统计问题第135问：病例对照研究的偏倚

Question

Researchers explored the association between maternal intake of folic acid supplements, multivitamins, and dietary folates in the prevention of cleft lip (with or without cleft palate) in the mothers’ babies. A population based case-control study was performed. All babies with cleft lip referred for corrective surgery in Norway between 1996 and 2001 were identified. For each case—a baby with a cleft lip—a maximum of two live births without facial clefts were randomly selected from the Norwegian medical birth registry for 1996 to 2001 to serve as controls. Controls were matched to each case by date of birth to within six weeks of delivery.

The mothers of all cases and controls were invited to take part in the study. If women agreed, they completed questionnaires several months after delivery, providing demographic characteristics, reproductive history, and details about smoking, alcohol and drug consumption, and other exposures during early pregnancy. Women were also asked to report whether they took folic acid supplements and, if so, the dosage consumed for each of the six months preceding pregnancy and each of the first three months of pregnancy. Similar information was also provided about multivitamins. Women were asked to recall their diet during the first three months of pregnancy by using a food frequency questionnaire so that dietary folates could be estimated.

The researchers reported that folic acid supplements in early pregnancy reduced the risk of isolated cleft lip (with or without cleft palate) by about a third. Other vitamins and dietary factors may also have provided additional benefit.

Which of the following types of bias, if any, may the above study have been prone to?

·a) Allocation bias

·b) Selection bias

·c) Recall bias

·d) Confounding

提示：正确答案不止一个。

Answer

Answers c and d are true, while a and b are false.

The aim of the study was to examine whether maternal intake of folic acid supplements, dietary folates, or multivitamins before and during pregnancy was associated with a reduced risk of cleft lip (with or without cleft palate). In a case-control study the participants are selected on the basis of their disease or condition. Babies born with a cleft lip were the cases, while those babies not born with a cleft lip were the controls. Babies were not allocated to a group, and therefore allocation bias did not occur (a is false). Allocation bias is the systematic difference between participants in allocation to treatment group and will occur in clinical trials if random allocation is not used.

Although case-control studies are typically prone to selection bias, it is unlikely to have existed in the above study (b is false). Selection bias would have occurred if those women invited to take part were not representative of the population of all possible mothers. The above study was a population based case-control study, with all babies born with a cleft lip in Norway between 1996 and 2001 identified as cases and with a sample of controls selected at random from the Norwegian medical birth registry for the same period. Therefore, those mothers invited to participate should have been representative of all mothers of possible cases and controls in the population. Selection bias must not be confused with non-response bias. Non-response bias would have occurred if after the invitation to participate in the questionnaire survey there was a difference for either cases or controls in the sociodemographic characteristics, behaviour, or attitudes between those mothers who agreed and those who refused to participate.

Mothers were asked to report their intake of folic acid supplements, dietary folates, and multivitamins before and during pregnancy. The recall of information was about past behaviour that occurred more than 15 months previously and therefore was likely to be inaccurate. Recall bias would have occurred if there was a systematic difference between mothers of the cases and mothers of the controls in the accuracy of the information that they recalled. The above study was prone to recall bias, it being a particular problem in case-control studies (c is true).The inaccuracy of the information provided by the mothers of the cases or the controls will have been amplified because of some association with the condition of cleft lip. It is possible that those mothers who delivered a child with a cleft lip (cases) over-reported their folic acid intake because of a sense of guilt through not doing more during pregnancy to protect their unborn child.

Babies were not randomised to outcome group (cases or controls). It is unlikely that the mothers of the two groups were similar in their demographic characteristics, reproductive history, and history of smoking, alcohol and drug use, and other exposures during early pregnancy. Therefore, the study was prone to confounding (d is true). A confounding variable or confounder would be related to the dependent variable—cleft lip—and also to the explanatory variable—folic acid supplement consumption. Alcohol consumption during early pregnancy is a risk factor for cleft lip and therefore may confound the relationship between folic acid supplementation and cleft lip. If alcohol consumption is a confounder it will be associated with the use of folic acid supplements: for example, mothers taking folic acid supplements might be less likely to drink alcohol during early pregnancy than those mothers not taking the supplements. This relationship is associative and not causative—that is, it simply exists as a result of the characteristics or behaviours of the mothers. Therefore, any increased risk of cleft lip in babies of women who didn’t take folic acid supplements may have been due not to the lack of supplements but to the babies’ greater exposure to alcohol in early pregnancy when compared with babies of women who took folic acid supplements.

所以答案是选择  c d

每天学习一点，你会更强大！