Taiwan Centers for Disease Control (TCDC) and Taiwan Food and Drug Administration collaboratively established a national passive surveillance system for adverse event following immunization (AEFI) in concert with the 2009 H1N1 vaccination program [9]. Patients or their parents, healthcare providers, manufacturers, and others were encouraged to report any health event that occurs in patients after receipt of 2009 H1N1 vaccines at any time interval to the system, regardless of causality. Reports were categorized as serious if the adverse event involved death, life-threatening illness, hospitalization, prolongation of hospitalization, permanent disability, or congenital anomaly [10]. Medical records were sought for reports coded as serious, reports suggestive of adverse events of special interest (AESIs) [11], and reports involving pregnancy-specific adverse events.

We searched the passive surveillance database for reports received from November 1, 2009 through August 31, 2010 on pregnant women vaccinated with the 2009 H1N1 vaccines. Each report was investigated by a standard protocol at time of report received. The protocol requested blood and tissue specimens, which were subject for further relevant diagnostic tests, placental or umbilical cord pathology, or fetal autopsy (Table S1). A TCDC physician (WTH) prospectively reviewed all reports, medical records, and results on the laboratory or pathologic investigations. Adverse events were classified by the timing of exposure to 2009 H1N1 vaccines during the first (0–13 weeks), second (14–27 weeks) and third (≥28 weeks) trimester of pregnancy, and by the outcomes including pregnancy-specific and nonpregnancy-specific adverse events. We defined spontaneous abortion as natural loss of conceptuses at <20^th week of pregnancy and stillbirth as nonviable conceptuses at ≥20^th week of gestation.

Data collection was conducted as part of a public health response to the 2009 H1N1 pandemic and therefore did not require approval by an institutional review board or informed consent.

In addition to the passive surveillance, TCDC developed a nationwide large-linked database (LLDB) of computerized vaccinations and medical records for 2009 H1N1 vaccine safety hypothesis testing [9]. The International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnoses of spontaneous abortion (codes 631, 632, 634*, 637*, 761.8) were prospectively collected from the daily updated National Health Insurance (NHI) database. This LLDB had recorded 62% of the 5.6 million doses of 2009 H1N1 vaccines administered to the Taiwan population from November 1, 2009 through August 31, 2010 (TCDC, unpublished data).

Because the proportion of the vaccinated population that experiences an AEFI cannot be directly estimated with the passive surveillance data [12], we used the capture-recapture methodology to assess the true numbers of spontaneous abortion after 2009 H1N1 vaccination for the unadjuvanted and MF59®-adjuvanted vaccines [13]. We matched all reports of spontaneous abortion from the passive AEFI surveillance to cases after 2009 H1N1 vaccination that occurred between November 1, 2009 and August 31, 2010 from the LLDB on a unique identifier assigned to each resident, date of vaccination, and date of diagnosis. The reported dates of vaccination and diagnosis in the two data sources were allowed to differ by up to 7 days to account for recall bias in the passive surveillance reports. We calculated Chapman estimators of the true number of spontaneous abortion after 2009 H1N1 vaccination as N = [(b+1)(c+1)/(a+1)]−1, in which a is the number of cases captured in both sources, b is the number of cases captured in the passive surveillance database, and c is the number of cases captured in the LLDB [14]. Variances for Chapman estimators were calculated using the formula derived by Serber as Var(N) = [(b+1)(c+1)(b−a)(c−a)]/[(a+1)²(a+2)] [15]. A log-transformation was used to obtain the 95% variance-based confidence intervals (CIs) of N so that the lower limit was always greater than the observed number of cases.

Black et al. estimated the background rate of spontaneous abortion to be 3.5 to 22.4 per 100 pregnancies, varying by age and country [16]. For the general pregnant population in Taiwan, background rates were 12.8 (95% CI, 12.8–12.9) spontaneous abortions per 100 pregnancies according to the published literature [17]. We calculated the capture-recapture estimated risks of spontaneous abortion in pregnant women who received the 2009 H1N1 vaccines. The denominator data on 2009 H1N1 vaccine doses administered to pregnant women between November 1, 2009 and August 31, 2010 were obtained from the National Influenza Vaccine Information System. The Influenza Vaccine Information System monitors 2009 H1N1 vaccine utilizations in real-time; the data distinguished doses administered by vaccine type (the unadjuvanted or MF59®-adjuvanted vaccine), but not by the timing of vaccination in relation to stage of pregnancy [9].

