For many years, there has been controversy in the medical community regarding the correlation of female hormonal factors with the outcome of women with malignant melanoma. There have been multiple reports that women with high hormone states, such as pregnancy, had thicker tumors and/or a worse prognosis compared with a group of control women.
The authors used a database that contained maternal and neonatal discharge records from the entire state of California from 1991 to 1999 and linked those records to the California Cancer Registry, which maintains legally mandated records of all cancers reported in California during the same time period. Four hundred twelve women with malignant melanoma diagnosed during or within 1 year after pregnancy were identified (145 antepartum, 4 at delivery, and 263 postpartum) and were compared with a group of age-matched, nonpregnant women with melanoma (controls). The database captured only pregnancies at ≥ 20 weeks of gestation.
When comparing women who had pregnancy-associated melanoma with the control group, the authors found no difference in the distribution of disease stage (82.0% of pregnant and postpartum women had localized melanoma vs. 81.9% of control women) or the tumor thickness (mean: 0.77 mm for pregnant women, 0.90 mm for postpartum women, and 0.81 mm for the control group). In a multiple regression model that controlled for age, race, stage, and tumor thickness, pregnancy had no impact on survival in women with melanoma. Lymph node assessment and positivity of lymph nodes also were equivalent between the two groups. Maternal and neonatal outcomes did not differ between pregnant women with melanoma and control women who were pregnant and had no history of malignancy. Small numbers of women with advanced melanoma and the inability to capture melanoma that occurred in pregnancies that were lost or were terminated prior to 20 weeks limited the conclusions primarily to women with localized melanoma.
The incidence of cutaneous malignant melanoma has been increasing over the last decade by an average of 3% per year.1 Historically, both the occurrence of melanoma and deaths caused by melanoma have been more common in men than in women. Reasons for this difference are not completely clear, although the tendency of men to develop melanoma on the head or trunk area, compared with melanoma of the extremity area, which indicates a relatively better prognosis, may account for some of the difference in mortality.2 Despite this, women in the U.S. have an incidence rate of 13.5 per 100,000 population, making it more common than carcinoma of the uterine cervix.1
The question of whether hormonal factors play a role in either the difference in incidence or the prognosis between men and women long has been debated. Case reports and retrospective reviews dating from the 1950s reported a poor prognosis for women with melanoma during pregnancy and emphasized the risk of metastasis from malignant melanoma to the placenta and fetus.3, 4 Those reports, along with the increased pigmentation of several areas of the body found naturally in pregnancy, have led to conjecture about the possible increased incidence and/or worsened prognosis associated with pregnancy either during or after the diagnosis of malignant melanoma.5 Although more comprehensive epidemiologic studies have failed to show an adverse effect of pregnancy on melanoma, there continue to be controversial recommendations made to patients on that presumption.5
In this report, we review data on women with malignant melanoma who were diagnosed during pregnancy and postpartum (from 9 months prior to 1 year after delivery) in California for the years 1991–1999, review the prognosis for those women compared with a control group of nonpregnant women with melanoma, and outline the available data with regard to stage distribution, histologic type, tumor thickness, surgical management, and maternal and fetal outcomes. This analysis represents the largest population-based evaluation of malignant melanoma associated with pregnancy to date.
MATERIALS AND METHODS
For the purposes of this study, we utilized a linked data set available from the California Office of Statewide Health Planning and Development (OSHPD) that consists of linked vital statistics and birth, infant death, and neonatal and maternal hospital discharge records (VS/PDD) for the years 1991–1999. This database documents all births beyond 20 weeks, including multiple births and stillbirths. Overall, 4,846,505 obstetric patients and 4,906,920 infants were reviewed through this database. These data were then linked by a probabilistic data record linkage to data from the California Cancer Registry (CCR). Cancer reporting in California by hospitals, physicians, and laboratories is mandated by state law, and the CCR also conducts active case-finding to ensure complete case ascertainment. Records from pathology laboratories throughout the state are reviewed routinely to identify all malignancies. Linkage between the obstetric database (VS/PDD) and the data from the CCR was performed by means of Integrity software (version 3.3; Vality Technologies, Inc., Boston, MA).6 By identifying women with malignant melanoma through the CCR files for the years 1991–1999, comparison between the two data sets of several variables, including Social Security number, date of birth, gender, and zip code of residence, could be performed, allowing us to identify women with malignant melanoma associated with pregnancy. Linked files contained coded identifiers to preserve patient anonymity. Data available through the CCR included the date of melanoma diagnosis, anatomic site, clinical and pathologic stage, histologic type, type of surgery performed, and survival after diagnosis.
