baseline model, we assume ε follows independent and identically distributed white noise, and we do not impose any parametric autoregressive structure on ε. We performed the Cumby-Huizinga test to diagnose potential residual serial autocorrelation in each model.³⁰ If statistically significant autocorrelation was identified, we applied the Newey-West estimator to obtain Heteroskedasticity and Autocorrelation Consistent (HAC) SEs.³¹ A sensitivity analysis was conducted by setting the second intervention point to June 2024 (the month of the national NVBP announcement), instead of August 2024 (the month of Hubei provincial rollout), to test whether the 2-month delay affected the results. All analyses were conducted on Stata V.18.0.

Patient and public involvement

No patients or members of the public were involved in the design, conduct, reporting or dissemination plans of this study. This is a retrospective policy evaluation study using available administrative data, and patient involvement was not required for the research objectives.

Results

This study divided the sample data into three distinct procurement periods: the pre-NVBP period, the first-NVBP period and the second-NVBP period. It had examined overall differences across these periods through between-group comparisons and employed ITS analysis to accurately identify both the short-term impact and long-term trends of the policy. Table 1 presents the monthly mean DDDs and DDDc for insulin and its subgroups across the three periods. DDDs increased significantly across all categories, while DDDc declined substantially, with domestic enterprises and primary institutions showing the largest relative changes (all p<0.01). Procurement amount for total insulin declined by 15.24% after the first NVBP and increased by 27.93% after the second NVBP. Detailed results are provided in online supplemental file S1 and online supplemental figure S1.

Table 1.

Comparative analysis before and after the two consecutive rounds of NVBP for insulin

Subgroups	Items	Pre-NVBP	First-NVBP	Second-NVBP	GR (%)		P value
					Pre/first	First/second	Pre/first	First/second
Defined daily doses (thousand DDD)
Insulin	Total insulin	3560.58	5891.61	7689.36	65.47%	30.51%	<0.01*	<0.01*
Generations	Human insulin	1523.64	2489.87	2693.62	63.42%	8.18%	<0.01*	>0.05
	Insulin analogue	2036.93	3401.74	4995.74	67.00%	46.86%	<0.01*	<0.01*
Enterprises	Domestic insulin	1068.64	2597.52	4093.05	143.07%	57.58%	<0.01*	<0.01*
	Imported insulin	2491.94	3294.09	3596.31	32.19%	9.17%	<0.01*	<0.05*
Grades	Tertiary	2628.41	3481.35	4531.76	32.45%	30.17%	<0.01*	<0.01*
	Secondary	687.42	1061.71	1388.58	54.45%	30.79%	<0.01*	<0.01*
	Primary	255.75	1340.38	1737.73	424.10%	29.64%	<0.01*	<0.01*
Defined daily dose cost/CNY
Insulin	Total insulin	10.135	5.187	5.080	−48.82%	−2.06%	<0.01*	<0.01*
Generations	Human insulin	6.317	3.661	3.436	−42.05%	−6.15%	<0.01*	<0.01*
	Insulin analogue	12.945	6.310	5.970	−51.26%	−5.39%	<0.01*	<0.01*
Enterprises	Domestic insulin	8.005	4.564	4.386	−42.99%	−3.90%	<0.01*	<0.01*
	Imported insulin	11.026	5.675	5.880	−48.53%	3.61%	<0.01*	<0.01*
Grades	Tertiary	10.549	5.393	5.266	−48.88%	−2.35%	<0.01*	<0.01*
	Secondary	9.296	5.111	4.995	−45.02%	−2.27%	<0.01*	<0.01*
	Primary	9.219	4.712	4.671	−48.89%	−0.87%	<0.01*	>0.05

*A statistically significant difference.

CNY, Chinese yuan; GR, growth rate; NVBP, National Volume-Based Procurement.

