Outpatient calcium-channel blockers and the risk of postpartum haemorrhage: a cohort study

Authors

  • BT Bateman,

    Corresponding author
    1. Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA
    2. Department of Anesthesiology, Critical Care, and Pain Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA
    • Correspondence: Dr BT Bateman, Brigham & Women's Hospital, Department of Medicine, Division of Pharmacoepidemiology & Pharmacoeconomics, 1620 Tremont Street, Suite 3030, Boston, MA 02120, USA. Email bbateman@partners.org

    Search for more papers by this author
  • S Hernandez-Diaz,

    1. Department of Epidemiology, Harvard School of Public Health, Boston, MA, USA
    Search for more papers by this author
  • KF Huybrechts,

    1. Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA
    Search for more papers by this author
  • K Palmsten,

    1. Department of Epidemiology, Harvard School of Public Health, Boston, MA, USA
    Search for more papers by this author
  • H Mogun,

    1. Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA
    Search for more papers by this author
  • JL Ecker,

    1. Department of Obstetrics, Gynecology, and Reproductive Sciences, Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA
    Search for more papers by this author
  • EW Seely,

    1. Department of Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, USA
    Search for more papers by this author
  • MA Fischer

    1. Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA
    Search for more papers by this author

Abstract

Objective

To determine whether outpatient exposure to calcium-channel blockers (CCBs) at the time of delivery is associated with an increased risk for postpartum haemorrhage (PPH).

Design

Cohort study.

Setting

United States of America.

Population or sample

Medicaid beneficiaries.

Methods

We identified a cohort of 9750 patients with outpatient prescriptions for CCBs, methyldopa, or labetalol for pre-existing or gestational hypertension whose days of supply overlapped with delivery; 1226 were exposed to CCBs. The risk of PPH was compared in those exposed to CCBs to those exposed to methyldopa or labetalol. Propensity score matching and stratification were used to address potential confounding.

Main outcome measures

The occurrence of PPH during the delivery hospitalisation.

Results

There were 27 patients exposed to CCBs (2.2%) and 232 patients exposed to methyldopa or labetalol (2.7%) who experienced PPH. After accounting for confounders, there was no meaningful association between CCB exposure and PPH in the propensity score matched (odds ratio 0.77, 95% CI 0.50–1.18) or stratified (odds ratio 0.79, 95% CI 0.53–1.19) analyses. Similar results were obtained across multiple sensitivity analyses.

Conclusions

The outpatient use of CCBs in late pregnancy for the treatment of hypertension does not increase the risk of PPH.

Introduction

Postpartum haemorrhage (PPH) is a leading cause of maternal morbidity and mortality in both the developed and developing world.[1-6] The incidence of PPH has increased substantially over the past decade in developed nations, primarily as the result of an increase in the occurrence of uterine atony.[3, 7-13] Consequently, identifying modifiable risk factors for PPH in general, and uterine atony in particular, is urgently needed.

Calcium-channel blockers (CCBs) are frequently prescribed during the third trimester of pregnancy for both hypertension[14] and, off-label, for labour tocolysis;[15, 16] recent data from the USA suggest that between 1 and 2% of all patients are exposed to CCBs as outpatients during the third-trimester.[17, 18] As the incidence of hypertensive disorders during pregnancy increases,[19, 20] exposure to CCBs is likely to become increasingly common.

Calcium-channel blockers inhibit the influx of calcium into uterine smooth muscle and prevent uterine contraction[21]; this is the basis for their use in prolonging pregnancy in patients with preterm labour. While this property is useful in the setting of preterm labour, following delivery, effective uterine contraction is required to compress the uterine vasculature and stem the flow of blood to the myometrium.[22] In the absence of effective contraction, the atonic uterus can bleed excessively, resulting in PPH. Because of the tocolytic properties of CCBs, several authors have suggested that exposure to CCBs close to the time of delivery may increase the risk of uterine atony and PPH.[23, 24] To date, however, there are few data examining this risk.

