Although multiple pregnancy rate in natural conception is less than one percent, this increases significantly in assisted reproductive cycles. The number and rate of multiple pregnancies have increased over the past two decades, during which the number of twin deliveries rose by 65 percent, and the rate of triplet and higher-order multiple pregnancies increased by more than 400 percent (1). In the majority of developed countries, 30-50% of all twin pregnancies occur as a result of infertility treatment (2). In the United States, multiple pregnancies were reported to occur in 35.4% of in vitro fertilization cycles, which is more than ten times the natural multiple pregnancy rate (3).

The majority of perinatal morbidities associated with multiple gestations are related to preterm delivery. Gestational age at delivery and birth weight are the two most important factors that affect perinatal, neonatal, and infant morbidity and mortality (4). In addition to the morbidity and mortality attributable to preterm delivery, fetuses in multiple gestations are vulnerable to a variety of complications, such as malformations and twin-to-twin transfusion syndrome. Maternal complications such as preeclampsia, gestational diabetes, postpartum hemorrhage, and maternal death are also increased in multiple gestations, and these risks are further increased in triplet or higher-order multiple pregnancies (5).

Multifetal pregnancy reduction (MFPR) is a procedure that reduces higher-order pregnancies, which improves perinatal outcomes. Multifetal pregnancy reduction in triplet and higher-order multiple pregnancies significantly reduces the risks of prematurity and a low birth weight, and may also be associated with a reduction in overall pregnancy loss and obstetric complications (6-8). Moreover, no significant differences were found in mean gestational age at birth, birth weight, and perinatal mortality rates between reduced and non-reduced twins (9). A recent meta-analysis concluded that multifetal pregnancy reduction to twins, compared with expectant management, seems to be an effective treatment option for women with a triplet pregnancy (10). These results suggest that MFPR of higher-order pregnancies to twins is a medically justifiable procedure.

Since the MFPR was first introduced in the mid-eighties (11), several modifications of methods have been developed. Various options are available in regard to the route of approach, the timing of procedure, and the use of embryotoxic agents such as potassium chloride (KCl). However, controversy still exists regarding which method is better than the others (12-15). We retrospectively analyzed different methods for MFPR to evaluate the outcomes of procedure, and to determine which produces the better outcomes.

The mean age of the subjects was 30.6±2.9 yr. Eightyone MFPRs were performed before 8 weeks' gestational age (54.7%) and 67 procedures were performed at 8 weeks or later (45.3%). In 76 patients, KCl was used as the embryo toxic agent (51.4%) and not used in the other 72 patients (48.6%). Maternal ages and characteristics of the MFPR procedures such as starting number of fetuses, finishing number of fetuses, and the number of procedures required to complete the procedure were not different among the four groups. Gestational ages at the times of procedure were different between the groups (Table 1).

Mean gestational ages at delivery and birth weights were not different between KCl and non-KCl groups and between Early and Late groups. As compared with KCl groups, non-KCl groups showed a significantly higher take-home-baby rate (86.1% vs. 69.7%, p=0.045). Immediate loss and pregnancy loss rates were lower in non-KCl groups than in KCl groups, but the differences were not statistically significant (5.6% vs. 10.5% and 12.5% vs. 23.7%, respectively). A significant difference was found between Early and Late groups in terms of immediate loss rates (2.5% vs. 14.9%, p=0.019). Early groups showed a lower pregnancy loss rate and a higher take-home-baby rate compared with Late groups, but the differences were not statistically significant (12.3% vs. 25.4% and 81.5% vs. 73.1%, respectively) (Table 2).

Fetal and maternal complications, such as fetal growth restrictions, discordant twins, congenital anomalies, gestational hypertension, and cervical incompetence were not different between KCl and non-KCl groups or between Early and Late groups. Extreme prematurity rates were significantly higher in KCl groups than in non-KCl groups (8.6% vs. 1.6%, p=0.020). PPROM occurred significantly more frequently in KCl groups than in non-KCl groups (27.6% vs. 9.7%, p=0.019) (Table 3).

The 'Early, non-KCl' group showed a significantly lower immediate loss rate, pregnancy loss rate, and PPROM rate as compared with the 'Late, KCl' group (3.3% vs. 14.5%, p=0.033; 11.7% vs. 27.3%, p=0.034; 10.0% vs. 27.3%, p=0.017, respectively). The take-home-baby rate was significantly higher in the 'Early, non-KCl' group than in the 'Late, KCl' group (86.7% vs. 70.9%, p=0.038) (Table 4).

Several methods of MFPR are available based on different combinations of three components, i.e., the use of embryotoxic agent, the timing of procedure, and the route of approach. However, it is difficult to determine which one is a better option because prospective randomized studies or simultaneous comparisons among these variable methods are difficult to perform. In particular, no randomized controlled trial has been conducted (17), and few comparative studies have been conducted to compare the effects of different factors on outcomes. The present study is a retrospective comparative study of different modalities comprised of two components of the MFPR procedure, i.e., the use of an embryotoxic agent, and the timing of procedure. To our knowledge, this is the first study to evaluate the impact of those two components simultaneously.