Comparative analysis of DDDs

After the implementation of two consecutive rounds of NVBP for insulin, the average monthly DDDs for insulin and its subgroups increased to various degrees. Among these: (1) During both the first-NVBP and the second-NVBP periods, the DDDs for insulin increased by 2.33 million DDD and 1.80 million DDD, respectively, corresponding to growth rates of 65.47% (p<0.01) and 30.51% (p<0.01). (2) During the first-NVBP period, the DDDs for human insulin and insulin analogues increased by 966 000 DDD and 1365 000 DDD, respectively, representing growth rates of 63.42% (p<0.01) and 67.00% (p<0.01). However, during the second-NVBP period, the DDDs for insulin analogues continued to grow at a high rate, reaching 46.86% (p<0.01); the DDDs for human insulin showed only a slight increase (8.18%, p>0.05), which was not statistically significant. (3) During both the first-NVBP and the second-NVBP periods, the DDDs for domestic and imported enterprises showed varied degrees of growth; however, the increase in DDDs for domestic enterprises (143.07%, p<0.01; 57.58%, p<0.01) was greater than that of imported enterprises (32.19%, p<0.01; 9.17%, p<0.05). (4) During both the first-NVBP and the second-NVBP periods, the DDDs at medical institutions of all grades experienced significant increases; in particular, primary medical institutions achieved over a fourfold increase during the first-NVBP period (p<0.01). The results are presented in table 1.

Comparative analysis of DDDc

During both the first-NVBP and the second-NVBP periods, the DDDc for all groups decreased to varying degrees, with the exception of the DDDc for imported insulin during the second-NVBP period. Among these: (1) During the first-NVBP period, the DDDc for all insulin products and their respective subgroups decreased significantly (−42.05% to −51.26%, all p<0.01), with the largest drop in insulin analogues (−51.26%) and the smallest in human insulin (−42.05%). (2) During the second-NVBP period, DDDc for insulin, human insulin, insulin analogues, domestic insulin and at tertiary and secondary medical institutions all showed slight decreases (−2.06% to −6.15%, p<0.01); the DDDc for primary medical institutions also decreased slightly (−0.87%, p>0.05), but this decrease was not statistically significant. (3) In contrast to the further decline observed among domestic enterprises, the DDDc for imported enterprises actually increased during the second-NVBP period (3.61%, p<0.01). The results are presented in table 1.

ITS analysis of DDDs

The two consecutive rounds of NVBP for insulin had significant impacts on insulin DDDs, as demonstrated by the following: (1) During the first-NVBP period, insulin DDDs immediately increased by 1.490 million DDD (p<0.01), but monthly DDDs did not show a sustained upward trend (p>0.05); during the second-NVBP period, insulin DDDs immediately increased by 1.025 million DDD (p<0.05) and maintained a monthly increase of 78 000 DDD (p<0.01). (2) During the first-NVBP period, DDDs for human insulin and insulin analogues increased immediately by 807 000 DDD (p<0.01) and 683 000 DDD (p<0.01), respectively; while monthly DDDs for human insulin showed a continuous downward trend (−16 000 DDD, p<0.05), monthly DDDs for insulin analogues exhibited a continuous upward trend (29 000 DDD, p<0.01). During the second-NVBP period, the DDDs for human insulin and insulin analogues increased immediately by 392 000 DDD (p<0.05) and 633 000 DDD (p<0.05), respectively; after the second NVBP, the monthly DDDs of human insulin did not increase significantly (p>0.05), whereas those of insulin analogues continued to rise (p<0.01). (3) For imported insulin, the ITS model estimated no significant monthly trend change after the first NVBP (p>0.05), and no significant immediate change (p>0.05) or monthly trend change (p>0.05) after the second NVBP. For domestic insulin, by contrast, the monthly trend rose significantly after the first NVBP (+15 100 DDD per month, p<0.05), and both the immediate increase (+571 000 DDD, p<0.01) and subsequent monthly trend (+95 100 DDD per month, p<0.01) were significant after the second NVBP. (4) Insulin DDDs in tertiary (p<0.01), secondary (p<0.01) and primary (p<0.05) medical institutions all showed an immediate and significant increase during the first-NVBP period. In contrast to insulin DDDs in tertiary and secondary medical institutions showed an upward trend on immediate changes (p<0.05; p<0.01) and monthly trend changes (p<0.01; p<0.05) during the second-NVBP period; primary medical institutions exhibited a sustained growth trend during the first-NVBP period (p<0.05), but this growth trend did not continue during the second-NVBP period. The ITS analysis diagrams are shown in figures 1 and 2 and the detailed model parameters are shown in table 2.

Figure 1.

ITS analysis of monthly DDDs. Human insulin DDDs (A), insulin analogue DDDs (B), domestic insulin DDDs (C), imported insulin DDDs (D). DDD, defined daily doses; ITS, interrupted time series.

