应用实例分析 - 临床流行病学研究 胎源性疾病 （ DOHAD)

骆忠诚上海交通大学附属新华医院

学习目标

通过应用实例分析 ， 加深对流行病学重要基本概念的理解

通过应用实例分析 ， 提高对流行病学研究重要方法的应用能力

内容

实例分析 一个经典的有关胎源性疾病的临床流行病学研究报告

实例分析 - 设计一个有关胎源性疾病的临床流行病学研究项目

胎源性疾病 / 发育源性疾病 (DOHAD)

The suscepatible ity to many chronic diseases in adulthood can be traced back to exposures during early life (during fetal and early postnatal life) 成年期的许多慢性疾病的易感性可以追溯到生命早期的暴露因素（胎儿期和出生后早期）

无致病基因突变

流行病学 - 重要作用之一 发现新联系 （ DOHAD ）

*Odds ratio for two hour glucose concentration of >7.8 mmol/l adjusted for current body mass index. (X2 for trend= 15.4; p<0.001).Reduced growth in early life is strongly linked with impaired glucose tolerance and non-insulin dependent diabetes (T2DM). 生命早期生长降低与成年期（ 64 岁）糖耐量受损和非胰岛素依赖型糖尿病密切相关。研究样本 · 466 男性 · （ 59-70 岁 · ， 平均 64 岁）– Hales CN, Barker DJ, et al. BMJ 1991;303:1019-22.

出生体重（ lb) 糖耐量异常 n/N （ % ）

调整比值比 (95% 可信区间 )Adjusted OR （ 95% CI ） *

低于 5.5 lb (2.5kg)

8/20 (40) 6.6 (1.5-28.0)

5.5-6.4 (2.9 kg) 16/47 (34) 4.8 (1.3-17.0)

6.5-7.4 (3.4 kg) 32/104 (31) 4.6 (1.4-16.0)

7.5-8.4 (3.9 kg) 26/117 (22) 2.6 (0.8-8.9)

8.5-9.4 (4.2 kg) 7/54 (13) 1.4 (0.3-5.6)

9.5+ lb (4.3 kg +) 4/28 (14) 1.0 ( 参照组 )

OBJECTIVE: To discover whether reduced fetal and infant growth is associated with non-insulin dependent diabetes and impaired glucose tolerance in adult life.

DESIGN: Follow up study of men born during 1920-30 whose birth weights and weights at 1 year were known.

SETTING: Hertfordshire, England. SUBJECTS: 468 men born in east Hertfordshire and still

living there. Men known to have diabetes were excluded. MAIN OUTCOME MEASURES: Fasting plasma glucose,

insulin, proinsulin, and 32-33 split pro-insulin concentrations and plasma glucose and insulin concentrations 30 and 120 minutes after a 75 g glucose drink.

Hales CN, Barker DJ, Clark PM, Cox LJ, Fall C, Osmond C, Winter PD. Fetal and infant growth and impaired glucose tolerance at age 64. BMJ 1991;303:1019-22

• RESULTS: 93 men had impaired glucose tolerance or hitherto undiagnosed diabetes. They had had a lower mean birth weight and a lower weight at 1 year. The proportion of men with impaired glucose tolerance fell progressively from 26% (6/23) among those who had weighted 18 lb (8.16 kg) or less at 1 year to 13% (3/24) among those who had weighed 27 lb (12.25 kg) or more. Corresponding figures for diabetes were 17% (4/23) and nil (0/24). Plasma glucose concentrations at 30 and 120 minutes fell with increasing birth weight and weight at 1 year. Plasma 32-33 split proinsulin concentration fell with increasing weight at 1 year. All these trends were significant and independent of current body mass. Blood pressure was inversely related to birth weight and strongly related to plasma glucose and 32-33 split proinsulin concentrations.

• CONCLUSIONS: Reduced growth in early life is strongly linked with impaired glucose tolerance and non-insulin dependent diabetes. Reduced early growth is also related to a raised plasma concentration of 32-33 split proinsulin, which is interpreted as a sign of beta cell dysfunction. Reduced intrauterine growth is linked with high blood pressure, which may explain the association between hypertension and impaired glucose tolerance.