The objective of this study is therefore to examine whether outpatient exposure to CCBs at the time of delivery is associated with an increased risk of PPH. To avoid the potential for confounding by indication[25] in this observational study, we focus this study on CCBs prescribed for hypertension and employ an active comparator control group of patients exposed to methyldopa or labetalol, which are first-line agents for the treatment of hypertension in pregnancy[26-28] and which, to the best of our knowledge, have not been implicated in causing PPH.

Methods

Cohort

The cohort was derived from the Medicaid Analytic eXtract (MAX), an administrative data set that contains information on Medicaid enrolment and utilisation claims. The data set contains claims for inpatient admissions and outpatient visits, as well as outpatient pharmacy dispensing claims. Using MAX data from 2000 to 2007, a cohort was created for the study of drug use and safety in pregnancy.[29] Briefly, women with an inpatient or outpatient claim for a delivery were linked to infants within states using the Medicaid Case Number (which is generally shared by families) and the infant's date of birth. Because neither gestational age at birth nor the date of a woman's last menstrual period (LMP) is directly recorded in the MAX, the LMP was assigned using a validated algorithm that is based on diagnostic codes in the maternal and infant records.[30]

We then restricted the cohort to women who were eligible for Medicaid continuously from 6 months after the estimated LMP month through the delivery month to allow accurate capture of maternal medical and obstetric conditions important in the planned analysis. To ensure complete ascertainment of relevant claims, we restricted our analysis to women with ≥28 days of enrolment each month, and without restricted benefits, private insurance, or certain state-specific managed care programmes (that under-report claims to MAX). The source cohort included 3 622 489 completed pregnancies.

We then restricted the cohort to those patients that were dispensed as outpatients either a CCB, labetalol, or methyldopa and whose days of supply overlapped with the date of delivery. Because we hypothesise that the risk of PPH associated with CCBs derives from direct action of the CCB on the myometrium following delivery, the aetiologically relevant window for exposure is the days immediately before delivery. A list of individual medications that comprise the CCBs are included in the Supplementary material, Appendix S1.

Because confounding by indication is a major threat to the validity of nonrandomised studies,[25] we selected as a reference group of patients exposed to methyldopa or labetalol, which are considered the preferred treatment for hypertension in pregnancy by many guidelines.[26-28] To ensure the comparability of exposed and reference groups, for the primary analysis we restricted our analysis to those patients who had a diagnosis code for gestational or pre-existing hypertension and further excluded patients with diagnoses other than hypertension that are indications for the prescription of CCBs including preterm labour/threatened labour, preterm delivery, migraine, arrhythmia, angina and Raynaud syndrome. We also exclude patients exposed to methyldopa or labetalol in addition to CCBs. Finally, to ensure accurate capture of the outcome of PPHs, we excluded patients that did not deliver in the hospital. After these restrictions and exclusions, the primary cohort included 9750 pregnancies (see Figure 1).

Figure 1.

Patient flow chart.

Study outcomes

The primary study outcome was the occurrence of PPH during the delivery hospitalisation. This was defined based on a diagnostic code of 666 or any subcode thereof, from the International Classification of Diseases, 9th revision (ICD 9). Diagnostic codes have been shown to have a high positive predictive value for PPH in administrative data.[31, 32] As a secondary outcome, we also examined PPH due to uterine atony, as denoted by ICD 9 code 666.1 or any subcode thereof, which mechanistically is the form of PPH most likely to be influenced by CCBs.

