The Burkina-Faso registry of PPCM prospectively includes women diagnosed with PPCM in the university hospital of Ouagadougou and Saint Camille hospital of Ouagadougou. The criteria for the diagnosis of the initial peripartum cardiomyopathy include the development of congestive heart failure within 1 month before delivery to 5 months after delivery in the absence of any another identifiable cause of heart failure; and evidence of depressed left ventricular function, defined as a left ventricular ejection fraction (LVEF) of less than 45% (Simpson method), as measured by echocardiography [7]. Women from the PPCM registry who were hospitalized during a subsequent pregnancy (despite medical advice not to become pregnant after PPCM) were prospectively included from 2012 to 2016. In-hospital to one week post discharge data were collected. We assessed clinical condition, left ventricular diameters and LVEF, tricuspid annular plan systolic excursion (TAPSE) and systolic pulmonary arterial pressure. Delta (%) values of these parameters were assessed as follows: (value at admission – baseline value) / baseline value. Maternal and fetal mortality, pregnancy issues, way of delivery and newborn weight were also appreciated. We compared alive and deceased women to identify the factors associated with maternal death. We also conducted a second analysis focused on fetal prognosis to identify factors associated with miscarriage or prematurity in these women.

To examine determinants of death events, we examined the area under the receiver-operating characteristic (ROC) curve (plot of sensitivity versus 1 − specificity for all possible cut-off values for classifying predictions) for LVEF, systolic Pulmonary arterial pressure (sPAP) and TAPSE with the best sensitivity and specificity according to the Youden index [8]. The cut-off value is given in the results section. Areas under the ROC curves were compared using the method of DeLong et al. [9] for paired data.

All analyses were performed using SPSS 20.0 0 (SPSS, Inc., Chicago, IL, USA) and MedCalc 13.3.1 (MedCalc Softaware, Mariakerke, Belgium).

During the inclusion period, 29 women followed in the PPCM registry were hospitalized while pursuing a subsequent pregnancy. The mean age was 26.7 ± 4.6 years, with predominance of low socio-economic status. The diagnosis of cardiomyopathy had been made before delivery in 4 women (all of them in the seventh months of pregnancy), during the first month after delivery in 20 women, and between two and six months after delivery in 5 women (3 women in the second month after delivery, 1 in the third month, and 1 in the fifth month). No case of preeclampsia, chronic hypertension, severe anemia or history of hyperthyroidism was identified. Clinical and echocardiographic characteristics of patients during subsequent pregnancy in both groups alive vs deceased (according to the occurrence of maternal death during subsequent pregnancy) are summarized in Table 1.

The mean LVEF in the total cohort of 29 women at the last follow-up before pregnancy was 49.9 ± 5.1% and decreased significantly to 37.6 ± 4.7% after the first subsequent pregnancy with a mean delta LVEF of − 19.3 ± 9%.

Fourteen (14) women died (48.3%). There was no statistically significant difference between alive and deceased women concerning socioeconomic status, left ventricular and right ventricular systolic function before subsequent pregnancy (baseline). At the time of hospitalization (during the subsequent pregnancy) (Table 1), congestive heart failure (p < 0.001), with lower left and right ventricular function was more frequently observed in women who died after pregnancy than in the group of surviving patients.

To find out if echocardiographic data (at the last follow-up and at admission) could predict mortality after pregnancy, we built ROC curves using variables with a strong statistical association to death in univariate analysis: LVEF (at last follow-up (=baseline), at admission and delta LVEF), sPAP (at last follow-up, at admission and delta sPAP), and TAPSE (at last follow-up, at admission and delta TAPSE). The AUC were compared to determine the best variable, and the best cut-off value. Among these variables, LVEF at admission (AUC = 0.95; 95% CI 0.87–1, p < 0.001), TAPSE at admission (AUC = 0.879; 95% CI 0.74–1, p = 0.001) and sPAP at admission (AUC = 0.85; 95% CI 0.71–0.99, p = 0.001) were strongly and significantly associated to death (Figs. 1, 2, 3). Comparing the AUC of these 3 variables at the time of admission (data not shown), there was no statistically significant difference. The best cut-offs for these variables are detailed within the figures.

