Journal Club

1. Journal Club DorresteijnJA, Visseren FL, Ridker PM, Wassink AM, Paynter NP, Steyerberg EW, van der Graaf Y, Cook NR. Estimating treatment effects for individual patients based on the results of randomised clinical trials. BMJ. 2011 Oct 3;343:d5888. doi: 10.1136/bmj.d5888. MengeBA, Grüber L, Jørgensen SM, Deacon CF, Schmidt WE, Veldhuis JD, Holst JJ, Meier JJ. Loss of inverse relationship between pulsatile insulin and glucagon secretion in patients with type 2 diabetes. Diabetes. 2011 Aug;60(8):2160-8. Epub 2011 Jun 15. 埼玉医科大学　総合医療センター　内分泌・糖尿病内科 Department of Endocrinology and Diabetes, Saitama Medical Center, Saitama Medical University 松田　昌文 Matsuda, Masafumi 2011年10月13日8:30-8:55 ８階　医局

3. 30 4S - Placebo 25 Rx - Statin therapy PRA – pravastatin ATV - atorvastatin Secondary Prevention 4S - Rx 20 Event rate (%) LIPID - Placebo 15 CARE - Placebo LIPID - Rx CARE - Rx Primary Prevention HPS - Placebo HPS - Rx TNT – ATV10 10 PROVE-IT - PRA WOSCOPS – Placebo TNT – ATV80 PROVE-IT – ATV AFCAPS - Placebo 6 5 AFCAPS - Rx WOSCOPS - Rx ASCOT - Placebo ASCOT - Rx 0 40 (1.0) 60 (1.6) 80 (2.1) 100 (2.6) 120 (3.1) 140 (3.6) 160 (4.1) 180 (4.7) 200 (5.2) LDL-C achieved mg/dL (mmol/L) LDL-Cholesterol and Coronary event rate Adapted from Rosensen RS. Exp Opin Emerg Drugs 2004;9(2):269-279 LaRosa JC et al. N Engl J Med 2005;352:1425-1435

4. GALAXY PROGRAMME

5. The GALAXY Programme™ is designed to confirm the hypothesis that: • the statin with the best effects on the atherogenic lipid profile and beneficial changes on inflammatory markers will • deliver profound benefits on atherosclerosis • in turn leading to the best reductions in cardiovascular morbidity and mortality.

6. Inclusion criteria:LDL≦130 mg/dl and high-sensitivity CRP≧ 2.0 mg/l ■Effect of Rosuvastatin(20mg) ・Compliance; 75% n = 8901 ｘ 2 = 17802 -37% -50% +4% -17% N Engl J Med 2008;359:2195-207.

7. The combined primary end point of myocardial infarction, stroke, arterial revascularization, hospitalization for unstable angina, or death from cardiovascular cause HR=0.56; 95% CI, 0.46 to 0.69; P<0.00001) 20 mg NNT:95 31 ･･25 (2years)(4years)(5years) N Engl J Med 2008;359:2195-207.

8. Lancet 2009; 373: 1175–82 LDL-C 70mg/dl For people choosing to start pharmacological prophylaxis, reduction in both LDL cholesterol and hsCRPare indicators of successful treatment with rosuvastatin.

9. However, to immediately translate these findings into clinical practice without appropriate and careful discussion of their implications is not prudent. TURBO STATIN Helene Trudel Jean-Pierre Despres www.thelancet.com Vol 373:1178 April 4, 2009

11. Objectives To predict treatment effects for individual patients based on data from randomised trials, taking rosuvastatin treatment in the primary prevention of cardiovascular disease as an example, and to evaluate the net benefit of making treatment decisions for individual patients based on a predicted absolute treatment effect.