In this study, disease stage assessment was based on the CCR-coded variable “SEER summary stage at diagnosis.” This variable was coded according to Surveillance, Epidemiology, and End Results (SEER) guidelines, which utilize a combination of clinical and pathologic information, including lymph node biopsy results when available.7 The CCR used SEER summary stage for patients diagnosed prior to 1994 and SEER Extent of Disease (EOD) guidelines for patients diagnosed in 1994 and later. EOD was converted to SEER summary stage and then categorized as localized, regional, or remote. This summary stage, which either was coded directly or was converted from EOD, was available for all years of the study. For analyses of tumor thickness (Breslow), only women who were diagnosed in 1994 and later were included. Breslow tumor thickness categories corresponding to the American Joint Committee on Cancer (AJCC) staging system that was in use prior to 2001 were utilized, because this represented the years from which the patients were drawn. Only women with invasive malignant melanoma were included, and women with in situ melanoma were excluded. Lymph node evaluation was assessed through the CCR variables “number of lymph nodes assessed” and “number of lymph nodes positive.” The number of patients with lymph node information, therefore, captured all surgical assessments, including sentinel lymph node technique, biopsy, or elective or therapeutic lymphadenectomy. Women with maternal melanoma that occurred within 9 months preceding obstetric delivery were termed “antepartum,” or were termed “postpartum” if the earliest date of diagnosis occurred within 12 months after delivery. Melanomas were classified as diagnosed “at delivery” if the earliest date of diagnosis was during the delivery hospitalization. Data for pregnant women with melanoma were compared with a group of age-matched, nonpregnant women who had melanoma (the control group). To select the control group, the women with pregnancy-associated melanoma were excluded from the data set of all women with melanoma during a particular year; then, the match was achieved by selecting the same proportion of women in each age group (ages < 20 years, 20–25 years, 25–35 years, 35–40 years, 40–45 years, and ≥ 45 years) for pregnant and nonpregnant women with melanoma.
The Cochran–Armitage test was used to test for trends.8 The Mantel–Haenszel age-adjusted and race-adjusted odds ratios were calculated to describe the risk of various maternal and neonatal outcomes if diagnosed with melanoma. Fisher exact tests and Student t tests were used to determine the differences in clinical outcomes between comparison groups. Hazard ratios were obtained by Cox proportional hazards model to measure the relative contribution of various factors to the probability of dying from melanoma.9 The cut-off date used for survival was December 31, 2001. Kaplan–Meier estimates of survivorship functions were used to draw survival distribution curves, and log-rank tests were performed to analyze the difference of survival functions.10 All analyses were performed by using SAS software (version 8.2; SAS Institute Inc., Cary, NC). This investigation was approved by the OSHPD, by the Human Subjects Committee of the University of California—Davis, and by the State of California Committee for the Protection of Human Subjects.
In total, 412 women were identified through the databases who were diagnosed with cutaneous malignant melanoma during pregnancy or within the first year postpartum (pregnancy-associated melanoma), including 145 women who were diagnosed within 9 months predelivery (antepartum), 4 women who were diagnosed at the time of delivery, and 263 women who were diagnosed within 1 year postpartum. This correlates with an overall occurrence rate of 8.5 per 100,000 pregnant women in the database.
