Background Disease control (DC) and spouse control (SC) are commonly used in case control study. The exposure level of these two kinds of control may be different, especially in smoking status. However, few studies concentrate on the association of DC and SC. Based on this situation, this study aims to explore the quantitative relationship of Odds Ratio (OR) between DC (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\({OR}_{dc}\)</EquationSource> </InlineEquation>) and SC (<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\({OR}_{sc})\)</EquationSource> </InlineEquation> for lung cancer. Methods In this study, the data of cases, DC and SC were collected from the National Retrospective Survey of Deaths, including one million deaths during 1986 to 1988. Referring to cross-walk model, four forms of model, stratified by urban/rural areas, genders and 5-year age categories, were established to explore the suitable one, for adjusting <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\({OR}_{dc}\)</EquationSource> </InlineEquation> and <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\({OR}_{sc}\)</EquationSource> </InlineEquation>. Results After matching, a total of 57,789 participants were collected. Accordingly, the smoking prevalence of case, SC and DC groups were 54.39%, 31.72%, 33.54%, respectively. In the urban areas, the total <InlineEquation ID="IEq5"> <EquationSource Format="TEX">\({OR}_{sc}\)</EquationSource> </InlineEquation> of males and females were higher than <InlineEquation ID="IEq6"> <EquationSource Format="TEX">\({OR}_{dc}\)</EquationSource> </InlineEquation>, while the quantitative relation was the opposite in the rural areas. The <InlineEquation ID="IEq7"> <EquationSource Format="TEX">\({OR}_{sc}\)</EquationSource> </InlineEquation> and <InlineEquation ID="IEq8"> <EquationSource Format="TEX">\({OR}_{dc}\)</EquationSource> </InlineEquation> value gradually increases with the age group, reaching its maximum at 60–64 in urban area and 50–59 in rural area. After establishing the cross-walk model, in urban area, <InlineEquation ID="IEq9"> <EquationSource Format="TEX">\({logit(OR}_{sc})=0.98logit({OR}_{dc})\)</EquationSource> </InlineEquation> and <InlineEquation ID="IEq10"> <EquationSource Format="TEX">\({logit(OR}_{sc})=-0.18+1.52logit({OR}_{dc})\)</EquationSource> </InlineEquation> were used for males and females, respectively. In rural area, <InlineEquation ID="IEq11"> <EquationSource Format="TEX">\({logit(OR}_{sc})=0.56+0.29logit({OR}_{dc})\)</EquationSource> </InlineEquation> and <InlineEquation ID="IEq12"> <EquationSource Format="TEX">\({logit(OR}_{sc})=0.60logit({OR}_{dc})\)</EquationSource> </InlineEquation> were used for males and females, respectively. Conclusions <InlineEquation ID="IEq13"> <EquationSource Format="TEX">\({OR}_{sc}\)</EquationSource> </InlineEquation> were commonly larger than <InlineEquation ID="IEq14"> <EquationSource Format="TEX">\({OR}_{dc}\)</EquationSource> </InlineEquation> in urban areas, whereas the opposite quantitative relation was observed in rural areas. Any modeling of the relationship between <InlineEquation ID="IEq15"> <EquationSource Format="TEX">\({OR}_{sc}\)</EquationSource> </InlineEquation> and <InlineEquation ID="IEq16"> <EquationSource Format="TEX">\({OR}_{dc}\)</EquationSource> </InlineEquation> is presented solely as a descriptive exercise for this historical dataset and should not be interpreted as a validated adjustment tool or a generally applicable method for translating effect estimates between control types. As an exploratory, proof-of-concept analysis, this study illustrates how effect estimates may vary across control definitions in case-control studies, highlighting a fundamental methodological caution.

Quantitative relationship of effect estimates from spouse and disease controls in case control study

期刊论文
出版模式：

开放获取
发表日期：

2026年4月9日

查看全文

Zemin Cai,
Xiao Zhang,
Xiaonong Zou,
Xia Wan

摘要

Background

Disease control (DC) and spouse control (SC) are commonly used in case control study. The exposure level of these two kinds of control may be different, especially in smoking status. However, few studies concentrate on the association of DC and SC. Based on this situation, this study aims to explore the quantitative relationship of Odds Ratio (OR) between DC ( \({OR}_{dc}\) ) and SC ( \({OR}_{sc})\) for lung cancer.

Methods

In this study, the data of cases, DC and SC were collected from the National Retrospective Survey of Deaths, including one million deaths during 1986 to 1988. Referring to cross-walk model, four forms of model, stratified by urban/rural areas, genders and 5-year age categories, were established to explore the suitable one, for adjusting \({OR}_{dc}\) and \({OR}_{sc}\) .

Results

After matching, a total of 57,789 participants were collected. Accordingly, the smoking prevalence of case, SC and DC groups were 54.39%, 31.72%, 33.54%, respectively. In the urban areas, the total \({OR}_{sc}\) of males and females were higher than \({OR}_{dc}\) , while the quantitative relation was the opposite in the rural areas. The \({OR}_{sc}\) and \({OR}_{dc}\) value gradually increases with the age group, reaching its maximum at 60–64 in urban area and 50–59 in rural area. After establishing the cross-walk model, in urban area, \({logit(OR}_{sc})=0.98logit({OR}_{dc})\) and \({logit(OR}_{sc})=-0.18+1.52logit({OR}_{dc})\) were used for males and females, respectively. In rural area, \({logit(OR}_{sc})=0.56+0.29logit({OR}_{dc})\) and \({logit(OR}_{sc})=0.60logit({OR}_{dc})\) were used for males and females, respectively.

Conclusions

\({OR}_{sc}\) were commonly larger than \({OR}_{dc}\) in urban areas, whereas the opposite quantitative relation was observed in rural areas. Any modeling of the relationship between \({OR}_{sc}\) and \({OR}_{dc}\) is presented solely as a descriptive exercise for this historical dataset and should not be interpreted as a validated adjustment tool or a generally applicable method for translating effect estimates between control types. As an exploratory, proof-of-concept analysis, this study illustrates how effect estimates may vary across control definitions in case-control studies, highlighting a fundamental methodological caution.