12. Setting As an example, data were used from the Justification for the Use of Statins in Prevention (JUPITER) trial, a randomised controlled trial evaluating the effect of rosuvastatin 20 mg daily versus placebo on the occurrence of cardiovascular events (myocardial infarction, stroke, arterial revascularisation, admission to hospital for unstable angina, or death from cardiovascular causes). Population 17 802 healthy men and women who had low density lipoprotein cholesterol levels of less than 3.4 mmol/L and high sensitivity C reactive protein levels of 2.0 mg/L or more. Methods Data from the Justification for the Use of Statins in Prevention trial were used to predict rosuvastatin treatment effect for individual patients based on existing risk scores (Framingham and Reynolds) and on a newly developed prediction model. We compared the net benefit of prediction based rosuvastatin treatment (selective treatment of patients whose predicted treatment effect exceeds a decision threshold) with the net benefit of treating either everyone or no one.

13. Prediction of treatment effect for individual patients 10 year treatment effect (absolute risk reduction)=baseline risk without treatment−residual risk with rosuvastatin treatment Baseline 10 year absolute risk for cardiovascular events (%) without treatment (General Population) Framingham risk score based method: risk as calculated from the Framingham risk score, published in the Adult Treatment Panel III guidelines17 Reynolds risk score based method: risk as calculated from the Reynolds risk score, derived for women from the Women’s Health Study18 and for men from the Physicians Health Study19 Optimal fit model*: (1−0.985433(5×exp[B]))×100%, where: B=0.09379363 × age in years + 3.34656382 (if male) − 0.03698750 × age in years (if male) + 0.81823698 (if current smoker) + 0.54045383 (if using blood pressure lowering drugs) + 0.60281674 (if family history of premature coronary heart disease) − 6.9932 Residual 10 year absolute risk for cardiovascular events (%) with rosuvastatin treatment Framingham risk score: 0.56 × baseline 10 year absolute risk for cardiovascular events (%) without treatment Reynolds risk score: 0.56 × baseline 10 year absolute risk for cardiovascular events (%) without treatment Optimal fit model*: (1−0.985433(5×exp[B])×100%, where: B=0.09379363 × age in years + 3.34656382 (if male) − 0.03698750 × age in years (if male) + 0.81823698 (if current smoker) + 0.54045383 (if using blood pressure lowering drugs) + 0.00932154 (if family history of premature coronary heart disease) − 7.484613 *Treatment effect of rosuvastatin expressed in optimal fit model for residual risk as different coefficient for family history of premature coronary heart disease and different constant subtraction factor

14. Reynolds risk score Framingham risk score

15. Development of the ‘Optimal fit’ model The trial data were regarded as a single prospective cohort with variable follow-up time. The occurrence of the primary combined end point of myocardial infarction, stroke, arterial revascularization, hospitalization for unstable angina, or death from cardiovascular causes was modelled using Cox proportional hazard regression. To minimize over-fitting, candidate predictors for the outcome and transformations of variables were based on existing prediction models (the Framingham risk score and the Reynolds risk score).[1-3] The starting model, therefore, contained the following variables: rosuvastatin treatment (yes/no), age, gender, current smoking status, log transformed systolic blood pressure (SBP), log transformed high density lipoprotein cholesterol (HDLc), log transformed total cholesterol (TC), log transformed high-sensitivity C-reactive protein, use of blood pressure lowering medication, and family history of premature coronary heart disease. Parallel to the Framingham risk score, interactions between age and gender and between log transformed SBP and blood pressure lowering medication were also considered.[1] Continuous predictors were truncated at the 1st and 99th percentile to limit the effect of outliers. The model was backward selected on the basis of the Akaike’s Information Criterion (AIC). To minimize over-fitting, the simplest model was chosen if the model AIC increased by less than 2 after deletion of a factor (Occam’s razor).[4] Interaction of treatment effect with any remaining covariates was considered after initial backward selection of the starting model. The resulting ‘Optimal fit’-model was used to predict each patient’s 2 year risk for cardiovascular events with and without rosuvastatin treatment. Based on the assumption of constant hazard and, thus, exponential risk over time, the 2 year risk estimates were extrapolated to 10 years. The predicted absolute risk reduction achieved by rosuvastatin treatment at 10 years (10-year treatment effect) was defined as the difference between the two risk predictions (Box 1).