Potassium chloride is widely used for MFPR, but the safety and efficacy of this agent are debatable. Cases of anencephaly and limb amputation have been reported, and total pregnancy loss may be resulted if the KCl solution accidentally reaches the amniotic fluid of remaining fetuses (18). It has been suggested that the development of an inflammatory response to the resorbing dead feto-placental tissue with subsequent release of cytokines and stimulation of prostaglandins is a cause of pregnancy loss, preterm delivery, and other complications following MFPR (19-21). All procedures in those studies were performed by KCl injection. Based on those previous findings and our results, we suggest that the use of KCl for MFPR may cause or aggravate the inflammatory process and induce PPROM and preterm birth.

The reason for the higher PPROM rate for KCl groups in the present study is not clear. However, several reports have demonstrated that PPROM is associated with matrix degrading enzymes, such as, plasminogen activators (PAs) and matrix metalloproteinases (MMPs) (22-24). Moreover, in animal studies, investigators found that KCl can induce a release of tissue plasminogen activator in the hypothalamo-neurohypophysial system (25) and an upregulation in MMP-9 activity in the retina, which promote retinal damage (26). Although no study has been conducted on the roles of KCl in relation to matrix degrading enzyme activity in the intrauterine system, we hypothesize that KCl can cause PPROM by inducing matrix degrading enzymes in the uterus after MFPR.

Optimal MFPR timing remains controversial. There have been studies that suggested that fetal reduction should be delayed until the late first trimester, based on the expectation of a natural reduction of fetus and a detection of anomalies (27, 28). However, studies that have focused on loss rates per gestational sac or embryo, not per patient, observed that the rates of spontaneous fetal demise in multiple gestations during early pregnancy were much lower than those reported by other studies that did not consider this condition (29-31). Moreover, reported rates of malformation after MFPR are lower than expected rates, which is probably due to the tendency to selectively reduce embryos with a higher risk of congenital malformation (12). Several studies reported that MFPR during early gestation, at 7-8 weeks, without KCl injection showed superior results to that performed later (13, 16). In the present study, better outcomes were observed in the Early groups, and we confirm these previous study results.

Our data show that 'early, non-KCl' is the best of the different MFPR methods. This is thought to have both advantages of early MFPR and that of non-KCl MFPR. Performing MFPR in earlier gestational age without KCl injection makes the procedure easier because the targeted fetus is relatively small and easily aspirated. If the procedure is performed in later gestation, the relatively large fetus cannot be easily terminated. The early, non-KCl method is also easier and faster than the other methods and is probably related to fewer procedure-related complications. The significantly lower immediate loss rate in the 'Early, non-KCl' group in our data supports this suggestion. Another advantage of the early method is that it avoids the use of KCl. Potassium chloride is more needed as gestational age increases because larger fetuses cannot be easily terminated by aspiration without KCl injection. As we have discussed earlier, using KCl is not a safe way to perform MFPR, and thus, the use of KCl should be avoided. Early MFPR favors this approach. The early, non-KCl method has benefits of both early and non-using KCl procedure, and thus it could be the best among the four methods investigated.

In the present study, all procedures were performed via the transvaginal route. Several studies have been conducted on different routes of approach, and shown controversial results (12, 15, 32). Experience and skill are also important factors in addition to the route of approach. Like any other procedures, MFPR has a learning curve, and thus outcomes improve with the number of cases. In a collaborative study of 3,513 cases from 11 centers, Evans et al. found the outcomes improved considerably with increasing experience (33). Therefore, if poor initial outcomes are inevitable due to the learning curve, simpler methods should be adopted. Reproductive physicians are more comfortable with transvaginal MFPR because the procedure is similar to ovum aspiration under transvaginal ultrasonographic guidance. Thus, transvaginal MFPR requires fewer technical considerations than transabdominal MFPR. Moreover, not every center performs several hundreds of MFPR procedures as in the large centers included in the previous studies. As cases of highorder multiple pregnancies become less frequent, the number of cases requiring MFPR will also decrease. In this situation, each center must choose the method that best fits its own situation. Transvaginal MFPR could be a better option for centers that perform relatively few procedures and have relatively less experiences. In a collaborative study involving more than 1,000 cases, Evans et al. (32) observed that transabdominal MFPR was more associated with a poor outcome than transcervical or transvaginal MFPR earlier in the series, which also support our suggestion.

Limitations of the present study should be mentioned. First, this is an observational study and not a randomized controlled trial; hence there may be confounding variables that might have influenced the results. However, the similarity between the groups in clinical characteristics such as maternal age, starting number of fetuses, finishing number of fetuses, and number of trials supported that selection bias would be minimal. Second, regarding sample sizes, 250 patients in each group achieve 80% power at a 5% significance level when the difference of take-home-baby rate is 10%. Sample size calculations with other proportional variables showed similar results. In the present study, there are several proportional data that showed a nonsignificant trend of about 10% difference. We could not rule out the possibility that the nonsignificant differences seen in those data could be attributed to a type II error. Further studies in a larger scale will be necessary.

In conclusion, we suggest that the early transvaginal non-KCl method is a better option for MFPR. We believe that it should be considered as the first choice modality in the majority of centers.