Miscarriage occurred in six cases. Among the 23 women with a living child after subsequent pregnancy, 13 had a normal vaginal delivery, and 10 were delivered by cesarean. The mean newborn weight was 2385 ± 286 g. Miscarriage and prematurity were associated with baseline LVEF and baseline TAPSE as shown in Table 1. Baseline LVEF had the best AUC to predict SSP outcomes among all variables (AUC = 0.75; 95% CI 0.58–092, p = 0.003), with a cut-off value of < 50% (sensitivity = 79%, specificity = 67%).Four newborn (that 3 preterm) died, all from neonatal infection.

Our study demonstrated that when women with SSP did not benefit from cardiac follow-up, the issue can be fatal.

Mortality was remarkably high in our study, probably because the women were lost from clinical follow-up during the pregnancy period.

In several studies [10‐12], SSP was responsible for maternal death in about 0.5 to 1%. This contrasts with our results. This can be explained by the fact that our patients were lost from follow-up and the lack of therapeutic means. Indeed follow-up enables an effective management of heart failure, pregnancy and delivery. During this follow-up, normalization of left ventricular function does not guarantee an uncomplicated SSP as demonstrated by some authors in America and Europe. In their study, approximately 20% of such patients are also at risk of moderate to severe deterioration of left ventricular function, which persists after delivery in 20 to 50% of cases [10‐12].

We demonstrated that left ventricular ejection fraction decreased during SSP.

The best determinant of maternal death in our study was the left ventricular ejection fraction at admission and the percentage of loss of left ventricular ejection fraction. The best cut-off value of left ventricular ejection fraction to predict mortality during first SSP in our study was a LVEF at admission < 40% [Sensitivity = 93%, Specificity = 87%].

All previous studies were based on left ventricular function, but right ventricular function was not studied. We demonstrate that right ventricle systolic function also has an important role in the prediction of maternal mortality during SSP. We appreciated right ventricular systolic function by TAPSE, but we also measured systolic pulmonary arterial pressure to determine if sPAP could predict mortality during SSP.

We found that all these parameters were associated to maternal death. These results underline that right ventricular function must be taken into account to estimate maternal outcomes of SSP. In a study by Elkayam et al. [10], medical pregnancy interruption was used in some cases to prevent maternal outcomes. In their study, maternal mortality was very low compared to our result, but medical pregnancy interruption was more frequent.

All our results demonstrate that PPCM women with SSP are very fragile. They must benefit from a rigorous follow-up including echocardiographic assessment.

Pregnancy outcomes were also poor. We noted that miscarriage and prematurity were associated with baseline left ventricular ejection fraction and baseline TAPSE. This can be explained by the fact that low baseline left ventricular ejection fraction compromises fetal development. The same reason can explain why our new borns had all small birth weight.

Elkayam et al. [10] demonstrated that abortion rate is very high (20.4%), and prematurity is two times more frequent in women with persistent baseline left ventricular dysfunction. In our study, not only baseline left ventricular ejection, but also baseline TAPSE was responsible for pregnancy outcomes.

Death occurred predominately in premature new borns (3/4). There was no maternal nor fetal factor linked to these deaths. However, as we know, prematurity is a factor of mortality in new born babies.

Our study included only a small number of PPCM women but it is to the best of our knowledge the biggest African study on SSP outcomes.

A recruitment bias has to be acknowledged: only hospitalized women were included, due to loss of follow-Up. Indeed, from the 166 women included in the registry, 38 were lost to follow-up. From these lost to follow-up women, 29 were hospitalized during a SSP and 9 were missing. None of the 128 patients followed in the registry had so far a SSP. Among the 9 patients missing, uncomplicated pregnancies could have occurred. Thus, the mortality rate of SSP described in our study could be slightly overestimated.