16. Table 1| Baseline characteristics of participants in Justification for the Use of Statins in Prevention trial. Values are medians (interquartile ranges) unless specified otherwise

17. high risk lower risk The median predicted 10 year absolute risk reduction for cardiovascular events was 4.4% (interquartile range 2.6-7.0%) based on the Framingham risk score, 4.2% (2.5-7.1%) based on the Reynolds score, and 3.9% (2.5-6.1%) based on the newly developed model (optimal fit model). Table 2| Calculation example of predicted 10 year treatment effect for two patient scenarios

18. Fig 1 Basic concept for weighing treatment effect against harm. Treatment effect usually increases with baseline risk, whereas harm is relatively constant for all patients. Those whose treatment effect exceeds treatment related harm (reflected by decision threshold) benefit from treatment

19. Fig 2 Calibration plots. Predicted and observed two year event free survival for cardiovascular events within 10ths of predicted survival using three models. P values derived from the Hosmer-Lemeshow test

20. Fig 3 Distribution of predicted 10 year absolute treatment effect (absolute risk reduction) based on Framingham risk score, Reynolds risk score, and optimal fit model, with coloured bars indicating predicted treatment effects for two different patient scenarios. JUPITER=the Justification for the Use of Statins in Prevention trial

21. we propose to rename the NNT that is associated with clinical equipoise as the number willing to treat (NWT). Fig 4 Decision curve: graphical representation of net benefit. For large values of numbers willing to treat (NWT), the net benefit of treating all patients is about equal to the net benefit of prediction based treatment. The net benefit of treating all patients becomes negative if the NWT is less than 20, whereas the net benefit of prediction based treatment is still positive for a NWT of 20 and converges to zero for smaller values of NWT

22. Calculation example of Net Benefit The net benefit of treatment according to Framingham based predictions versus treat no one (decrease in event rate – treatment rate x T) is, thus, 0.0287 - .465 x 0.05 = 0.0054. the number willing to treat (NWT) is set to 20. This means that the decision-threshold (T) is a 5% absolute risk reduction or more. The net benefit assessment method is described in detail by Vickers et al.[1] Here we provide an example of how we applied this method to our data. In this example, the number willing to treat (NWT) is set to 20. This means that the decision-threshold (T) is a 5% absolute risk reduction or more. The net benefit of Framingham risk score-prediction based treatment compared to treat all patients is calculated as follows: 1) First we observe the 2-year event rate in the placebo treated group, which is the event rate in the situation that no one is treated with rosuvastatin. The event rate of 2.47% in this example can be extrapolated to an event rate of 11.75% over 10 years follow up. 2) Second we observe the 2-year event rate in the rosuvastatin treated group. This is the event rate in the situation that everybody is treated with rosuvastatin. The event rate in the present example (1.39%) can be extrapolated to an event rate of 6.77% over 10 years follow up. The decrease in event rate of 4.99% was achieved at the cost of a treatment rate of 100%. 3) The net benefit of treat all patients versus treat no one (decrease in event rate – treatment rate x T) is, thus, 0.0499 - 1 x 0.05 = -0.0001. Notably, the minus sign means that in if the NWT is 20, treating no one is preferable over treating everyone. 4) Next, we observe the event rate in patients whose treatment allocation was congruent to their predicted treatment effect. This includes 4,106 intervention group patients whose predicted 10-year treatment effect exceeded the decision-threshold (i.e. 5% absolute risk reduction) and 4,726 control group patients whose predicted 10-year treatment effect was lower than the decision-threshold. The 2-year event rate in these groups of patients combined (n=8,832) was 1.84% (8.88% if extrapolated to 10 years). This means that Framingham risk score prediction-based treatment compared to treating no one reduced the event rate by 2.87% at the cost of a treatment rate of 46.5%. 5) The net benefit of treatment according to Framingham based predictions versus treat no one (decrease in event rate – treatment rate x T) is, thus, 0.0287 - .465 x 0.05 = 0.0054. This figure can also be found in results table 3 of the article. Notably, it means that if the NWT is 20, Framingham risk score prediction-based treatment is preferable over both treating everyone and treating no one.