Covariates

We identified five groups of potential confounders to the planned analysis: patient demographics, medical characteristics, obstetric conditions, medications (which can act as markers of disease and/or disease severity), and measures of medical care use during the third trimester; these were selected because they may influence the choice of antihypertensive medication and the risk of PPH. Demographic characteristics analysed included age and race/ethnicity as reported in the MAX. Maternal medical conditions were identified on the basis of diagnosis codes recorded in patients' inpatient or outpatient claims at any time during the third trimester and included pre-existing hypertension, pregestational diabetes, obesity, chronic renal disease and fibroids. Maternal obstetric conditions/procedures were identified in the same manner and included gestational hypertension, mild pre-eclampsia (including unspecified hypertensive disorders), severe pre-eclampsia, gestational diabetes, previous caesarean delivery, placenta praevia, multiple gestations, caesarean delivery and induction of labour. Note, while some of these conditions/procedures (e.g. caesarean delivery and induction of labour) will by definition occur after the patient was prescribed the antihypertensive, they may be markers of the severity of important confounding variables (i.e. induction of labour may be a surrogate for severe hypertension) and are therefore appropriate to consider as potential confounders in the analysis. However, to ensure that this approach did not introduce selection bias we performed a sensitivity analysis excluding these conditions/procedures in the model used to estimate the propensity score (see below).[33]

We also examined measures of medical care use during the third trimester which may also serve as markers of comorbid disease/disease severity including number of hospitalisations (including the delivery hospitalisation), number of outpatient visits, and number of distinct prescription medications (other than antihypertensives) dispensed.[34]

Statistical analysis

To adjust for differences between patients treated with CCBs and methyldopa/labetalol, we used two different approaches based on propensity scores. For both of these analyses, a propensity score was estimated using logistic regression with CCB exposure as the dependent variable and all potential confounders described above included without further selection. In the first approach, patients exposed to CCBs were matched on propensity score to those exposed to methyldopa or labetalol in a fixed 1:3 ratio using a nearest neighbour algorithm with a caliper of 0.05 difference in propensity score.[35] This resulted in a matched cohort containing 4900 patients. The odds ratio and 95% confidence interval for PPH or PPH with atony associated with CCB exposure was then estimated directly. In the second approach, all patients with a propensity score value that corresponded to the 2.5 centile or lower of propensity score distribution in the CCB exposed and the 97.5 centile or higher of the propensity score distribution in the methyldopa/labetalol exposed were identified and excluded from this analysis. Trimming subjects from the tails of the propensity score distribution has been demonstrated to provide an additional reduction in residual confounding.[36] In the remaining cohort, propensity score deciles were defined. A logistic regression model was then fitted adjusting to deciles of propensity score as a categorical covariate in the model and the odds ratio and 95% CI for PPH and PPH with atony associated with CCB exposure was determined.

Sensitivity analyses

Four-sensitivity analyses were performed to test the robustness of our findings. First, as the sensitivity of diagnostic codes for hypertensive disorders is imperfect, we repeated our analysis without requiring that patients have a code for pre-existing or gestational hypertension (n = 12 298). This was done to ensure that excluding patients without these diagnostic codes did not introduce bias to the analysis. Second, we performed a sensitivity analysis restricted to only those patients that refilled their prescription for CCBs or methyldopa/labetalol at least once during the third trimester (n = 7230). This was intended to reduce misclassification of exposure due to noncompliance based on the assumption that if the patient refilled their prescription at a time close to delivery it is likely that they were taking their antihypertensive medication as prescribed. Third, we performed an analysis that did not include clinical events or conditions that would have developed after the prescription of the antihypertensive in the model used to estimate the propensity score to ensure that we were not adjusting for intermediates in our analysis; these covariates included mild or severe pre-eclampsia, caesarean delivery, induction of labour and the number of hospitalisations or outpatient visits during the third trimester. Fourth, we included the entire cohort before any exclusions or restrictions (= 3 622 489). Exposure was defined by prescription for CCB whose days of supply overlapped with delivery (as done in the primary analysis) and the referent included all other patients (with or without hypertension). The propensity score for this sensitivity analysis included the same covariates as in the primary analysis plus variables indicating preterm delivery, preterm labour/threatened labour, migraine, arrhythmia, angina, Raynaud disease and in-hospital delivery.

All analyses were performed using sas version 9.2 (SAS Institute, Cary, NC, USA). Institutional review board approval for this research was in place.

Results

Cohort characteristics

The cohort consisted of 9750 patients with outpatient prescriptions for CCBs, methyldopa or labetalol for pre-existing or gestational hypertension whose days of supply overlapped with delivery; of these 1226 were exposed to CCBs (Figure 1), with the balance exposed to methyldopa/labetalol. Overall, 259 (2.7%) patients in the analysed cohort experienced PPH and 201 had PPH due to uterine atony (2.1%).

