Examples of Study Design Choices That Can Affect Data Comparability Among Selected US Studies of Menarche

StudyStudy Design TypeQuestionnaire Data Used in Analysis (Respondent)Calculation/AnalysisAge Increment ReportedAge Range of Population, ya
Fels Institute Study; Reynolds and Wines13Longitudinal“Menstrual record” (NR)MeanMidpoint between visits (ie, half-year)8-18
Guidance Study; NicolsonLongitudinalNR (NR)MeanMidpoint between visits8-17
and Hanley14 (1953)
NHES I, II, and III;CS; nationallyI: recall (self); II and III:I: mean; II and III: median/I: year; II and III:1:18-79; II and III:
MacMahon63 (1973),representative,current status and recallfitted logistic curve;midpoint ofthe year11-14; 8.6-
Lee etal173 (1980)longitudinal(parent); “menstrual history” (self)mean(added 0.5 years); half-year17.8
Bogalusa Heart Study;CS (1 cohort)bRecall (self)MedianYear5-14
Foster etal61 (1977)
Pediatric Research in Office Settings (PROS) Network; Herman-CS; well-child or problem visit at volunteer physicians officeCurrent status (self)Means/probit analysisYear3-12
Giddens etal5 (1997)
Bogalusa Heart Study; Wattigney et al93 (1999)Mixed longitudinal/CS (2 nonoverlapping CS cohorts)bLimited to postmenarcheal girls; recall (self)Mean and analysis of variance to assess differences between 2 cohortsEarly (8-11 years); intermediate (12-13 years); late (14-17 years)8-17
Bogalusa Heart Study; Freedman et al4 (2002)Mixed longitudinal/CS (7 overlapping CS cohorts)bCurrent status (self)Median/logistic regressionMidpoint of yearThird grade (∼8) to 17
NHANES III; Wu etal47 (2002)CS; nationally representativeCurrent status (self); recall (self)Current status data: mean/ probit analysis; recall data: mean/failure time modelCurrent status: year; postmenarcheal girls: midpoint of year (added 0.5 years)10-16
NHANES III; Chumlea et al46 (2003)CS; nationally representativeCurrent status (parent for 8- to 9-year-olds; self for 10- to 19-year-olds)Median/probit analysisMidpointofa 3-mo age group8-19
  • NR indicates not reported; CS, cross-sectional.

  • a Year ranges described in whole numbers indicate the beginning ofthe first age to the end ofthe second age (eg, 3-12 is 3.0years to 12.99yearsofage).

  • b The Bogalusa Heart Study used an overlapping panel design.There were 7 CSexaminationsofchildren (orcohorts) in a specific age range,which were timed closelytogether so that some ofthe participants were examined multiple times (ie, longitudinally). Freedman et al4 analyzed data from all 7 CS cohorts; Wattigneyetal93 analyzed data from 2 nonoverlapping CS cohorts; Foster et al61 analyzed data from 1 CS cohort.