In Table 1, the demographics and tumor characteristics of our patient population are illustrated. Women with pregnancy-associated melanoma were similar in terms of age, race, and disease stage; as expected, 92% of women were non-Hispanic white women, and only small numbers of women were identified in any other ethnic groups. Most of the women were between the ages of 25 years and 35 years. Although it is not shown in the table, the majority of pregnant women with melanoma were multiparous and started their prenatal care in the first trimester. The stage of disease identified in the pregnant patients (diagnosis within 9 months prior to delivery and up to 12 months after delivery) showed no significant difference compared with age-matched, nonpregnant women with melanoma. In a breakdown of patients either antepartum (0–9 months prior to delivery) or postpartum (up to 12 months after delivery) who presented with regional or remote disease, a total of 6.9% of antepartum women presented with regional or remote disease, and 8.3% of the antepartum group (n = 145 women) died of disease. In the postpartum group (n = 263 women), 3.8% presented with regional or remote disease, and 6.1% died of disease. Neither of these groups differed statistically from the nonpregnant control group, in which 5.8% of women had regional or remote disease, with 9.8% mortality.
Table 1. Demographic Variables and the Risk for Maternal Malignant Neoplasma
Age-matched nonpregnant women
NOS: not otherwise specified.
There were 412 pregnant and postpartum patients in total, including 145 patients who were diagnosed 9 months before delivery, 4 patients who were diagnosed at the time of hospitalization for delivery, and 263 patients who were diagnosed within 1 year postpartum. The total number of age-matched nonpregnant patients is 2451.
< 25 yrs
> 35 yrs
Other (Native American/unknown)
≤ 0.75 mm
> 1.5 mm
Lower limb and hip
Scalp and neck
Skin of trunk
Upper limb and shoulder
Hutchinson melanotic freckle
Malignant melanoma, NOS
Superficial spreading melanoma
Also illustrated in Table 1 are the Breslow thickness distributions for pregnant versus nonpregnant women with melanoma from the years 1994–1999. These also were similar statistically. Table 2 amplifies this comparison, showing that the mean and median tumor thicknesses (Breslow) were similar between pregnant and nonpregnant women for localized disease and regional/remote disease. The mean tumor thickness in the pregnant women with localized disease was 0.77 mm, compared with 0.81 mm for the nonpregnant control women.
Table 2. Tumor Thickness: Pregnant and Age-Matched Nonpregnant Women with Malignant Melanoma in California, 1994–1999a
Age-matched nonpregnant women
Diagnosed 0–9 mos before and at delivery
Diagnosed 1 yr postpartum
A Student t test did not find significant differences in the means between pregnant and nonpregnant groups within the two stage groups.
No. of women
Tumor thickness (mm)
No. of women
Tumor thickness (mm)
The site of the primary tumor also is compared in Table 1. Overall, there were no significant differences in anatomic site, such as lower limb versus truncal occurrences, when comparing pregnant women with nonpregnant women.
The histologic types of melanoma were compared between women with pregnancy-associated melanoma and age-matched, nonpregnant controls and also are included in Table 1. Forty-one percent of pregnancy-associated melanomas were classified as superficial spreading melanoma, which was similar to the rate of 43% of melanomas in the nonpregnant control group.
Figure 1 shows the survival curve for women with all stages of pregnancy-associated melanoma compared with the nonpregnant control group. Survival data were available for 303 of the pregnant/postpartum patients and for 1799 of the nonpregnant patients. The vast majority of women with melanoma in pregnancy for which survival data were available (n = 289 women) had localized disease. A survival distribution comparison in these patients showed that nonpregnant women with localized melanoma trended toward a shorter survival compared with their pregnant counterparts, but this trend was not significant (log-rank test; P = 0.16) (Fig. 2). There also was no significant difference in survival between women with regional and remote stage disease associated with pregnancy and nonpregnant women (log-rank test; P = 0.82), although the numbers in this group were quite small (n = 14 pregnant women and 83 nonpregnant women).
Various prognostic factors were assessed for their effects on the survival of patients with melanoma using a Cox proportional hazards model. These factors are illustrated in Table 3. Tumor thickness was the strongest predictive factor for survival. Patients with tumors that measured 0.76–1.5 mm in thickness died at a rate that was approximately 2.6 times the rate of patients with tumors that measured < 0.76 mm in thickness, and patients with tumors that measured > 1.5 mm in thickness were 9.7 times more likely to die than patients with tumors that measured < 0.76 mm in thickness. Stage (localized vs. regional and remote) also was a strong predictive factor, with a hazard ratio of 4.8 for patients who had regional or remote disease compared with patients who had localized disease. Age at diagnosis did not have significant effect on survival for the patients with melanoma in this population. Pregnancy also did not appear to affect survival after controlling for age, race, disease stage, and tumor thickness.