There were several important differences in the baseline characteristics of the patients exposed to CCBs compared with those exposed to methyldopa or labetalol (Table 1). Those treated with CCBs were more likely to be Black (and less likely to be White or Hispanic) and to carry a diagnosis of pre-existing hypertension, pre-existing diabetes, or previous caesarean delivery. They were less likely to have gestational hypertension. They were also less likely to deliver via caesarean or to take iron supplements. These imbalances were no longer present in the propensity-score-matched cohort; the absolute difference in proportions for all covariates considered was ≤2.0% (Table 2).

Table 1. Baseline characteristics of patients in the overall cohort
 CCB, n (%)Methyldopa or labetalol, n (%)Difference (%)
  1. a

    Cell count cannot be disclosed in accordance with data use agreement forbidding the reporting of cells with <11 patients; counts for Native Hawaiian, more than one race, age <15 or >44 years also not shown due to small cell sizes.

Total12268524 
Age group, years
15–1989 (7.3)643 (7.5)−0.3
20–24298 (24.3)2119 (24.9)−0.6
25–29313 (25.5)2163 (25.4)0.2
30–34246 (20.1)1829 (21.5)−1.4
35–39201 (16.4)1295 (15.2)1.2
40–4470 (5.7)434 (5.1)0.6
Race
White491 (40.1)4186 (49.1)−9.1
Black590 (48.1)2795 (32.8)15.3
Asian13 (1.1)107 (1.3)−0.2
Hispanic58 (4.7)831 (9.8)−5.1
Hispanic and other race15 (1.2)182 (2.1)−0.9
Native American25 (2)164 (1.9)0.1
Unknown27 (2.2)210 (2.5)−0.3
Maternal medical conditions
Pre-existing hypertension1020 (83.2)6860 (80.5)2.7
Pregestationaldiabetes200 (16.3)1211 (14.2)2.1
Obesity125 (10.2)870 (10.2)0
Chronic renal disease38 (3.1)197 (2.3)0.8
Fibroids20 (1.6)103 (1.2)0.4
Maternal obstetric conditions
Gestational hypertension587 (47.9)4379 (51.4)−3.5
Mild pre-eclampsia595 (48.5)4268 (50.1)−1.5
Severe pre-eclampsia74 (6)508 (6)0.1
Gestational diabetes268 (21.9)1896 (22.2)−0.4
Previous caesarean314 (25.6)2000 (23.5)2.2
Placenta praevia16 (1.3)122 (1.4)−0.1
Multiple gestation a 77 (0.9)−0.3
caesarean delivery478 (39)3557 (41.7)−2.7
Induction of labour284 (23.2)1984 (23.3)−0.1
Maternal medications
Heparin a 51 (0.6)0.1
Low-molecular-weight heparin a 41 (0.5)0.3
Aspirin a 48 (0.6)0.2
Oral hypoglycaemics46 (3.8)264 (3.1)0.7
Insulin136 (11.1)797 (9.4)1.7
Iron207 (16.9)1620 (19)– 2.1
Vitamins555 (45.3)3903 (45.8)– 0.5
Medical care usage
Number of hospitalisations
 11002 (81.7)7076 (83)−1.3
 2188 (15.3)1189 (14)1.4
 3+36 (2.9)259 (3)−0.1
Number of outpatient visits
 0–9236 (19.3)1603 (18.8)0.4
 10–13298 (24.3)2319 (27.2)−2.9
 14–17339 (27.7)2112 (24.8)2.9
 18+353 (28.8)2490 (29.2)−0.4
Number of prescription medications (other than antihypertensives)
0558 (45.5)3953 (46.4)−0.9
1412 (33.6)2709 (31.8)1.8
2+256 (20.9)1862 (21.8)−1.0
Table 2. Baseline characteristics of patients in the propensity-score match cohort
 CCB, n (%)Methyldopa or labetalol, n (%)Difference (%)
  1. a

    Cell count cannot be disclosed in accordance with data use agreement forbidding the reporting of cells with <11 patients; counts for Native Hawaiian, more than one race, age <15 or >44 years also not shown due to small cell sizes.