Questions 问题 (1)

• What is the study design ？本研究设计是？• Why OR? why not RR? 为什么用比值比（ O

R ）？而不是相对风险度 （ RR ） ？• Can we have crude OR and crude RR? 我们可

以计算粗 OR （未调整比值比） , 粗 RR （未调整相对风险度 ）吗？

• Why adjusted OR? Why not adjusted RR ? 为什么用调整 OR ？为什么不用调整 RR ？

• Do we need adjustments ? Always necessary? 需要调整吗？调整总是必要吗？

Questions 问题 (2)

• What the difference Between OR and RR? 比值比与相对危险度有什么区别？• What’s the difference between Crude OR vs.

Adjusted OR ? 粗 OR 与调整 OR 有什么区别？• What’s the difference between Crude RR vs.

Adjusted RR ? 粗 RR 和调整 RR 之间的区别是什么

• Is current BMI a confounding factor? 现在的体重指数是一个混杂因素吗？

*Odds ratio for two hour glucose concentration of >7.8 mmol/l adjusted for current body mass index.

出生体重（ lb) 糖耐量异常 n/N （ % ）

Adjusted *OR

Crude OR

CrudeRR

低于 5.5 lb (2.5kg) 8/20 (40%) 6.6 ? ?

5.5-6.4 (2.9 kg) 16/47 (34%) 4.8 ? ?

6.5-7.4 (3.4 kg) 32/104 (31%) 4.6 ? ?

7.5-8.4 (3.9 kg) 26/117 (22%) 2.6 ? ?

8.5-9.4 (4.2 kg) 7/54 (13%) 1.4 ? ?

9.5+ lb (4.3 kg +) 4/28 (14%) 1.0 1.0 1.0

Calculation of Crude OR and RR for impaired glucose tolerance from published data (based on Table 2 in Hales and Barker, BMJ 1991 paper)

Exposure Group No-exposure group Birth weight (Birth weight >=9.5 lb) Crude OR Crude RR Adjusted OR

Event Total n1-a Rate1 Event Total n0-c Rate0 (a/b)/(c/d) (a/n1)/(c/n0) *a n1 b a/n1 c n0 d c/n0 ad/bc an0/cn1

<5.5 lb 8 20 12 40% 4 28 24 14% 4.0 2.8 6.65.5-6.4 lb 16 47 31 34% 4 28 24 14% 3.1 2.4 4.86.5-7.4 lb 32 104 72 31% 4 28 24 14% 2.7 2.2 4.67.5-8.4 lb 26 117 91 22% 4 28 24 14% 1.7 1.6 2.68.5-9.4 lb 7 54 47 13% 4 28 24 14% 0.9 0.9 1.4*adjusted for current body mass index

Answers to Questions （ 1 ）

• What is the study design 研究设计是？ Retrospective cohort study 回顾性队列研究

• Why OR? why not RR? Either good for reasoning, the latter is more accurate in defining the relative risk disparity. 为什么用比值比（ OR ）？而不是相对风险度 （ RR ？ 推理任一均可， RR 在定义相对风险差距更准确。

• Can we have crude OR and crude RR? Yes, in cohort studies or RCT, you can calculate RR. 我们可以计算粗 OR （未调整比值比） , 粗 RR （未调整相对风险度 ）吗？ 是的，在队列研究或试验，可以计算出 RR 。

• Why adjusted OR? Why not adjusted RR ? Either is good for reasoning, it is easier to calculate the adjusted OR. 为什么是调整 OR ？ 而不是调整 RR ？推理任一均可，调整 OR 更容易计算 .

For cohort study data, you can use log binomial model, to obtain the adjusted RR.

