A secondary analysis was performed on data from the longitudinal cohort Pain After Delivery (PAD) study.16
Institutional Review Board approval was obtained for the original study, and the present secondary analysis was considered exempt. Parturients were recruited from September 2004 to December 2005 at four university hospitals (Winston-Salem, NC, USA; New York, NY, USA; Geneva, Switzerland; Brussels, Belgium) and provided written informed consent. All parturients hospitalized for cesarean or vaginal delivery at each hospital were invited to participate, except during weekends and holidays when research personnel were unavailable. These women were queried within 72 hours after delivery about acute pain after delivery, pain history before and during pregnancy, other medical history, and demographic information (Entrance Questionnaire, Appendix
). At the two U.S. hospitals, post-delivery questions were administered by research personnel in the form of a scripted interview; patients at the European sites completed the questionnaire in written format. Patient medical records were also reviewed by on-site research personnel, who directly transcribed data pertaining to age, body mass index, and gestational age. Women from the two U.S. hospitals were contacted 8 weeks later and interviewed by telephone using a computer-assisted scripted interview that inquired about persistent pain.
Headache prior to pregnancy was identified from several responses to the Entrance Questionnaire. Women who either indicated they had “chronic pain that started before this pregnancy” (question 4) or were “seeing a doctor regularly for any medical problems in the year before your pregnancy” (question 8) and listed head, headache, or migraine as the pain location or condition (questions 4a, 8b, and 8b1) were considered to have headache prior to pregnancy. Women who answered “yes” to “taking daily pain medications before this pregnancy” (question 5) or “taking any medications regularly (daily or > 3 days a week) during the year before pregnancy” (question 9) and listed medications exclusively used for migraine (e.g., a triptan) (questions 5a, 9a, and 9b) also were considered to have headache prior to pregnancy.
The principal outcome measures were headache during pregnancy, headache within 72 hours after delivery, and headache at 8 weeks after delivery. Headache during pregnancy was identified from questions 15 and 16 of the Entrance Questionnaire. Women were considered to have headache during pregnancy if they identified “headache” as the location of pain during pregnancy (question 15d) or reported taking regular medications during pregnancy for headache or migraine (questions 16a and 16b). Headache within 72 hours of delivery was identified from endorsements of “headache” as the site of current pain subsequent to delivery (question 24f). Women were considered to have headache at 8 weeks after delivery if they identified headache or migraine as an ongoing health problem or were currently taking a headache or migraine medication at the 8-week follow-up phone interview.
In addition to headache variables, the Entrance Questionnaire assessed many demographic and delivery characteristics. This information included self-reported general state of health, presence of chronic pain before pregnancy, use of daily pain medications before pregnancy, presence of chronic pain in family members, level of menstrual pain, regular doctor visits for medical problems before pregnancy, use of medications regularly (> 3 times per week) before pregnancy, smoking more than 1 pack of cigarettes a week in the year before pregnancy, level of alcohol consumption in the year before pregnancy, pain from previous operations, pain level during pregnancy, regular medications during pregnancy, and prolonged pain after previous birth. The questionnaire also included a somatization scale (questions 3a-j).17
Data on receipt of regional anesthesia injections were obtained from a separate Delivery Information Questionnaire (as described in Eisenach et al., 200816
This is a secondary analysis of the Pain After Delivery (PAD) study, and a description of the primary analyses, including the original statistical power considerations, has been reported previously.16
Due to the secondary nature of this analysis (i.e., the study was not originally powered to examine these hypotheses), care was taken to report and interpret point estimates of effect size with 95% confidence intervals. Descriptive statistics are presented as mean (SD) for normally distributed scaled data and frequency counts with percentages for categorical data.
To examine the associations between patient characteristics and the presence of headache during pregnancy, a multivariable logistic regression model was created. The model was specified based on forced entry of this set of a priori selected predictors. The model controlled for study site, and care was taken to consider the correlations among predictors prior to analysis (i.e., none of the predictors were so highly correlated as to present estimation problems in the analysis). To predict the presence of headache within 72 hours after delivery, a multivariable Cox proportional-hazards model was created. Based on the modest number of parturients who experienced headache after delivery (n = 92), care was taken to specify a model that employed a suitably limited number of predictors (to avoid overfitting the data). The proportional-hazards assumption was tested and satisfied prior to interpretation of the model.
To express the relative risks of the various predictors, odds ratios (logistic regression model), hazard ratios (Cox proportional-hazards model), and risk ratios all with respective 95% confidence intervals are reported. The RR was chosen to express the relative risk of experiencing headache at one measurement occasion conditional on previous occasions because of the relative frequency of the observed event rates (odds ratios and risk ratios are only similar for rare events) and ease of interpretation that this metric provides. For example, a RR of 3.0 indicates that experiencing a headache is three times more likely in one group versus another. All analyses were conducted using SAS 9.2 (SAS, Inc. Cary, NC). Where appropriate, all hypotheses are two-tailed with p < 0.05.