Table 3. Hazard Ratios for Death Stratified by Pregnancy, Tumor Size, and Race for Women Diagnosed with Melanoma, 1991–1999a
Pregnant women did not differ significantly from nonpregnant women in terms of survival.
Diagnosed 0–9 mos prior to and at delivery
Diagnosed 1 yr postpartum
≤ 0.75 mm
> 1.5 mm
< 35 yrs
≥ 35 yrs
We also assessed the timing of the melanoma diagnosis in relation to the pregnancy and compared outcomes depending on the time of diagnosis (Fig. 3). Complete data for gestational age at delivery were available for 399 of 412 patients. In an attempt to discern whether the diagnosis of melanoma may have caused any patients to deliver earlier (i.e., iatrogenic prematurity), a comparison was made between the timing of the diagnosis and the percentage of women in each group who delivered prematurely. A similar number of patients were diagnosed during the first, second, and third trimesters of pregnancy, and no difference was seen between these groups with regard to the likelihood of premature delivery (approximately 6–11%). Women who were diagnosed with melanoma postpartum also were not more likely to have been delivered prematurely (approximately 6–12%).
Maternal and neonatal outcomes were compared between women with malignant melanoma who were diagnosed during pregnancy (diagnosis 0–9 months prior to delivery and at delivery) and control women who did not have melanoma. Table 4 shows the rates for various outcome measures and compares them by age-adjusted and race-adjusted odds ratios with pregnant women without melanoma. Overall, there were no significant differences in any endpoint, including cesarean delivery, length of stay, risk of low birth weight or very low birth weight, prematurity, hospital charges, or neonatal death. We also examined the same maternal and neonatal outcomes in the women who were diagnosed with melanoma in the first postpartum year, and there also were no significant differences.
Table 4. Maternal and Neonatal Outcomes among Women with Malignant Melanomaa
No. of patients (%)
Age and race-adjusted OR (95% CI)
OR: odds ratio; 95% CI: 95% confidence interval.
Data were available for 148 patients.
Blood transfusion at delivery
Length of stay at delivery hospitalization > 5 days
Preeclampsia or eclampsia
Low birth weight
Very low birth weight
Readmission within 28 days
Hospital transfer after birth
Postpartum hospital stay > 5 days
Hospital stay > 5 days during first year
Hospital charge > $1,000
Total for first yr
Table 5 illustrates the occurrence rate of melanoma per 100,000 deliveries in California for each year from 1991 to 1999. There was no significant increase noted in the number of diagnoses of malignant melanoma in pregnant women in California during that period.
Table 5. Time Trend for Melanoma Associated with Pregnancy, 1991–1999a
No. of women
Occurrence rate per 100,000
No. of women giving birth
The 2-sided P value for the Cochran–Armitage trend test is 0.285. The women evaluated were those who were diagnosed 9 months before and at the time of delivery and within 1 year postpartum.
We also analyzed the types of surgery performed for the treatment of malignant melanoma and the date of surgery compared with the dates of diagnosis and delivery. Surgery data were available for 141 of 145 antepartum patients and for 257 of 263 postpartum patients. For patients who were diagnosed with melanoma antepartum (0–9 months prior to delivery; n = 141 women), the mean number of days between diagnosis and surgical removal was 14.3 days. In patients who were diagnosed postpartum (0–12 months after delivery; n = 257 women), 14.6 days was the mean between diagnosis and surgery. In contrast, for age-matched, nonpregnant women, the mean was 3.7 days between diagnosis and surgery. This was significantly lower than either group with pregnancy-associated melanoma (P < 0.001; t test) We also examined the rate of lymph node evaluation compared with tumor thickness in pregnant women compared with the nonpregnant control group. Table 6 shows that, among women who had pregnancy-associated melanoma that measured ≤ 0.75 mm in thickness, 90.9% did not have lymph nodes removed (examined), similar to the 93% of women with nonpregnancy-associated melanoma who had tumors with the same thickness. Among the patients who had tumors that measured 0.76–1.5 mm in thickness, 73.3% of pregnant women with melanoma, compared with 54.9% of postpartum women with melanoma and 68.9% of nonpregnant control women with melanoma, did not undergo lymph node removal; and, among patients who had tumors that measured > 1.5 mm in thickness, 50.0%, 52.0%, and 53.7% did not undergo lymph node evaluation, respectively. None of these differences reached statistical significance. At all depths of tumor thickness, the risk of metastases in the lymph nodes was similar between pregnant women and nonpregnant women. Small numbers of lymphadenectomies overall limited this comparison. With regard to sites of metastases, we were unable to obtain data on the specific sites of metastases in the 20 patients who had regional or remote disease during pregnancy.