Figure 2.

ITS analysis of monthly DDDs. Total DDDs (E), tertiary DDDs (F), secondary DDDs (G), primary DDDs (H). DDD, defined daily doses; ITS, interrupted time series.

Table 2.

Model validation and parameter estimation for DDDs of NVBP for insulin (thousand DDD)

Items	Intercept (β0)		Time1 (β1)		Level change at first-NVBP (β2)		Trend change after first-NVBP (β1+β3)		Level change at second-NVBP (β4)		Trend change after second-NVBP (β1+β3+β5)
	Es β0	P value	Es β1	P value	Es β2	P value	Es (β1+β3)	P value	Es β4	P value	Es (β1+β3+β5)	P value
Total insulin	2943.01	<0.01∗	51.46	<0.01∗	1489.95	<0.01∗	13.23	>0.05	1025.36	<0.05∗	78.28	<0.01∗
Human insulin	1190.47	<0.01∗	27.76	<0.01∗	807.24	<0.01∗	−15.53	<0.05∗	392.24	<0.05∗	3.86	>0.05
Insulin analogue	1752.54	<0.01∗	23.70	<0.01∗	682.72	<0.01∗	28.77	<0.01∗	633.12	<0.05∗	74.41	<0.01∗
Domestic insulin	589.44	<0.01∗	39.93	<0.01∗	813.29	<0.01∗	15.11	<0.05∗	570.98	<0.01∗	95.06	<0.01∗
Imported insulin	2353.56	<0.01∗	11.53	>0.05	676.67	<0.01∗	−1.88	>0.05	454.38	>0.05	−16.78	>0.05
Tertiary	2488.86	<0.01∗	11.63	>0.05	679.21	<0.01∗	1.74	>0.05	699.04	<0.05∗	43.61	<0.01∗
Secondary	677.28	<0.01∗	0.10	>0.05	342.54	<0.01∗	3.04	>0.05	198.93	<0.01∗	11.38	<0.05∗
Primary	−223.13	>0.05	39.74	<0.01∗	447.92	<0.05∗	9.39	<0.05∗	160.47	>0.05	14.05	>0.05

Parameter estimates from the segmented regression of the ITS analysis, with monthly DDDs (thousand DDD) as the dependent variable.

^*A statistically significant difference.

β₃ and β₅ are the interaction terms between time and each NVBP indicator. (β₁+β₃) is the trend slope after the first NVBP, and (β₁+β₃+β₅) is the trend slope after the second NVBP.

DDD, defined daily doses; Es, estimated; ITS, interrupted time series; NVBP, National Volume-Based Procurement.

A sensitivity analysis that set the second intervention to June 2024 (the national announcement month) found no significant immediate level changes for either total DDDs or total DDDc. This contrasts with the main model, which used August 2024 (the Hubei provincial implementation) and showed significant immediate responses. This pattern supports our choice of the provincial rollout date as the effective intervention point. Detailed results are shown in online supplemental file 2.

ITS analysis of DDDc

The two consecutive rounds of NVBP for insulin had significant impacts on the immediate change in DDDc of insulin, but had limited effects on the monthly change in DDDc, as demonstrated by the following: (1) The ITS model estimated significant immediate declines in total insulin DDDc at both NVBP points (first: ¥−4.818, p<0.01; second: ¥−0.174, p<0.01), but the post-intervention monthly trends were not significant (both p>0.05). (2) During both the first-NVBP and the second-NVBP periods, the DDDc decreased significantly in the implementation month; compared with the relatively stable DDDc of insulin analogues, the monthly DDDc of human insulin declined continuously after both NVBP rounds (first: ¥−0.006 per month; second: ¥−0.005 per month; both p<0.01). (3) During the first-NVBP period, the DDDc for both domestic and imported insulin enterprises immediately declined (domestic: ¥−3.515; imported: ¥−5.502; both p<0.01); unlike the DDDc of imported enterprises which remained relatively stable, the monthly DDDc of domestic enterprises showed a sustained upward trend (¥0.010, p<0.05). During the second-NVBP period, the DDDc of domestic enterprises immediately declined again (¥−0.278, p<0.01) and maintained a continuous downward trend (¥−0.005, p<0.01); however, the DDDc of imported enterprises did not change significantly (¥−0.056, p>0.05) and showed a continuous upward trend (¥0.032, p<0.01). (4) During the first NVBP, DDDc declined immediately across all institution grades (tertiary: ¥−5.360; secondary: ¥−4.154; primary: ¥−3.498; all p<0.01); in contrast to the stable DDDc levels observed in tertiary and secondary medical institutions, the DDDc in primary medical institutions actually showed an upward trend (¥0.023, p<0.01). During the second-NVBP period, there was no significant change in the DDDc of tertiary medical institutions; DDDc in secondary medical institutions decreased immediately (¥−0.201, p<0.01), but the monthly DDDc (p>0.05) showed no significant change; DDDc in primary medical institutions decreased immediately (¥−0.467, p<0.01), but the monthly DDDc (¥0.015, p<0.01) showed an upward trend. The ITS analysis diagrams are shown in online supplemental figure S2 and the detailed model parameters are shown in online supplemental table S2.

