Extent of Beta Cell Destruction Is Important but Insufficient to Predict the Onset of Type 1 Diabetes Mellitus Authors: David J. Klinke Published: 2008, PLOS ONE DOI: 10.1371/journal.pone.0001374 ABSTRACT Type 1 diabetes mellitus is characterized by an inability to produce insulin endogenously. Based on a series of histopathology studies of patients with recent onset of the disease, it is commonly stated that the onset of clinical symptoms corresponds to an 80-95% reduction in beta cell mass. Motivated by the clinical importance of the degree of beta cell destruction at onset, a meta-analysis was used to determine the validity of this common wisdom. The histopathology results identifying insulin containing islets in patients younger than 20 years of age were extracted from three different studies. The results for 105 patients were stratified by duration of diabetic symptoms and age at onset. Linear regression and a non-parametric bootstrap approach were used to determine the dependence of residual beta cell mass to age at onset. The percentage reduction in beta cell mass was highly correlated (p<0.001) with the age of onset with the greatest reduction in beta cell mass in the youngest patients. As this trend had not been previously observed, an alternative physiology-based model is proposed that captures this age-dependence. The severity in beta cell reduction at onset decreased with age where, on average, a 40% reduction in beta cell mass was sufficient to precipitate clinical symptoms at 20 years of age. The observed trend was consistent with a physiology-based model where the threshold for onset is based upon a dynamic balance between insulin-production capacity, which is proportional to beta cell mass, and insulin demand, which is proportional to body weight. INTRODUCTION Type 1 diabetes mellitus is characterized by an impaired ability to produce insulin due to the progressive and selective destruction of beta cells in the pancreatic islets of Langerhans [1] . The destruction of beta cells is attributed to a variety of environmental and genetic risk factors [2] and results in the inability to maintain glucose homeostasis. While there are promising therapeutics in the drug pipeline (e.g. anti-CD3 [3] ), there are currently no treatments available to stop the progressive destruction of insulin-producing beta cells. The standard of care for many patients with Type 1 diabetes mellitus is currently exogenous insulin or insulin analog therapy [4] . The requirement for exogenous insulin, associated with reduced control of glucose homeostasis, results in an increased incidence of significant long-term side effects [5] . Any effort to prolong endogenous insulin production may have a significant impact on the long-term quality of life for patients with Type 1 diabetes mellitus. Clinical management of this disease faces two significant challenges: the increase in incidence of Type 1 diabetes mellitus across the globe [6] , [7] and the lack of a cure. Improved understanding of the natural history of this disease underpins possible solutions to these challenges. One of the persistent challenges with understanding the natural history of Type 1 diabetes mellitus is the inability to observe directly the events in the human pancreas that lead to the onset of hyperglycemia. It is clear that a reduction in endogenous insulin production precipitates the onset of hyperglycemia. Knowledge detailing the pathogenesis in humans is limited to a small number of biopsy studies from individuals with recent disease onset and to data from individuals who died soon after diabetes onset [8] – [11] . Based upon these landmark studies [12] , it is a widely held belief that the onset of hyperglycemia occurs when 80–95% of an individual's beta cells are destroyed [13] , [14] . The level of destruction may not be so severe as a series of recent studies suggest that impairment in beta cell function may also contribute to the emergence of hyperglycemia [15] – [17] . Quantifying the relative reduction in beta cell mass has important implications for the design of clinical trials to prevent or to improve management of the disease [18] . Given that common wisdom suggests that the reduction in beta cell mass is so severe at onset, one might conclude that the ability to enhance beta cell function or preserve the remaining beta cells would have limited therapeutic potential. In addition, the extent of reduction at onset can influence the debate [19] over the risk versus benefit analysis for novel drugs that prolong endogenous insulin production. Given the clinical importance of this question, the objective of this study was to re-examine the