Total12253675 
Age group, years
15–1989 (7.3)292 (8)−0.7
20–24297 (24.2)905 (24.6)−0.4
25–29313 (25.6)919 (25)0.5
30–34246 (20.1)739 (20.1)0
35–39201 (16.4)588 (16)0.4
40–4470 (5.7)211 (5.7)0
Race
White491 (40.1)1464 (39.8)0.2
Black589 (48.1)1766 (48.1)0
Asian13 (1.1)38 (1)0
Hispanic58 (4.7)235 (6.4)−1.7
Hispanic and other race15 (1.2)50 (1.4)−0.1
Native American25 (2)51 (1.4)0.7
Unknown27 (2.2)57 (1.6)0.7
Maternal medical conditions
Pre-existing hypertension1019 (83.2)3062 (83.3)−0.1
Pregestational diabetes199 (16.2)607 (16.5)−0.3
Obesity125 (10.2)399 (10.9)−0.7
Chronic renal disease37 (3)120 (3.3)−0.3
Fibroids20 (1.6)61 (1.7)0
Maternal obstetric conditions
Gestational hypertension587 (47.9)1744 (47.5)0.5
Mild pre-eclampsia595 (48.6)1779 (48.4)0.2
Severe pre-eclampsia74 (6)220 (6)0
Gestational diabetes267 (21.8)784 (21.3)0.5
Previous caesarean314 (25.6)896 (24.4)1.3
Placenta praevia16 (1.3)37 (1)0.3
Multiple gestation a 20 (0.5)0
caesarean delivery478 (39)1394 (37.9)1.1
Induction of labour284 (23.2)828 (22.5)0.6
Maternal medications
Heparin a 25 (0.7)0
Low-molecular-weight heparin a 28 (0.8)0
Aspirin a 24 (0.7)0.1
Oral hypoglycaemics46 (3.8)143 (3.9)−0.1
Insulin135 (11)417 (11.4)−0.3
Iron207 (16.9)608 (16.5)0.4
Vitamins555 (45.3)1612 (43.9)1.5
Medical care usage
Number of hospitalisations
 11001 (81.7)3029 (82.4)−0.7
 2188 (15.4)545 (14.8)0.5
 3+36 (2.9)101 (2.8)0.2
Number of outpatient visits
 0–9236 (19.3)697 (19)0.3
 10–13298 (24.3)898 (24.4)−0.1
 14–17338 (27.6)1028 (28)−0.4
 18+353 (28.8)1052 (28.6)0.2
Number of prescription medications (other than antihypertensives)
0558 (45.6)1712 (46.6)−1.0
1411 (33.6)1215 (33.1)0.5
2+256 (20.9)748 (20.4)0.5

Association of CCB exposure with PPH

Overall, 27 patients exposed to CCBs (2.2%) and 232 patients exposed to methyldopa or labetalol (2.7%) experienced PPH. The unadjusted odds ratio (OR) for PPH associated with CCB exposure was 0.81, 95% confidence interval 0.54–1.21. After adjustment for confounders through propensity score stratification and matching, there was not a meaningful association between CCB exposure and PPH (OR 0.79, 95% CI 0.53–1.19 and OR 0.77, 95% CI 0.50–1.18, respectively; Table 3).