队列研究的数据，您可以使用 log 二项式模型，得到调整后的 RR 。

How to calculate adjusted RR, in this study ?在本研究中， 如何计算调整后的 RR ？

Answers to Questions （ 2 ）

• Do we need adjustments in OR or RR? Most times, yes. 我们需要调整吗？大多数时候，是的。

• Is adjustments always necessary? No, sometimes unnecessary.调整是必要的吗？不，有时不必要。

• What the difference Between OR and RR? OR does not always represent RR. OR can be calculated in any study designs, RRs can not be calculated directly in case control studies. OR 和 RR有什么区别？ OR 有时不能代表 RR 。可以在任何临床流行病学研究设计中计算 OR 。在病例对照研究不能直接计算 RR 。

• What’s the difference between Crude OR vs. Adjusted OR ? The adjusted OR more often (but not always) represents the true association 。未调整 OR 或与调整 OR 区别是什么？调整 OR 更经常（但并不总是）代表真正的联系。

• What’s the difference between Crude RR vs. Adjusted RR ? 调整 RR 更经常（但并不总是）代表真正的联系。

真理？假象？ Causal Inference Considerations 因果推理思考

•Information bias? 信息偏倚 less likely•Confounding factors ？混杂因素 possibly•Consistency of association ? 联系的一致性 yes•Strength of association? 关联强度 OK•Dose-response relationship? 剂量 - 反应关系 yes•Temporally order consistent? 时间一致性 yes

•Deterministic / probabilistic? 决定性 no•Necessary? 必要性 no•Sufficient? 充分 no•Specificity? 特异性 no

•Biological plausibility? 生物合理性 yes•Surrogate risk factor? 替代风险因素 may be•Animal model experiment? 实验动物模型 yes

Birth weight 出生体重Crude

ORAdjusted*

OR

<5.5 lb

5.5-6.4 lb

6.5-7.4 lb

7.5-8.4 lb

4.0

3.1

2.7

1.7

6.6

4.8

4.6

2.6

ORs for impaired glucose toleranceAdjusted ORs > Crude ORsIs Current BMI a confounder?

*adjusted for current body mass index.

Confounder or Effect Mediator?混杂因素 , 或影响介质 ?

Glucose tolerance 糖耐量 Blood pressure 血压

Current BMI现体重指数

Birth weight 出生体重

Confounders 混杂因素 (e.g. ethnicity 如种族 )

When you inappropriate adjust for a factor in the causal pathway, you could produce a false association, or exaggerated association 当你不适当的调整一个在因果通路途径上的因素，你可能会产生一个虚假的关联，或夸张的关联。

It may be inappropriate to adjust for current BMI in estimating the effect of birth weight on current glucose tolerance or blood pressure. 调整现在的体重指数以估计出生体重对目前的糖耐量或血压的影响可能是不合适的。

Reversal paradox 逆转谜题 

Why evidence for the fetal origins of adult disease might be a statistical artifact: the "reversal paradox" for the relation between birth weight and blood pressure in later life. Tu YK, West R, Ellison GT, Gilthorpe MS. Am J Epidemiol;161(1):27-32.

Some researchers have recently questioned the validity of associations between birth weight and health in later life. They argue that these associations might be due in part to inappropriate statistical adjustment for variables on the causal pathway (such as current body size), which creates an artifactual statistical effect known as the "reversal paradox." Computer simulations were conducted for three hypothetical relations between birth weight and adult blood pressure. The authors examined the effect of statistically adjusting for different correlations between current weight and birth weight and between current weight and adult blood pressure to assess their impact on associations between birth weight and blood pressure. When there was no genuine relation between birth weight and blood pressure, adjustment for current weight created an inverse association whose size depended on the magnitude of the positive correlations between current weight and birth weight and between current weight and blood pressure. When there was a genuine inverse relation between birth weight and blood pressure, the association was exaggerated following adjustment for current weight, whereas a positive relation between birth weight and blood pressure could be reversed after adjusting for current weight. Thus, researchers must consider the reversal paradox when adjusting for variables that lie within causal pathways.

Surrogate risk factors 替代风险因素 ?

Glucose tolerance糖耐量

Birth weight 出生体重

Surrogate risk factors

替代风险因素

Shared genetic variants cause both LBW and impaired glucose tolerance ?

Glucose tolerance糖耐量

Birth weight 出生体重

Genetic variants遗传变异

Causal Mechanisms /pathways? epienetic changes ， etc.

Glucose tolerance糖耐量

Birth weight 出生体重

Epigenetic changes, etc.表观遗传改变，等 .

