A literature search reveals that the only study of the cost effectiveness for diabetes interventions in an Asian developing country is the Indian Diabetes Prevention Programme . However, this study was only based on 3 years of short-term follow-up and, therefore, does not reflect the long-term economic profile of diabetes prevention. Our study has conducted the first economic analysis of systematic prevention for diabetes in China and includes a large amount of high-quality data on benefits and costs, and also includes various sensitivity analyses.
Significant differences of remaining survival years for individuals with IGT or diabetes were found between the three lifestyle intervention strategies (diet, exercise and duo-intervention) and the other two groups (one-off screening with no intervention strategy and the control) at the initiation ages of 25 and 40, but not 60 years. Consistent with previous studies [12–14, 16], our results showed that the extra time for pre-diabetes before diabetes onset was existent, and it was 0.49–2.94 years at the three initiation ages of the simulated intervention strategies. This extra time was less than that found in the USA where it was 6.3 years for prevention initiated at age 30 years, 4.72–5.98 years at age 45 years, and 1.83 years at age 60 years or older . A likely explanation for this difference could be that diabetes was diagnosed much earlier and effective medications were prescribed more widely in higher income countries, while effective management and screening of diabetes and IGT are fairly limited in developing countries [3, 46–48]. For the screening alone strategy, the extra time was 0.04 years, which is comparable to the range of 0.02–0.08 reported in the UK and Taiwan [19, 20, 26]. Consequently, the strategy of screening alone, which results in early monitoring or treatment for diabetes, can improve health . In terms of preventing complications, there is now a broad consensus that the earlier diabetes is detected and treated, the greater the likelihood that complications will be prevented or delayed . Our study has further confirmed that screening and interventions increase the time before developing any complications in IGT subjects.
The estimated costs of screening and lifestyle interventions in this study ($360–$370 per year) are much lower than those of the Diabetes Prevention Programme conducted in the USA ($2780 per year) . This lower cost is primarily because of the following reasons: (i) the personnel costs in developing countries are lower, (ii) our study was based on simulated community interventions, which included group counselling instead of the more costly one-to-one clinical trials, and (iii) the average cost of diabetes intervention in a typical Chinese community was estimated from a later starting period than that of the American programme that started during the initial phase of intervention, which is a more costly period since it included frequent laboratory tests and clinical follow-ups. Although the prevention costs are lower in developing countries, some patients may still be unwilling to pay the costs especially for long-term preventions, such as in China. This is mainly owing to the fact that most Chinese have to pay all prevention expenses out of their own pockets irrespective of the insurance plan that they have . This situation differs from that in many European countries where insurance companies and other health providers cover most costs of the prevention programmes .
In the present study, all simulated screening strategies reduced the lifetime costs by approximately $390 or more per screened subject at all the initiation ages as compared with the control; in other words, all the prevention programmes were cost saving. These results are not only considered economically attractive by international standards , but are also seen as better than some cost-effective diabetes prevention programmes in high-income countries, such as in Taiwan ($17,113 per QALY gained) , Australia ($10,142 per QALY gained) , USA ($9731 per QALY gained)  and UK ($8358 per QALY gained) . It is noteworthy that the least savings gained by interventions in our study were approximately $2000 per subject among different initiation ages, which is still more than that of some cost-saving countries like Mexico ($1000 saved per subject) , Switzerland ($1040 saved per subject) and Germany ($600 saved per subject) . This benefit is due to the long-term effects beyond the intervention period for postponing or averting diabetes and related complications, which bring about substantial high medical costs in China [4, 5], even when the interventions were performed for just 6 years.
Among the three simulated intervention programmes, screening with exercise had the greatest savings at all three starting ages compared with the control. However, the differences between these three lifestyle interventions were insignificant. A possible explanation for this could be that the incidences and mortalities of the main diabetes-related complications that we used were the same, since these parameters were calculated after combining the different lifestyle intervention groups [35, 36].
Targeting a population at an appropriate age for receiving preventive intervention should be more effective for lowering costs than by targeting a non-specific age. Because the best net health benefits and the greatest saving costs were realized at the age of 25 years, selection of appropriate age groups should not ignore the young adults in developing countries, even though the current recommendations in developed countries are that screening should begin at the age of 45 years . As for older subjects, lifestyle intervention was also observed to result in a reduction of costs and more favourable health consequences, which is similar to the pharmacological interventions . However, these results were not as marked as those for the young and middle-aged, as the older patients with pre-diabetes were not as susceptible to improvements because they responded poorly to a single lifestyle intervention as compared with receiving two or more interventions together . Furthermore, along with the higher incidence and mortality of disease and the significant complications or comorbidities [41, 51], the older subjects with IGT were more likely to remain in the death or diabetes-related disorder states since there was no intervention at all in the screening alone and control groups.
Sensitivity analysis found that long-term outcomes were not sensitive to the changes of reference assumptions and they still supported the conclusion, i.e. screening and intervention strategies are cost saving. Because of the chronic and asymptomatic nature of diabetes, screening performance and compliance are important issues for diabetes prevention. Similar to previous studies , we found that a lower detection level of IGT from screening and lower compliance with screening resulted in higher costs and lower health effectiveness. However, the prevention programmes may still be worth expanding in China, since they were shown to be cost saving. Meanwhile, maintaining high rates of screening performance and compliance will lead to favourable clinical and economic effects on diabetes prevention.
There were also a few limitations of this study. First, the utilities we used were estimated based on the US population, and they may be on the low side compared with Chinese circumstances [57, 58]. Thus, the overall QALYs would increase when the China-specific utilities are available. Nonetheless, the lower utilities might not have significant impact on the stability of the models, since the saving costs were not very sensitive to most changes of utilities.
A second limitation of this study was that the longer life span induced by preventions may result in additional expenditures called ‘future costs’ , which probably should have been considered in the economic analysis . Unfortunately, because of the lack of comprehensive data to estimate the future costs accurately , we only included the costs of diabetes prevention and treatments during the simulated 40 years. However, it has been shown that economic analysis, excluding future costs and effects, could still maintain internal consistency .
A third limitation was that the influence of repeated screening was not assessed in this study. The saving costs for longer time interval screening ought to be similar to that of one-off screening, while it might decrease for short interval screening, because the prevalence of undiagnosed diabetes and IGT could possibly be lower and result in a higher cost per case [18, 20]. The sensitivity analyses suggested that the saving costs of all strategies were insensitive to the prevalence of IGT and diabetes, that is, whether a subject undergoes a repeat screening might not actually affect our results.
Finally, this study did not consider the impact of behavioural or biomarker risk factors such as smoking and haemoglobin A1c concentration [24, 51]. Also, the potential benefits of screening and management of related disorders like hypertension and hyperlipidaemia were not taken into account [3, 18]. Further investigations of China-specific clinical data with reference to the related subgroups should be explored.
Despite these limitations, the model truly reproduces the effects of diabetes screening and lifestyle interventions. Comprehensive validations that were performed further promoted the accuracy of the model. The findings of this study should be applicable to real lives in China and other developing countries, and it should be able to assist governments of developing countries on strategic decision making regarding health resource allocation over the long term.