Table 3. Odds ratios of PPH in patients treated with pre-existing to gestational hypertension calcium channel blockers as compared with patients treated with methyldopa or labetalol
 CCBMethyldopa/labetalolOdds ratio (95% CI)
PPH (overall)
Unadjusted27/1226232/85240.81 (0.54–1.21)
Propensity score stratified27/1182228/80720.79 (0.53–1.19)
Propensity score matched27/1225105/36750.77 (0.50–1.18)
PPH from uterine atony
Unadjusted23/1226178/85240.90 (0.58–1.39)
Propensity score stratified23/1182175/80720.90 (0.58–1.41)
Propensity score matched23/122574/36750.93 (0.58–1.49)

There were 23 patients (1.9%) exposed to CCBs and 178 patients (2.1%) exposed to methyldopa/labetalol who experienced PPH due to uterine atony (unadjusted OR 0.90, 95% CI 0.58–1.39). Again, after accounting for confounders through the use of propensity score stratification and matching, there was no meaningful association between CCB exposure and PPH due to uterine atony (OR 0.90, 95% CI 0.58–1.41 and OR 0.93, 95% CI 0.58–1.49, respectively; Table 3).

Sensitivity analyses

When we performed our analyses without requiring a documented diagnosis of pre-existing or gestational hypertension in the inpatient or outpatient claims, results were comparable to those of the main analysis, with no meaningful association between CCB exposure and either PPH overall or PPH due to uterine atony after accounting for relevant confounders through propensity score matching or stratification. Likewise, there was no association between CCBs and either of these outcomes when we restricted our analyses to those who refilled their antihypertensive medication at least once during the third trimester, when we removed variables from our propensity score model that occurred temporally after the initiation of the antihypertensive, or when we analysed the entire cohort adjusting for the indications for CCB exposure (Table 4).

Table 4. Sensitivity analysis (1) without restricting to those with diagnosis of pre-existing and gestational hypertension and (2) those who refilled the medication of interest at least once during the third trimester, (3) excluding from propensity score model clinical conditions or events that occur after the initiation of the medication, and (4) including the entire cohort without restriction
 CCBMethyldopa/labetalolOdds ratio (95% CI)
Without restricting to those with diagnosis of pre-existing to gestational hypertension
PPH (overall)
Unadjusted43/1947268/10 3510.85 (0.61–1.18)
Propensity score stratified43/1865260/98600.88 (0.63–1.23)
Propensity score matched32/1545106/46350.90 (0.61–1.35)
PPH from uterine atony
Unadjusted31/1947210/10 3510.78 (0.53–1.14)
Propensity score stratified31/1865203/98600.85 (0.57–1.25)
Propensity score matched25/154579/46350.95 (0.60–1.49)
Restricting to those with diagnosis of pre-existing or gestational hypertension and who refilled the medication of interest at least once during the third trimester
PPH (overall)
Unadjusted22/915161/63150.94 (0.60–1.48)
Propensity score stratified22/883159/59940.94 (0.60–1.49)
Propensity score matched22/90970/27270.94 (0.58–1.53)
PPH from uterine atony
Unadjusted19/915126/63151.04 (0.64–1.70)
Propensity score stratified19/883124/59941.05 (0.64–1.73)
Propensity score matched19/90957/27271.00 (0.59–1.69)
Excluding from propensity score model clinical conditions or events that occur after the initiation of the medication
PPH (overall)   
Unadjusted27/1226232/85240.81 (0.54–1.21)
Propensity score stratified27/1189228/81430.81 (0.54–1.21)
Propensity score matched27/ 122696/36780.84 (0.55–1.29)
PPH from uterine atony
Unadjusted23/1226178/85240.90 (0.58–1.39)
Propensity score stratified23/1189175/81430.92 (0.59–1.43)
Propensity score matched23/122675/36780.92 (0.57–1.47)
 CCBUnexposed to CCB 
Including the entire cohort, without restriction
PPH (overall)
Unadjusted380/15 05284 836/3 607 4371.08 (0.97–1.19)
Propensity score stratified269/968059 814/2 315 6061.06 (0.94–1.20)
Propensity score matched380/15 0521117/45 1561.02 (0.91–1.15)
PPH from uterine atony
Unadjusted280/15 05267 003/3 607 4371.00 (0.89–1.13)
Propensity score stratified196/968047 575/2 315 6061.03 (0.90–1.19)
Propensity score matched280/15 052823/45 1561.02 (0.89–1.17)

Discussion

Main findings

Despite concerns that exposure to CCBs before delivery may increase the risk of PPH due to their property of causing uterine relaxation, there has not, to our knowledge, been a large, well-controlled study examining this risk. In this cohort study of 9750 patients, after careful control of confounding factors and across a range of sensitivity analyses, we did not observe an increase in the risk of PPH or of PPH from atony in patients exposed to CCBs at the time of delivery compared with those exposed to methyldopa or labetalol. These data suggest that concern about an increase in the risk of PPH should not affect obstetricians' decision about whether to prescribe a CCB for the control of hypertension in the third trimester.