23. Table 3| Results of net benefit assessment

24. Fig 5 Implications for clinical practice. Justification for the Use of Statins in Prevention trial shows that treatment of all patients is the strategy of choice if the 10 year number willing to treat (NWT) is 50 or more. Treating no one is preferable if the 10 year NWT is 15 or fewer. If the NWT is between 15 and 50, prediction based treatment results in most net benefit No

25. If the 10 year NWT to prevent one cardiovascular event is 20, patients with a baseline (for example, Framingham score or Reynolds score) risk of 11.4% or more (95% confidence interval 9.3% to 16.1%) benefit from treatment. Likewise, if the 10 year NWT is 30, patients with a baseline risk of 7.6% or more (6.2% to 10.8%) benefit from treatment.

26. Presentation of treatment-effect prediction models should preferably include the following: • Model structure and coefficients (Box 1). These are needed to build online treatment-effect calculators and implement the model in electronic patient record systems. • Measures of model performance, such as a calibration curve (Figure 2) and Hosmer-Lemeshowtest to assess calibration and a c-statistic to assess discrimination. Other performance measures may also be used. • A histogram of the distribution of predicted treatment effect in the trial population (Figure 3). This could help to determine whether the predicted treatment effect for an individual patient is lower, equal to or higher than average. • A decision-curve covering a clinically relevant range of NWT (Figure 4). The methods for calculating net benefit and plotting a decision curve are described in detail by Vickers et al.

27. Results The median predicted 10 year absolute risk reduction for cardiovascular events was 4.4% (interquartile range 2.6-7.0%) based on the Framingham risk score, 4.2% (2.5-7.1%) based on the Reynolds score, and 3.9% (2.5-6.1%) based on the newly developed model (optimal fit model). Prediction based treatment was associated with more net benefit than treating everyone or no one, provided that the decision threshold was between 2% and 7%, and thus that the number willing to treat (NWT) to prevent one cardiovascular event over 10 years was between 15 and 50.

28. Conclusions Data from randomised trials can be used to predict treatment effect in terms of absolute risk reduction for individual patients, based on a newly developed model or, if available, existing risk scores. The value of such prediction of treatment effect for medical decision making is conditional on the NWT to prevent one outcome event. Trial registration number Clinicaltrials.gov NCT00239681.

29. Message/Comments JUPITER試験（ロスバスタチンの心血管疾患一次予防効果の検討）を例に、無作為化試験のデータを使用した治療効果予測能と予測に基づいた治療意思決定の純便益評価を検討。データを基に開発した予測モデルによる個々の患者の絶対リスク低減から治療効果が予測でき、治療意思決定における効果予測の評価は、積極的に治療を受ける患者数によることが示唆された。

31. 2型糖尿病患者の食事に対するインスリンとグルカゴンの反応2型糖尿病患者の食事に対するインスリンとグルカゴンの反応 2型糖尿病(n=12)* 非糖尿病(n=11) 360 330 食事 300 血糖値 (mg/dL) 270 240 110 80 150 インスリン反応の抑制／遅延 120 90 インスリン (µU/ml) 60 30 0 140 130 120 グルカゴン (µµg/ml) グルカゴンの過剰分泌 110 100 90 –60 0 60 120 180 240 時間(分) *5例の患者についてインスリン測定を行った。 Müller WA et al N Engl J Med 1970;283:109–115.より改変

32. the 1Department of Medicine I, St. Josef-Hospital, Ruhr-University Bochum, Bochum, Germany; the 2Department of Biomedical Sciences, The Panum Institute, University of Copenhagen, Copenhagen, Denmark; and the 3Department of Medicine, Endocrine Research Unit, Mayo School of Graduate Medical Education, Mayo Clinic, Rochester, Minnesota. Diabetes 60:2160–2168, 2011

33. OBJECTIVE—In patients with type 2 diabetes, glucagon levels are often increased. Furthermore, pulsatile secretion of insulin is disturbed in such patients. Whether pulsatile glucagon secretion is altered in type 2 diabetes is not known.