Discussion

This study found that following the first-NVBP for insulin, procurement volume for insulin increased sharply in the implementation month and then stabilised. These findings are consistent with those of other domestic studies,²¹ further confirming that NVBP policy effectively stimulates patient demand for medications by significantly reducing drug prices. The mechanism of ‘volume for price’ has also been widely validated in international studies: research by US scholars has shown that reducing out-of-pocket costs for patients can improve adherence to insulin therapy and increase insulin usage³²; data from several Asian countries indicate that insulin usage is highly correlated with drug prices.³³ After the implementation of the second NVBP for insulin, drug prices declined slightly and procurement volume of insulin increased again and maintained a monthly growth trend. This result differs from the Wuhan pilot, which achieved a comparable average price reduction (4.53%) through a single-round, city-level negotiation using stepwise volume-retention rules rather than competitive bidding, but did not increase total insulin procurement.⁹ In contrast, the NVBP adopted competitive bidding with nationwide coverage, and its second round built on the demand momentum created by the 48% price reduction in the first round—a structural advantage absent from the Wuhan pilot. We propose the possible reasons are: (1) The significant price reductions resulting from first NVBP for insulin laid a solid foundation for the sustained and widespread use of insulin in terms of both price¹⁸ and reputation¹⁰; (2) the continuously rising number of people with diabetes,³⁴ combined with government requirements to improve diabetes recognition and control rates,³⁵ has continuously driven demand for insulin; (3) the NVBP policy has significantly improved drug accessibility, expanding insulin access to patients in previously underserved areas.³⁶ Therefore, we recommend that the NVBP policy for insulin be further consolidated and sustained.

This study found that during the first-NVBP period, the monthly procurement volume of both insulin analogues and human insulin increased significantly; however, during the second-NVBP period, only the insulin analogues continued to rise significantly, while that of human insulin remained relatively stable. In other words, the NVBP policy led to a significantly faster growth rate in the procurement volume of insulin analogues compared with human insulin. The findings of this study are consistent with those of other domestic and international studies, which indicate that price is a key factor affecting the structure of insulin prescriptions.^{21 37} In addition to price, we attribute this structural shift to several other factors: (1) A global shift toward more advanced treatments, as authoritative clinical guidelines both domestically and internationally recognise that insulin analogues are superior to human insulin in reducing the risk of hypoglycaemia^{34 38}; (2) evolving patient expectations, characterised by an increased emphasis on high-quality treatment regimens and drug side effects³⁹; (3) advancements in industrial capabilities, with China becoming one of the few countries globally capable of large-scale production of biosimilars.⁴⁰

This study also found that during the first-NVBP period, the monthly procurement volume of insulin from both domestic and imported enterprises increased significantly. During the second-NVBP period, the monthly procurement volume of insulin from both domestic and imported enterprises continued to rise, but the growth rate of domestic enterprises was significantly faster than imported enterprises. Domestic enterprises accounted for 44% of total volume after the first round of NVBP and further increased their share to 53% following the second round. This growth reflects both strategic policy guidance and accumulated market confidence. Since the launch of the 14th Five-Year Plan, the government has designated the biopharmaceutical industry as a strategic emerging industry and introduced a series of supporting policies.⁴¹ At the same time, after multiple rounds of NVBP, the product quality and supply capabilities of domestic enterprises have been validated in large-scale procurement, which has further driven the growth of their market share.