histopathological evidence that has long provided the basis for the conceptual models for the pathophysiology of Type 1 diabetes mellitus. A meta-analysis was used to extract and assess the significance of embedded trends within these landmark studies. It was found that the trends observed within these landmark studies do not fully support the common wisdom that onset of clinical symptoms of Type 1 diabetes occurs when 80–95% of an individual's beta cells are destroyed. An alternative model is proposed where the level of destruction varies with age and depends on the insulin-producing capacity, which is proportional to beta cell mass, and insulin demand, which is proportional to body weight. RESULTS The reduction in beta cell mass for recent onset patients with Type 1 diabetes were extracted from three landmark studies identified in the literature. For the two studies by Foulis and coworkers, the percentage reduction in beta cell mass was based on the ratio of the number of insulin deficient islets to the total number of islets observed following immunohistochemical staining of serial sections of the pancreata from Type 1 diabetic patients [8] , [9] . Concurrent observation of insulin-deficient and insulin-containing islets provides an internal control to correct for age-dependent and experimental variability. This calculation assumed that the density of islets in pancreata from Type 1 diabetic patients is representative of age-matched normal controls. To justify this assumption, it is noted that the number of islets and the density of somatostatin-secreting delta cells, glucagon-secreting alpha cells, and other cells of the endocrine pancreas are similar between Type 1 diabetic patients and normal controls [9] . In the study by Gepts [10] , the total pancreatic volume and number of beta cells were reported per area of pancreatic tissue for both non-diabetic controls and Type 1 diabetic patients. The calculated reduction in beta cell mass was based on the changes in density of beta cells within pancreata from diabetic patients, as the total pancreatic volume for recent onset patients was reported to be similar to non-diabetic controls. The percentage reduction in beta cell mass was determined based on the ratio of observed density of beta cell to the expected beta cell density based on a patient's age. As shown in Figure 1 , non-linear regression (solid line) was used to estimate the density of beta cells in non-diabetic controls (filled circles) as a function of age. In non-diabetic controls, the density of beta cells in the pancreas declines following birth and reaches a plateau at 25 years of age. The density of beta cells for patients diagnosed with Type 1 diabetes mellitus prior to their death (squares) are also shown in Figure 1 . The density of beta cells were reported for non-diabetic controls (filled circles) and Type 1 diabetics (open squares) reported by Gepts [10] . A non-linear regression was used to estimate the expected density of beta cells in non-diabetic controls as a function of age (solid line). In diabetics, the percentage reduction in beta cell mass is estimated by the ratio of the observed beta cell density in diabetics to the expected density provided by the non-linear regression. The reduction in beta cell mass is highly variable among a subset of 63 patients identified to have died within three weeks of diagnosis for Type 1 diabetes mellitus (recent onset patients) and shows dependence on age of diagnosis (see Figure 2 ). The linear regression results suggest that a greater reduction in beta cell mass is required for the onset of hyperglycemia in younger patients compared to older patients. Quantitatively, linear regression of the recent onset patients resulted in an intercept and slope of 85.5 (95% C.I. = 75.3 to 95.2) percent and −2.15 (95% C.I. = −3.02 to −1.30, p <0.001) percent year −1 , respectively. The subset of recent onset patients exhibited considerable inter-individual heterogeneity whereby 95% of the population falls within ±45% of the average reduction in beta cell mass for a given age (dashed lines). The data points, shown as black symbols, summarizes the percentage reduction in beta cell mass in 63 subjects that died within three weeks of diagnosis of type 1 diabetes mellitus from three different studies (squares [10] , diamonds [9] , and circles [8] ). The solid line represents the linear regression to the patients with recent onset type 1 diabetes mellitus. The 95% confidence intervals that enclose the population of recent onset type 1 diabetics patients are shown by the dashed lines. Individually, the sets of data analyzed in this study exhibit known heterogeneity