34. RESEARCH DESIGN AND METHODS—Twelve patients with type 2 diabetes and 13 nondiabetic individuals were examined in the fasting state and after mixed meal ingestion. Deconvolution analyses were performed on insulin and glucagon concentration time series sampled at 1-min intervals.

36. Study design All antidiabetic treatment was withdrawn at least 48 h before the start of the study. In insulin-treated patients, short-acting insulin was withheld on the evening preceding the tests, whereas all long–acting insulin preparations were withdrawn at least 24 hours before the experiments. OGTT for nondiabetic controls Subjects were studied on two occasions: (1) a fasting experiment over 60 min, with venous blood samples being collected at 1-min intervals; and (2) a mixed meal test, with venous blood samples being collected over 240 min (from 30 to 90 min after meal ingestion, blood samples were collected at 1-min intervals). Both tests were performed in randomized order, separated by an interval of 1–4 weeks to allow for regeneration of the blood pool. Experimental procedures Each test was performed in the morning after an overnight fast, with subjects in a supine position throughout the experiments. A large forearm vein was kept patent using 0.9% NaCl. On visit one, venous blood samples were taken at t = 25 and 0 min (9 mL each) as well as at 1-min intervals from 0 to 60 min (3 mL each). On visit two, a mixed meal (total caloric content 537 kcal; 42% fat, 41% carbohydrates, and 17% protein) was ingested together with 200 mL water over 5 min. Venous blood samples (9 mL each) were drawn frequently, as shown in Fig. 1. In addition, venous blood samples (3 mL each) were collected at 1-min intervals from t = 30 to 90 min after meal ingestion.

37. FIG. 1. Plasma concentrations of glucose (A and B), insulin (C and D), and glucagon (E and F) in 12 patients with type 2 diabetes (open symbols) and 13 nondiabetic control subjects (filled symbols) studied in the fasting state (A, C, and E) as well as after mixed meal ingestion (B, D, and F). Arrows indicate the time of meal ingestion. Data are presented as means 6 SEM. Statistics were carried out using repeated-measures ANOVA and denote (A) differences between the groups, (B) differences over the time course, and (AB) differences due to the interaction of group and time. Asterisks indicate significant differences at individual time points (P < 0.05).

38. fasting state nondiabetic control subject type 2 diabetes FIG. 2. Individual concentration time series of insulin (A and B) and glucagon (E and F), as well as the respective insulin secretion rates (C and D) and glucagon secretion rates (G and H), derived from deconvolution analysis, in a representative nondiabetic control subject (left) as well as in a patient with type 2 diabetes (right). The measurements were taken at 1-min intervals over 60 min in the fasting state.

39. fasting state nondiabetic control subject type 2 diabetes FIG. 2. Individual concentration time series of insulin (A and B) and glucagon (E and F), as well as the respective insulin secretion rates (C and D) and glucagon secretion rates (G and H), derived from deconvolution analysis, in a representative nondiabetic control subject (left) as well as in a patient with type 2 diabetes (right). The measurements were taken at 1-min intervals over 60 min in the fasting state.

40. after mixed meal nondiabetic control subject type 2 diabetes FIG. 3. Individual concentration time series of insulin (A and B) and glucagon (E and F), as well as the respective insulin secretion rates (C and D) and glucagon secretion rates (G and H), derived from deconvolution analysis, in a representative nondiabetic control subject (left) as well as in a patient with type 2 diabetes (right). The measurements were taken at 1-min intervals from t = 30 to 90 min after mixed meal ingestion.

41. nondiabetic control subject type 2 diabetes after mixed meal FIG. 3. Individual concentration time series of insulin (A and B) and glucagon (E and F), as well as the respective insulin secretion rates (C and D) and glucagon secretion rates (G and H), derived from deconvolution analysis, in a representative nondiabetic control subject (left) as well as in a patient with type 2 diabetes (right). The measurements were taken at 1-min intervals from t = 30 to 90 min after mixed meal ingestion.

44. FIG. 4. A: Individual postprandial secretion rates of insulin (left y-axis) and glucagon (right y-axis), derived from deconvolution analysis, in a representative nondiabetic control subject. B: The corresponding linear regression analysis between the minutely sampled postprandial concentrations of insulin and glucagon in the same individual.