This study found that during the first-NVBP period, the insulin procurement volume at tertiary, secondary and primary medical institutions all increased significantly; during the second-NVBP period, the insulin procurement volume at medical institutions of all grades continued to grow in step. This trend of synchronous growth has broken through the long-standing supply bottleneck faced by primary medical institutions. Tertiary and secondary hospitals maintain mature and stable drug supply chains. In contrast, primary medical institutions have long struggled with fragmented procurement volumes, high distribution costs and unreliable drug supply, making them the weakest link in insulin accessibility.⁴² Previous studies have confirmed that affordability and accessibility are key factors limiting the widespread use of insulin⁴³; however, existing research on NVBP for insulin has largely focused on improving affordability, and empirical evidence regarding accessibility at primary medical institutions remains scarce. The synchronous increase in usage observed across all grades of medical institutions in this study is a direct reflection of policies driving the benefits of NVBP to the grassroots level. In 2024, the government explicitly encouraged primary medical institutions to fully participate in NVBP,⁴⁴ and in 2025, provinces further promoted the extension of NVBP to primary medical institutions. These measures have contributed to improving insulin access in primary care institutions.

We acknowledge that the ITS design cannot fully separate the NVBP effect from other time-varying influences such as post-COVID-19 healthcare recovery, seasonal care patterns and expanded diabetes screening. Broader secular trends, including rising diabetes prevalence,³⁴ evolving clinical guidelines and insurance reimbursement reforms, may also have independently contributed to the growth in insulin usage. However, the consistency of the results across all subgroups and the stability of the pre-intervention trends suggest that the NVBP was an important contributor to the observed changes.

Strengths and limitations

Our study has several strengths. First, we collected provincial-level data covering nearly 60 million permanent residents⁴⁵ to evaluate the effectiveness of the NVBP policy. To our knowledge, this is the first study to use provincial-level data to evaluate the effectiveness of two consecutive rounds of insulin NVBP. In addition to examining the policy’s impact on insulin consumption, medication patterns and market structure, this study also is among the first to provide an analysis of the policy’s effects from the perspective of institutions at different grades. Second, this study employs a segmented regression analysis design using the ITS model, which is suitable for examining how interventions affect outcomes both immediately and over time. The ITS design separates the intervention effect from the underlying pre-intervention trend, providing both immediate level-change and trend-change estimates, with intuitive graphical representations.^{27 46} Finally, the sample data in this study spans 68 months, allowing the results to demonstrate the long-term outcomes following policy implementation.

Our study also has some limitations. First, this study was based on the procurement data from a single province. Due to significant regional variations in insulin usage patterns,²² the generalisability of findings from a single province to the national level warrants further consideration. Second, the outcome indicators were derived from procurement records. Procurement volumes reflect institutional ordering rather than actual patient insulin consumption, and the mandatory fulfilment of agreed NVBP volumes does not fully capture independent prescribing choices or genuine market acceptance. These limitations warrant further research using patient-level dispensing data. Third, although the full sample data in this study covers a long time span, the data collected following the second NVBP for insulin spans only 16 months. Although this duration satisfies the statistical requirements for ITS analysis, the long-term effects of the policy during this NVBP period still warrant verification through extended follow-up studies. Finally, as a quasi-experimental design, the ITS analysis cannot fully exclude time-varying confounders; these factors and their implications are addressed in the final paragraph of the Discussion.

Conclusions

This study systematically evaluated the long-term policy effects of the NVBP for insulin based on provincial-level procurement data spanning 68 months. The findings confirm that two consecutive rounds of NVBP for insulin have led to sustained price reductions, with steady long-term growth in overall procurement volume. Further subgroup analysis revealed the following: Insulin utilisation patterns have significantly improved, with the growth rate of insulin analogue procurement being significantly faster than that of human insulin; market structure has been continuously optimised, with the growth rate of insulin procurement by domestic enterprises being significantly faster than that of imported enterprises; medical supply has developed in a balanced manner, with insulin procurement volumes showing sustained and steady growth across all tiers of medical institutions.

Data availability statement

Data are available upon reasonable request. The raw procurement data used in this study are subject to confidentiality restrictions from the Hubei Provincial Centralized Procurement Platform and cannot be publicly shared. However, the de-identified, aggregated and cleaned datasets generated and analysed during the current study are available from the corresponding author on reasonable request.

Ethics statements

Patient consent for publication

Not applicable.

Ethics approval

Not applicable.