due to different measurement techniques and potential patient selection bias. Thus, it is important to assess how combining these individual sets of data influences the linear regression results. A computational Bayesian approach, based upon a Markov chain Monte Carlo (MCMC) algorithm, was used to estimate how the uncertainty in the linear regression parameters changed upon combining sets of data [20] . A 2-dimensional (2-D) marginal posterior density (2-D MPD) was used to estimate the confidence regions of the linear regression parameters. In Figure 3 , a single contour line indicates the 2-D MPD region within the parameter space that contains 95% of the total marginal posterior density (i.e., the 95% confidence region). The MCMC algorithm was repeatedly applied to the Gepts data only (red curve), the Foulis data only (blue curve), and the combined data (black curve). This repetitive approach was used to assess how the estimates of the parameter values change when considering a composite data set derived from individual sets of data. The confidence regions for the slope and intercept were estimated from the Gepts data (red) [10] , Foulis et al. data (blue) [8] , [9] , and combined data (black). The curves indicate the parameter values that are enclosed within 95% of the posterior probability density function. The confidence regions associated with the combined data is a non-linear combination of the confidence regions associated with the Gepts data only and the Foulis et al. data only. To quantify the effect of the inclusion criteria for the analysis, a series of linear regressions were applied to a subset of patients who died prior to the cutoff. The cutoff was varied between one week and 312 weeks after diagnosis. These linear regression results for various cutoff times are shown in Table 1 . The linear regression of the entire population of 105 patients also exhibits the same general trends, where the extent of reduction in beta cell mass at onset decreases with age. The slope and intercept for the linear regression relationship of the entire population were determined to be −1.10 (95% C.I. = −1.86 to −0.35, p =  0.001) percent year −1 and 79.8 (95% C.I. = 70.8 to 89.4) percent, respectively. C.I. = confidence interval N data  = number of patients included in subset TABLE: The linear regression parameters and corresponding 95% confidence intervals reported as a function of change in the cutoff date. Cutoff Weeks N data Slope Bootstrap 95% C.I. Intercept Bootstrap 95% C.I. 1 30 −2.58 (−1.36 to −3.54) 90.09 (74.84 to 100.73) 2 47 −2.59 (−1.66 to −3.27) 91.91 (79.64 to 99.99) 3 63 −2.15 (−1.28 to −2.88) 85.46 (73.58 to 94.25) 6 72 −1.92 (−1.06 to −2.65) 84.00 (72.15 to 93.14) 10 78 −2.05 (−1.21 to −2.74) 84.03 (72.51 to 92.55) 15 87 −1.45 (−0.84 to −1.98) 79.11 (67.82 to 88.25) 26 95 −1.46 (−0.87 to −1.97) 80.34 (69.78 to 89.14) 52 97 −1.42 (−0.84 to −1.94) 80.63 (69.97 to 89.32) 104 103 −1.16 (−0.58 to −1.66) 79.88 (69.28 to 88.22) 312 105 −1.10 (−0.52 to −1.60) 79.81 (69.20 to 88.31) DISCUSSION The results from this meta-analysis suggest that clinical presentation of Type 1 diabetes mellitus is correlated with a greater percentage loss in beta cell mass in infants compared to adults. An 85% reduction in beta cell mass in infants can lead to hyperglycemia. In contrast, as little as a 40% reduction by 20 years of age is sufficient to induce hyperglycemia. This meta-analysis also demonstrates the significant heterogeneity in pathophysiology within a population of Type 1 diabetic patients. In older patients, a lower reduction in beta cell mass gives rise to the same clinical phenotype. This trend was also observed qualitatively by Foulis et al. [9] . Given this inconsistency with existing models for the natural history of the disease [2] , an alternative theory may be proposed to explain this behavior. Growth of the human body is a dynamic non-linear process where different parts of the body grow at different rates. Of particular relevance to Type 1 diabetes mellitus, body weight changes at a different rate than beta cell mass. One possible explanation for this observed trend in residual beta cell mass at onset could be attributed to the dynamic imbalance between the number of beta cells and the insulin requirements for a growing body. Individually, beta cell density and total volume of the pancreas both vary as a function of age. The age-related growth in total beta cell mass, as shown in Figure 4 (dotted line), can be estimated by the product of the beta cell density and the total volume of the pancreas [10] . The changes in beta cell mass (dotted line) and body weight in kg (solid line) are shown