AP is one of the most important etiologies for pancreatic diabetes. Comparing to the general incidence of DM (4–9%), the incidence in AP patients was much higher, especially within one year. It was reported that the overall incidence of DM after AP could be up to 23%, and the one-year incidence was about 15% (6). In the present study, 8.6% patients with first-attack AP developed new-onset DM within one year. The lower incidence might partly be attributed to the loss of potential positive patients due to the short-term follow-ups. The lower proportion of severe AP patients in this study (19.95% of APACHE II grade ≥ 8 points and 9.85% of CT grade ≥ 4 points) (Table 1) might be another reason for the lower incidence of DM since the positive correlation between the morbidity rate and the disease severity of AP (9–11).
It was generally accepted that AP could lead to destruction of pancreatic endocrine cells and therefore increase the incidence of DM due to the local and systemic immune response. This type of pancreatic diabetes had been classified as type 3c diabetes (3–5). As reported previously, risk factors for the development of DM after AP could be generally classified into two categories named patient susceptibility and disease features. In the present study, demographic and clinical data were comprehensively enrolled as candidate risk factors for further analyses, which might provide more valuable information. Our study showed that hyperlipemia, GGT ≥ 40U/L, serum glucose ≥ 6.1mmol/L, CT grade ≥ 2 or 4 points and APACHE II grade ≥ 8 points were independent risk or protective factors (Table 2). As for hyperlipidemia, a common etiology for AP, were also found to be correlated to the elevated risk of developing DM after AP (17–19). Das et al (6) proposed that AP may be a triggering factor for DM in patients with certain susceptibility conditions such as autoimmune disease, genetic susceptibility, certain metabolic factors such as obesity and hypertriglyceridemia, and/or changes in pancreatic structure and function. Another study also concluded that AP may trigger a reaction in gene susceptible people already at risk of developing DM (6). Combining the present results and previous studies, we speculated that hyperlipidemia might serve as a susceptibility factor for the development of DM after AP, and was considered a suitable index for the risk prediction of these patients.
Previous studies reported that disease severity of AP, which characterized as the extend of pancreatic necrosis and endocrine dysfunction, was considered correlated with the risk of new-onset DM (6, 8–11, 21–23). DM might be induced when the amount of nonfunctional or dead pancreatic β-cells increased to the specific threshold value due to the severe necrosis of the pancreatic tissues caused by AP (24). Subsequent pancreatic tissue atrophy was also an important modality for the decreased normal pancreatic β-cells and insulin secretion due to the widespread pancreatitis, and was considered to be correlated to the development of DM (25). In addition, pancreatic endocrine dysfunction was more common in severe AP patients compared to those with mild AP (26). Specially, the morbidity rate of endocrine dysfunction was found up to 33% in patients with pancreatic necrosis (27). In this study, clinical data reflecting the disease severity of AP were enrolled for analyses, including CT grades and APACHE II grades. These two factors were more practical since they were preferable factors reflecting the local and systemic inflammatory responses and the pathophysiological status, and could be well quantified and were considered to be more objective. Results showed that higher CT grades and APACHE II grades were independent risk factors for the development of DM after AP. We concluded that disease severity of AP was crucial for the development of DM, and objective factors such as CT grades and APACHE II grades might be effective indexes for the establishment of risk prediction models.
The elevated serum glucose in AP patients was used to be considered as transient hyperglycemia caused by carbohydrate metabolism disorders due to the acute stress, excessive secretion of catecholamines and subsequent pancreatic microcirculation disorders (28, 29). However, it was found that AP patients with transient hyperglycemia exhibited higher incidence of DM, since a considerable number of these patients could not fully recover from high blood glucose, or might experience a short-term recovery (30). As previously reported, insulin resistance was more common in patients after AP, and might serve as a crucial factor in the onset of DM (26). The present study also showed that serum glucose ≥ 6.1mmol/L was also extracted as an independent risk factor for the development of DM after AP, which further supported the previous studies, and was considered an important risk prediction index. Besides, GGT < 40U/L was also extracted as an independent risk factor. As reported previously, higher serum GGT level was found to be positively associated with the elevated incidence of DM (31), while another study found that DM was associated with GGT levels within the normal concentration range (32). Currently, no definite study was found to explore the GGT level and the incidence of pancreatic diabetes specially cause by AP, which was classified as type 3c DM. Whether the above results were applicable in this specific patient cohort remained to be further investigated, and more subgroup analyses were urgently needed in the future.
Nomogram is a simple graph that showing the incidence of indicated clinical events by adding the corresponding scores derived from each risk factor, and has been widely used in disease diagnosis and prognosis prediction. Currently, only a few studies were conducted to predict the risk of DM after first-attack AP (13, 14). However, there existed several limitations which weakened the strength of the corresponding results in these studies. Firstly, the enrolled potential risk factors were not comprehensive enough, especially those quantizable and objective factors such as disease scoring systems based on laboratory and imaging indexes. Besides, the established prediction models were not fully validated, especially on statistical and clinical aspects. In the present study, more practical clinical indexes were enrolled for analyses, including two important scoring systems for the evaluation of the severity of AP. Results also confirmed that they were independent risk factors and were subsequently used for the establishment of the prediction nomogram. Besides, both internal and external validations in terms of the discrimination and calibration capabilities were conducted and showed favorable results, which confirmed the accuracy of the established nomogram on the statistical aspect. Furthermore, DCA was also conducted in this study to evaluate the clinical applicability by quantifying the net income of the prediction nomogram according to the threshold probability (33). Results showed that the decision curve of this nomogram was favorable and considered to be clinically effective for predicting the incidence of new-onset DM after first-attack AP.
There also existed several limitations in the present study. Firstly, this was a relatively small sample-sized retrospective study, and the proposed risk prediction nomogram needed to be validated in a prospective study with a larger sample size. Besides, the definite latency period between AP attack and DM onset remained unclear, further studies with longer follow-ups were needed to investigate the disease features and the corresponding risk factors. Thirdly, the external validation was conducted only in a single center, which might weaken the application scope of this nomogram. External validations in multicenter patient cohorts would provide more convincing evidence for the nomogram.
In conclusion, we established a practical nomogram to predict the risk of new-onset DM after first-attack AP based on independent risk factors derived from demographic and clinical data. This nomogram was further confirmed to be effective in terms of the discrimination and calibration capabilities, as well as the clinical applicability, which would contribute to the identification and management of these high-risk patients. Multicenter studies with a larger sample size and more comprehensive risk factor analyses and external validations were needed in the future.