Table 6. The Number of Regional Lymph Nodes with Evidence of Involvement by Tumor Thickness in Pregnant versus Nonpregnant Women with Melanoma in California, 1994–1999a
Breslow tumor thickness
Diagnosed 0–9 mos before and at delivery
Diagnosed within 1 yr postpartum
Age-matched nonpregnant women
LN: lymph node.
Fisher exact tests showed no significant differences between the pregnant group and the age-matched, nonpregnant group.
≤ 0.75 mm
No LNs examined
Missing LN status
No LNs examined
Missing LN status
> 1.5 mm
No LNs examined
Missing LN status
Missing Breslow tumor thickness
No LNs examined
Missing LN status
The median age at diagnosis of melanoma is 67 years, but up to 45% of patients are diagnosed in the first 4 decades of life.1 For women, the delay of child-bearing has caused an increase in the average age of the pregnant woman. Rates of melanoma in the general population also have been rising in the U.S. over the last 2 decades. Putting these factors together, there has been concern that the incidence of melanoma in pregnancy may rise. According to our data, the pregnancy-associated occurrence rate of melanoma in California was 8.5 per 100,000 pregnant women in the database. The National Cancer Institute SEER statistics from 1975 to 2000 show that the incidence rate in women ages 20–45 years ranges from 6.0 per 100,000 women to 21.2 per 100,000 women.1 The occurrence rate that we observed of 8.5 per 100,000 pregnant women does not reflect a true incidence in pregnancy in California due to our inability to capture women who had melanoma diagnosed during pregnancies that were not carried beyond 20 weeks. Evaluating our data over the individual years 1991–1999, we found no evidence for an increased rate of the diagnosis of this malignancy during pregnancy in California. This apparent lack of an increase in pregnant women, compared with the increase seen in the general population in the U.S. during this period, again, may be because our database did not allow us to capture melanomas in women who did not carry their pregnancies beyond 20 weeks. However, our study population consisted of women age < 50 years (child-bearing), and most of the increased incidence in melanoma in the general population has been in age groups > 50 years.1
Additional limitations to this population-based analysis of women with pregnancy-associated melanoma include the possibility that women with melanoma in pregnancy who did not carry their pregnancies beyond 20 weeks theoretically may have had more advanced tumors and/or a poorer prognosis, and this may have changed the pregnancy-associated population outcome as a whole, depending on whether those women were captured by the database. In addition, the nonpregnant control group in the current analysis ware generated from the database of women who did not have a pregnancy reported beyond 20 weeks. Some of those control women theoretically may have been in early pregnancy at the time of diagnosis. Also, as with any database, there is a small percentage of data missing that cannot be included in the comparison. However, the ability to document the outcome of over 400 women with pregnancy-associated melanoma to our knowledge represents the largest evaluation in the literature to date and can provide important insight into the neoplastic characteristics as well as the oncologic and maternal/neonatal outcomes for this unique group of women with primarily localized melanoma.