a function of age. The beta cell mass is the product of beta cell density [10] and volume of the pancreas [10] . The change in body weight as a function of age is an average value from male and female growth charts [25] . In addition, changes in body weight can influence how insulin regulates glucose homeostasis. Cellular response to insulin is proportional to the concentration of insulin. Concentration is a derived intrinsic variable and defined as the number of proteins per unit of volume. As the body grows, the volume in which proteins are distributed increases. Therefore, the number of insulin molecules must also increase to maintain a constant concentration (i.e., elicit the same biological effect). Insulin synthesis and release by beta cells exhibits a non-linear dependence on glucose concentration [21] , [22] . If we assume that the cellular capacity to synthesize insulin does not vary with age, each beta cell is capable of contributing a fixed number of insulin proteins per unit of time. Under normal conditions, beta cells may synthesize only a fraction of their maximum capability. As the number of beta cells decreases, the remaining beta cells may increase their individual contributions to maintain a normal concentration of insulin in the blood. As the beta cell mass progressively declines, individual contributions rise until a maximum level of protein synthesis is reached in the remaining beta cells. A subsequent decline in beta cell mass results in an inability of the body to maintain an appropriate concentration of insulin in the plasma. Clinically, metabolic parameters, such as fasting glucose, are observed to increase dramatically at the onset of hyperglycemia [23] , [24] . Admittedly, these studies were performed in high risk populations and may not generalizable to the entire Type 1 diabetes patient population. By focusing on patients who died within three weeks of onset, we hypothesize that the onset of hyperglycemia is a titration of the beta cell mass required to maintain glucose homeostasis for a given body weight. A quantitative relationship can be created to predict the “excess” beta cell mass (EBCM(t)) as a function of age. The “excess” beta cell mass corresponds to the reduction in beta cell mass that is required before hyperglycemia occurs. This relationship is derived from a dynamic mole balance between the source of insulin, which is proportional to beta cell mass (BCM Total (t)), and sinks for insulin, which are proportional to body weight (BWt(t)) [25] (see Appendix S1 ). Quantitatively, this relationship can be expressed as: (1) where K is a proportionality constant that provides an estimate of the required beta cell mass per kilogram of body weight. Using the patient results for the recent onset group, the value for the parameter K was estimated to be 499 units of BCM kg −1 (95% C.I. = 458 to 605). The resulting relationship for EBCM(t) as a function of age is shown in Figure 5 (solid line). The trend obtained by linear regression (dotted line) and the observed reduction in beta cell mass in pancreata obtained from the subset of recent onset patients are also shown for comparison. Comparison of the estimate of excess beta cell mass (solid curve) compared against the trend obtained by linear regression (dotted line) for the measured reduction in beta cell mass in 63 patients that died within three weeks of diagnosis of type 1 diabetes mellitus from three different studies (squares [10] , diamonds [9] , and circles [8] ). Qualitatively, the EBCM relationship exhibits a similar dependence with age, as the youngest patients exhibited the greatest potential reduction in beta cell mass. The trend also declines with increasing age, resulting in only a 40% excess in beta cell mass by the age of 20. In other words, the beta cell mass initially grows at a faster rate relative to the whole body. The overall beta cell mass plateaus while the overall body weight steadily increases through 20 years of age. The net result of the different growth dynamics is that the “excess” beta cell mass declines with age. Qualitatively, the observed reduction in beta cell mass in recent onsets diabetics appears to be normally distributed around the predicted “excess” beta cell mass. In the youngest patients (age 23 weeks (blue circles). In the recent onset group, the residual values appear to distribute normally around the predicted values. In addition, patients with duration greater than 23 weeks exhibited, on average, a greater loss in beta cell mass than predicted by the “excess” beta cell mass relationship (p<0.001). While not statistically significant, patients with an intermediate duration (3