Strengths and limitations

Our study should be interpreted in the context of the limitations inherent in its design. First, although we have information on the dispensing of medications that is free from recall bias, we do not have information about whether the patient was compliant and took the medication as prescribed. To help overcome this limitation, we performed a sensitivity analysis including only those patients that refilled their prescription at least once during the third trimester, which strongly suggests that the patient is in fact consuming the medication. This did not tangibly change the results from the main analysis. Another potential limitation is that we do not have direct information on the reason why the CCB was prescribed. We therefore restricted our analysis to those patients that did not have diagnoses for indications for CCBs other than hypertension (preterm labour/threatened labour, arrhythmia, etc) and required a diagnostic code for pre-existing hypertension or gestational hypertension for the primary analysis. Although there is still some potential for misclassification of the indication for CCBs, such misclassification is likely to be small and unlikely to significantly alter our results. Further, while we use a sophisticated and rigorous statistical methodology, propensity scores, to account for confounding, we are dependent on the recording of claims in the inpatient and outpatient records for demographic and medical conditions; this may result in under-ascertainment or misclassification of some confounders. We attempted to overcome this potential limitation by minimising the degree of confounding at the design stage by restricting our analysis to CCBs used for hypertension using an active comparator (treatment with methyldopa/labetalol) as a referent. There is also the potential in drug safety studies, such as the one presented here, for confounding by contraindication (i.e. if clinicians are concerned about the risk of PPH in a patient, they may not prescribe CCBs). However, such bias is unlikely to significantly affect our results, as most of the major risk factors for PPH are measured and adjusted for in our analysis. Further, we do not have direct information on the interval between the last CCB use and delivery. That said, our analysis does capture the effect of outpatient CCB use on the outcome of PPH as CCBs are used in routine practice. Last, after adjusting for relevant confounders, our study cannot exclude up to an 18% increase in the risk of PPH associated with CCB exposure. However, the MAX dataset used for this study is one of the largest of its kind. There were >9000 patients included in the main cohort who were exposed to CCBs, methyldopa and labetalol and for whom we had longitudinal information from throughout the third-trimester and delivery hospitalisation. The failure to show even a trend towards increased risk of PPH associated with CCBs across a range of analytical approaches and sensitivity analyses, suggests that such an increased risk is unlikely to be present.

Interpretation

There are several potential explanations for the lack of association between CCB exposure proximate to the time of delivery and PPH. First, uterotonic agents are routinely administered following delivery as prophylaxis against uterine atony and haemorrhage.[37] It is possible that these agents readily overcome the tocolytic effects of CCBs, which would yield a null result. Second, as our study examines CCBs administered in the outpatient setting, it is possible that for many women plasma levels are relatively low by the time delivery occurs (if the medication was discontinued during labour or if they were transitioned to another medication while in hospital). Whatever the explanation, it is important that outpatient use of CCBs is not associated with PPH given the frequency with which these medications are prescribed during the third trimester.[17, 18]

Conclusion

In conclusion, our results suggest that the outpatient use of CCBs in late pregnancy for the treatment of hypertension does not increase the risk of PPH. An obstetrician or internist's decision about whether to use a CCB or another type of antihypertensive in late pregnancy should not be influenced by this consideration.

Acknowledgements

None.

Disclosures of interests

The Pharmacoepidemiology Program at the Harvard School of Public Health receives funding from Pfizer and Asisa. SHD has consulted for Novartis, GSK-Biologics and AstraZenaca for unrelated projects.

