Hawley W, Nuovo J, DeNeef C P, Carter P
Department of Family Medicine, University of Washington School of Medicine, Seattle.
J Am Board Fam Pract. 1993 Mar-Apr;6(2):123-35.
We designed a study to assess the association of oral contraceptive use and the development of breast cancer for women in the following groups: (1) ever oral contraceptive users, (2) long-term oral contraceptive users, and (3) oral contraceptive users before a first full-term pregnancy.
A MEDLINE search of studies published in English from 1966 to 1990 was conducted using the following key words: "oral contraceptive and breast carcinoma." Eligible studies included all published case-control reports of nonduplicated data on a population (hospital or community-based). The following data were extracted from each report: country, age of subjects, number of cases and controls, whether it was a hospital or community-based study, and results. Two evaluators using a quality-assessment instrument independently and blindly reviewed the methods and data analysis section from each eligible study. In the category of "ever oral contraceptive users," an estimate of the pooled relative risk with 95 percent confidence intervals (CIs) was calculated. In the categories of "duration of oral contraceptive use" and "duration of oral contraceptive use before a first full-term pregnancy," Spearman's rank correlation coefficient (rs) was calculated.
For the categories of "ever oral contraceptive users" and "long-term oral contraceptive users," no association between the use of oral contraceptives and the development of breast cancer could be detected (pooled relative risk "ever oral contraceptive users" = 1.07, 95 percent CI 0.78 to 1.36, rs "duration of use" -0.153, P = 0.189). For the category of "oral contraceptive use before a first full-term pregnancy," a significant correlation was found (rs = +0.497, P = 0.011). Many reports failed to demonstrate adequate protection against the biases most relevant to case-control methods (namely, recall bias, interviewer bias, surveillance bias, and nonresponse bias) and therefore received low-quality scores.
This meta-analysis suggests a possible increased risk for breast cancer in women who use oral contraceptives before a first full-term pregnancy. The data, however, are confounded by studies that are generally of low quality. Further studies addressing the risk for breast cancer in oral contraceptive users need to be designed with methods that limit the biases inherent in case-control studies.
我们设计了一项研究，以评估以下几组女性口服避孕药的使用与乳腺癌发病之间的关联：（1）曾经使用过口服避孕药的女性；（2）长期使用口服避孕药的女性；（3）首次足月妊娠前使用口服避孕药的女性。
利用关键词“口服避孕药与乳腺癌”对1966年至1990年以英文发表的研究进行了MEDLINE检索。符合条件的研究包括所有已发表的关于人群（基于医院或社区）的非重复数据的病例对照报告。从每份报告中提取以下数据：国家、受试者年龄、病例数和对照数、研究是基于医院还是社区以及结果。两名评估人员使用质量评估工具，独立且盲法地审查了每项符合条件研究的方法和数据分析部分。在“曾经使用过口服避孕药的女性”类别中，计算了合并相对风险及95%置信区间（CI）的估计值。在“口服避孕药使用时长”和“首次足月妊娠前口服避孕药使用时长”类别中，计算了斯皮尔曼等级相关系数（rs）。
在“曾经使用过口服避孕药的女性”和“长期使用口服避孕药的女性”类别中，未检测到口服避孕药的使用与乳腺癌发病之间存在关联（“曾经使用过口服避孕药的女性”的合并相对风险 = 1.07，95%CI为0.78至1.36，“使用时长”的rs = -0.153，P = 0.189）。在“首次足月妊娠前使用口服避孕药”类别中，发现存在显著相关性（rs = +0.497，P = 0.011）。许多报告未能充分证明可抵御与病例对照方法最相关的偏倚（即回忆偏倚、访谈者偏倚、监测偏倚和无应答偏倚），因此获得了低质量评分。
这项荟萃分析表明，首次足月妊娠前使用口服避孕药的女性患乳腺癌的风险可能会增加。然而，这些数据因普遍质量较低的研究而受到混淆。需要设计进一步的研究来探讨口服避孕药使用者患乳腺癌的风险，采用能够限制病例对照研究固有偏倚的方法。