Intrauterine environmentGlucocorticoids, hormones ， etc糖皮质激素 , 激素 等

Unknown confounders未知的混杂因素

Hales CN, Barker DJ. The thrifty phenotype hypothesis. Br Med Bull. 2001;60:5-20

The thrifty phenotype hypothesis proposes that the epidemiological associations between poor fetal and infant growth and the subsequent development of type 2 diabetes and the metabolic syndrome result from the effects of poor nutrition in early life, which produces permanent changes in glucose-insulin metabolism. These changes include reduced capacity for insulin secretion and insulin resistance which, combined with effects of obesity, ageing and physical inactivity, are the most important factors in determining type 2 diabetes. Since the hypothesis was proposed, many studies world-wide have confirmed the initial epidemiological evidence, although the strength of the relationships has varied from one study to another. The relationship with insulin resistance is clear at all ages studied. Less clear is the relationship with insulin secretion. The relative contribution of genes and environment to these relationships remains a matter of debate. The contributions of maternal hyperglycaemia and the trajectory of postnatal growth need to be clarified.

Project - Maternal glucose tolerance, oxidative stress, and programming of the Metabolic, CIHR Funded 2006-2010

Follow-up of Women/Infants at Delivery (n=308)

Follow-up of infants at 1 year of age (n=242)

Biomarker Assays

Follow-up of infants at 3 months of age (n=281)

Follow-up of women at 32-35 weeks of gestation (n=320)

Recruitment - Women bearing a singleton fetusat 24-28 weeks of gestation (n=339) (Montreal)

Data Analyses and Reports

Luo et al. Diabetes Care 2010

Maternal Glucose Tolerance in Pregnancy Affects Fetal Insulin Sensitivity

Luo et al. Diabetes Care 2010

Maternal BMI was also inversely correlated Fetal Insulin Sensitivity, but less strongly so

IGF-1 (but not IGF-2) levels in maternal and fetal circulations were elevated in gestational

diabetes GD = gestational diabetic

35.524.3

45.2 36.8

9.67.2

1.0

10.0

100.0

1000.0

GD Non-GD GD Non-GD GD Non-GDMaternal blood Maternal blood Cord

24-28 weeks 32-35 weeks blood

Pla

sm

a I

GF

-1 (

nm

ol/

l)

Luo ZC, et al. J Clinical Endocrinology Metabolism 2012

Higher maternal IGF-1 (not IGF-2) levels predict increased risk of LGA/macrosomia

Luo ZC, et al. J Cinical Endocrinology and Metabolism 2012

Leptin and adiponectin levelswere positively correlated

in maternal versus fetal circulations

Luo ZC, et al. Obesity 2013

Fetal insulin sensitivity was negatively associated with cord blood leptin

(p<0.01) but not adiponectin (p>0.4) levels.

Luo ZC, et al. Obesity 2013

Summary 总结 (1) OR can be calculated in any clinical epidemiologic

studies (cross-sectional, case control, cohort), but OR may “overstate” the effect size when the outcome is common. OR 可以在任何临床流行病学研究在计算（横截面，病例对照，队列研究，临床试验）， OR 或许“夸大”效应时，如果结果是常见的。

Adjusted OR sometimes may distort the true association, if the adjusted “confounding factor” is not a true confounding factor, but an effect mediator in the causal pathway 。调整有时可能会扭曲真正的关联，如果调整后的“混杂因素”不是真正的混杂因素，而是一个在因果通路上的中介因素。

Summary 总结 (2) Be careful with the adjustment factors in regression

models; some factors should not be included in the adjustments. 在回归模型中，要注意小心调整因素。有些因素不应包括在调整模型中。

RR is a better measure of relative risk disparity, and can be calculated in any cohort studies or RCTs. RR是一个更好的衡量相对风险差异指标，可以在任何队列研究或随机对照试验中计算。

It is important to know what’s key aspect is not known and needs to be known in the study area in designing your own clinical epidemiologic study 。在设计你自己的临床流行病学研究项目时， 重要的是要知道在该研究领域， 什么关键方面有什么未知的需要解决的问题。