Contribution to authorship

BTB, SHD, KFH, KP and MAF participated in the design, analysis, and writing. HM participated in the analysis and writing. JLE and EWS participated in the design and writing.

Details of ethics approval

The study was judged by the Partner's IRB to not meet the definition of human research (Protocol # 2011-P-001953/1; MGH).

Funding

This study was supported by the National Institutes of Health T32 GM007592 (BTB) and K24 HL096141 (EWS). The MAX pregnancy cohort was supported by the Agency for Healthcare Research and Quality (AHRQ) (Grant R01HS018533). Kristin Palmsten was supported by a Training Grant T32HD060454 in Reproductive, Perinatal and Pediatric Epidemiology from the National Institute of Child Health and Human Development, National Institutes of Health.

Commentary on ‘Outpatient calcium-channel blockers and the risk of postpartum haemorrhage: a cohort study’

Postpartum haemorrhage (PPH) is the leading cause of maternal mortality worldwide and is on the rise in industrialised countries. The reasons for this rise are unclear (Joseph et al., BJOG 2007;114:751–9).

Calcium-channel blockers (CCBs; usually nifedipine) are recognised as effective tocolytics for preterm labour, reducing the incidence of both birth within 7 days and birth before 34 or 37 weeks (King et al., Cochrane Database Syst Rev 2003;(1):CD002255). It is therefore logical to be concerned that PPH may complicate CCB use for either tocolysis or other indications, such as hypertension.

Whether or not there is a link between CCB use and PPH is not known. PPH was reported as an outcome in a small preterm labour trial (40 women) that found no increase in PPH compared with ritodrine therapy (Weerakul et al., Int J Gynaecol Obstet 2002;76:311–13). PPH was not a pre-specified outcome in the largest of the Cochrane antihypertensive therapy reviews but in one randomised controlled trial, PPH was more common when nifedipine was continued during labour (43.8%) compared with being stopped at labour onset (13.8%) (Abalos et al., Cochrane Database Syst Rev 2007;(1):CD002252). PPH was less common with nimodipine versus MgSO4 for eclampsia prevention (Belfort et al., Am J Obstet Gynecol 1998;178:S7).

In their cohort study of Medicaid-eligible women, Bateman et al. investigate an association between CCB use and PPH among women with International Classification of Diseases, 9th revision, codes for pre-existing or gestational hypertension, and who received CCBs as outpatients but whose prescription covered the delivery date. Women with preterm delivery were excluded. The CCB-exposed women (= 1225) were compared with ‘antihypertensive controls’ who took methyldopa or labetalol (= 3675). The study was well conducted, using propensity scores to balance potential baseline confounders of the drug exposure–PPH relationship. The incidence of PPH overall was low, but did not differ between women exposed to CCBs (2.3%) and antihypertensive controls (2.9%) (odds ratio 0.77, 95% CI 0.05–1.18). Similar results were seen for PPH due to uterine atony (1.9% vs 2.0%, respectively; odds ratio 0.93, 95% CI 0.58–1.49).

The analysis by Batemen et al. has limitations. The antihypertensive exposure was based on outpatient prescriptions with no inpatient information. There is no detail about individual CCBs used. Women who delivered preterm were excluded; yet, superimposed pre-eclampsia is a major cause of iatrogenic prematurity and an important indication for CCB use in hospital. Almost two-thirds of CCB-exposed women were excluded, raising issues of generalisability for women with both hypertension and co-existent indications for CCB use, or, importantly, women with preterm labour who receive CCBs as inpatients. No increase in PPH was shown compared with methyldopa or labetalol, rather than compared with no antihypertensive or no CCB for another indication. The study was also powered to find only a large (50%) increase in PPH.

In summary, despite its limitations, the cohort study by Bateman et al., provides some reassurance for practitioners who use CCBs for outpatient management of hypertension in pregnancy. Future research must address the issue of inpatient use, particularly for preterm labour.

Disclosure of interests

The author has nothing to disclose.

  • LA Magee

  • The University of British Columbia, Vancouver, BC, Canada

Ancillary