Among 13,199 pregnant women who received the unadjuvanted 2009 H1N1 vaccine, 15 spontaneous abortions, 9 stillbirths, 3 neonatal deaths, and 3 nonpregnancy-specific adverse events were reported (Table 2). The median days from 2009 H1N1 vaccination to the occurrence of spontaneous abortion was 17 days (range, 0–45 days). In 7 of the 15 spontaneous abortion reports, advanced maternal age (≥35 years) were identified. One spontaneous abortion occurred in a woman who was 36 years of age and received the vaccine at 10^th week of pregnancy; cytogenetic analysis identified trisomy 21. The median days from 2009 H1N1 vaccination to the occurrence of stillbirth was 18 days (range, 1–70 days). All of the 9 stillbirths reported more than one risk conditions associated with stillbirth [18], including maternal age ≥35 years (n = 3), chorioamnionitis (n = 2), preterm premature rupture of membranes (n = 2), gestational diabetes mellitus (n = 1), hyperthyroidism (n = 1), preeclampsia (n = 1), multiple gestation (n = 1), small for gestational age fetus (n = 1), fetal hydrocephalus due to intraventricular hemorrhage (n = 1), placental insufficiency (n = 1), and oligohydramnios (n = 1). Cause of death for the three neonates varied (Table 2). No major birth defect was observed for stillborn fetuses or live-born infants who died in the neonatal period.

The three reports involving nonpregnancy-specific adverse events included allergic vasculitis (n = 1), numbness of fingers (n = 1), and dizziness, tremor, and rhinorrhea (n = 1). The allergic vasculitis patient received the unadjuvanted 2009 H1N1 vaccine at 10^th week of pregnancy and reported onset of adverse events 10 days after vaccination. She was treated with systemic corticosteroids, which led to an elective termination of pregnancy due to perceived risk of corticosteroids on fetal development.

Reports involving the 1,275 pregnant women who received the MF59®-adjuvanted 2009 H1N1 vaccine included one spontaneous abortion, two stillbirths, one neonatal death, one nonpregnancy-specific adverse event, and two inadvertent immunizations (Table 2). Chorioamnionitis was reported in the spontaneous abortion that occurred at 15^th week of pregnancy, 46 days after 2009 H1N1 vaccination. Onset days from 2009 H1N1 vaccination to the occurrence of stillbirth was 7 and 32 days, respectively. Maternal age was ≥35 years in one stillbirth; for the other, no relevant risk condition associated with stillbirth could be identified. For the infant who died within one month of birth, cause of death was cerebral hemangioblastoma with intracranial hemorrhage. The nonpregnancy-specific adverse event was generalized skin rash in a woman 28 years of age, 30 days after receipt of the MF59®-adjuvanted 2009 H1N1 vaccine at 11^th week of pregnancy.

Sixteen cases of spontaneous abortion after 2009 H1N1 vaccination were identified in the passive surveillance database and 135 cases were identified in the LLDB; 6 matches occurred between the two sources. We estimated that the true number of spontaneous abortion in pregnant women who received the 2009 H1N1 vaccines to be 329 (95% CI, 196–553) and reporting completeness of the passive surveillance to be 5% (95% CI, 3%–8%). The estimated risk of spontaneous abortion for the 14,474 pregnant women who received 2009 H1N1 vaccines was 2.3 (95% CI, 1.4–3.8) per 100 pregnancies, compared with a local background rate of 12.8 (95% CI, 12.8–12.9) per 100 pregnancies.

Table 3 showed the number of spontaneous abortion ascertained from the combinations of capturing data sources for different types of 2009 H1N1 vaccines. We estimated that risk of spontaneous abortion was 2.2 (95% CI, 1.3–4.0) and 2.0 (95% CI, 2.0–2.0) per 100 pregnancies associated with the unadjuvanted and MF59®-adjuvanted vaccine, respectively.