Because case reports in the 1950s suggested a poor outcome for pregnant women with melanoma, some authors suggested that melanoma may be a hormonally stimulated tumor and may manifest a more aggressive phenotype in women with high levels of hormones, such as those that occur in pregnancy.5 In the 1970s and 1980s, several retrospective reviews in small numbers of pregnant women also reported a worsened prognosis for patients with melanoma during pregnancy.11–13 Others found that estrogen and progesterone receptors can be detected in malignant melanoma tissue.14 However, to our knowledge, very few cases of malignant melanoma in pregnancy were actually documented in the literature prior to 1985. Since that time, there have been multiple studies documenting overall good outcomes for women with malignant melanoma during pregnancy. A report from the World Health Organization (WHO) melanoma program compared 92 women who were diagnosed with Stage I malignant melanoma during pregnancy with women who were diagnosed before or after their pregnancies.15 In that WHO study, it was found that the survival curve for women who were diagnosed during pregnancy was worse compared with women who were diagnosed and treated before pregnancy. In that analysis, however, the mean tumor thickness in the pregnant women was 2.38 mm, compared with 1.49 mm in the nonpregnant women (P = 0.002). After correcting for this difference, there was no difference noted with regard to survival. Another retrospective study from investigators at Duke University compared 100 pregnant women who had melanoma with 86 age-matched, nonpregnant women and found an increased number of lymph node metastases and a shorter disease-free interval in pregnant women with melanoma; however, those investigators found that the mean tumor thickness was 2.17 mm in the pregnant group, compared with 1.52 mm in the nonpregnant group (P = 0.05).16 Overall survival was not found to be significantly different between pregnant women and control women; and, in multivariate analysis, tumor thickness alone was a significant predictor of survival. In addition, a review of five controlled trials of malignant melanoma in pregnancy also revealed that tumors tended to be thicker in pregnant women than compared with controls and that pregnancy at the time of diagnosis was not a significant prognostic factor in multivariate regression analysis.17
The data presented herein from our investigation revealed no differences with regard to the mean or median tumor thickness between women who were diagnosed during pregnancy or during the postpartum period compared with nonpregnant women. In addition, and in contrast to any potentially worsened prognosis, we found that, for women who were diagnosed during or within 1 year after pregnancy, their survival was at least as good as the survival of nonpregnant women with melanoma.
There were several additional, reassuring results found in our data set. First, the timing of the diagnosis of melanoma either during or immediately after pregnancy did not appear to influence the risk of mortality. There also did not appear to be any iatrogenic prematurity or any significantly increased maternal or neonatal complications associated with the diagnosis of malignant melanoma in pregnancy. In addition, it is reassuring to find that at least 82% of the pregnant women who were diagnosed with melanoma were diagnosed with only localized disease, comparable to nonpregnant age-controlled patients. Again, it must be emphasized that this database would not have captured women who did not carry their pregnancies to at least 20 weeks; therefore, if women with advanced disease were more likely to terminate their pregnancies or otherwise suffer pregnancy loss before that time, then this percentage may be skewed toward disease with a more favorable prognosis. An additional reassuring finding was that, although there was a slightly longer time between diagnosis and surgical removal for both pregnant and postpartum women compared with nonpregnant control women who had melanoma, management was not altered in terms of the number of women undergoing lymph node evaluation and, because the outcomes were equivalent, no apparent clinical impact resulted.
Overall, the findings of the current study demonstrated good outcomes for the great majority of women with pregnancy-associated melanoma who were diagnosed in California between 1991 and 1999. The ability to diagnose and treat this disease with local surgery, as opposed to the need for radiation and chemotherapeutic approaches in most patients, allows for prompt diagnosis and treatment despite the pregnant state. In contrast to earlier reports, we did not find an increase in the incidence of thicker tumors during pregnancy, and we did not observe any negative impact on survival rates. There were no reported patients with placental metastases in the maternal discharge diagnoses; however, the rare occurrence of this complication of malignant melanoma would be expected to develop in pregnant women with advanced melanoma. Our study contained only 20 women with regional or remote disease.
We believe that the current study data from a large population database further contribute to the body of literature that dispels the concept that malignant melanoma is either more common, more aggressive, or more lethal during pregnancy. In addition, it supports the current recommended approach of early diagnosis and surgical treatment without delay, even if this occurs during pregnancy. There are no findings in the current data to conclude that continuing a pregnancy after the diagnosis and surgical removal of a localized melanoma would be harmful to either the mother or the fetus. The relatively few numbers of women in this study with either regional or remote disease precluded any conclusions regarding management in this group.
The authors thank David Harris of the California Cancer Registry and Barbara Burrall, M.D. (Department of Dermatology, University of California—Davis Medical Center), for their assistance in data preparation (Mr. Harris) and article review (Dr. Burrall).