45. nondiabetic control subjects type 2 diabetes FIG. 5. Mean cross-correlogram (solid lines, filled symbols) between insulin and glucagon concentration time series in 13 nondiabetic control subjects (A) and 12 patients with type 2 diabetes (B) studied over 60 min after a mixed test meal. Data are presented in relation to the upper and lower 95% confidence intervals (dotted lines). *Significant (P < 0.05) interaction between the insulin and glucagon time series at a given lag period. These analyses demonstrate a significant relationship between the insulin and glucagon concentration time series after meal ingestion in nondiabetic individuals, whereas the time relationship between both hormones is absent in patients with diabetes.

46. Approximate entropy. Regularity of insulin and glucagon plasma concentration time series was assessed by the model-independent and scale-invariant statistics approximate entropy (ApEn). ApEn measures the logarithmic likelihood that runs of patterns that are close (within r) for m contiguous observations remain close (within the tolerance width r) on subsequent incremental comparisons. A precise mathematical definition has been given by Pincus (32). A larger absolute value of ApEn indicates a higher degree on process randomness, i.e., less pattern orderliness. ApEn is rather stable to noise that is within tolerances. Detrending the time series by first differencing before ApEn calculation has been used to limit nonstationarity. Cross-ApEnstatistics. To evaluate relative regularity of glucagon and insulin concentrations, we used the cross-approximate entropy statistic. Cross-ApEn is a bivariate analog of ApEn and quantifies joint pattern synchrony between two separate but linked time series after standardization (z score transformation of the data) (33,34). The interpretation of cross-ApEn as implemented here depends on whether pattern recurrence is assessed in the forward or reverse direction. By way of convention, we computed forward cross-ApEn using serial insulin concentrations as the template to assess pattern reproducibility in glucagon concentrations, and calculated reverse cross-ApEn by using successive glucagon concentrations as the template to evaluate pattern recurrence in insulin concentrations, as previously described (14). Note that, for physiological reasons, the foregoing schema of pairing relates the concentration of an input signal to the secretion rate of the output signal. Individual values for forward cross-ApEn were compared with respective reverse cross- ApEn values in each person to infer the driving force in the interaction between the secretion of both hormones.

47. FIG. 6. Cross-ApEn statistics for the interaction of insulin and glucagon concentration time series in 13 nondiabetic control subjects (A and C) and 12 patients with type 2 diabetes (B and D) studied over 60 min in the fasting state (A and B) as well as over 60 min after mixed meal ingestion (C and D). Analyses were carried out by entering either glucagon (forward) or insulin (reverse) first. P values were calculated using paired Student t test. fasting state nondiabeticcontrol type 2 diabetes Forward cross-ApEn was significantly higher (less synchronous) than reverse cross-ApEn (P < 0.001; Fig. 6), indicating that insulin may drive glucagon rather than vice versa. after mixed meal

48. RESULTS—Both insulin and glucagon were secreted in distinct pulses, occurring at ;5-min intervals. In patients with diabetes, postprandial insulin pulse mass was reduced by 74% (P <0.001). Glucagon concentrations were increased in the patients during fasting and after meal ingestion (P < 0.05), specifically through an increased glucagon pulse mass (P <0.01). In healthy subjects, the increase in postprandial insulin levels was inversely related to respective glucagon levels (P <0.05). This relationship was absent in the fasting state and in patients with diabetes.

49. CONCLUSIONS—Glucagon and insulin are secreted in a coordinated, pulsatile manner. A plausible model is that the postprandial increase in insulin burst mass represses the corresponding glucagon pulses. Disruption of the insulin–glucagon interaction in patients with type 2 diabetes could potentially contribute to hyperglucagonemia.

50. Message/Comments １分おきの採血ができると、インスリンとグルカゴン分泌の解析ができるようである。 インスリンがグルカゴンの分泌を制御している可能性があるらしい。 ２型糖尿病ではこの制御が壊れている。 モデルベースなので実際には疑問が残るが。