• © 2009 by American Society of Clinical Oncology

Reply to M. Lens et al

We appreciate the opportunity to respond to the thoughtful comments by Lens et al¹ on our recently published report.² Their letter concludes that on the basis of three large studies,^2–4 pregnancy at time of diagnosis of malignant melanoma does not appear to influence prognosis. It is reassuring that the estimate for cause-specific survival reported by Lens et al in their letter on the basis of the Swedish study³ findings is as low as 1.17, with no difference between those pregnant or not pregnant at diagnosis.

Because we analyzed cause-specific survival not only in patients with malignant melanoma but in patients with other tumor types as well, our primary multivariable Cox proportional hazards regression model only adjusted for age at diagnosis (in 5-year intervals), time period (1967 to 1984, 1985 to 1994, and 1995 to 2004), and extent of disease (localized, regional, and distant metastases), as described in our report.² To explore the issue additionally, we performed subanalyses including Breslow thickness and tumor localization in patients for whom this information was available. Our estimates, unadjusted for thickness or body site, showed borderline significance for those pregnant at diagnosis. However, with adjustment for body site, hazard ratio (HR) was reduced to 1.45 (95% CI, 0.96 to 2.21), which underscores the conclusion drawn by Lens et al¹ (ie, no difference in surivival yielded).

Breslow thickness was first described in 1970, but it was not included as a compulsory measure on pathology reports for clinical use until several years later.⁵ Because our study² included patients diagnosed as early as 1967, Breslow thickness was available only for 55% of women pregnant at diagnosis and for a similar proportion of the corresponding control patients. In comparison, Breslow thickness was available for approximately 40% of patients in the Swedish study by Lens et al.³ Because Breslow thickness today is considered the most crucial prognostic factor for malignant melanoma, we agree that lack of information is a major limitation in studies analyzing survival, in which adjustment for this possible confounder is impossible. However, such information has been unavailable, at least in registry-based series. Nevertheless, we believe that a nonselected study population, such as ours, along with a large number of patients, can provide valuable information in addition to smaller hospital-based studies, which allow for a greater number of prognostic factors.

In the past, the Cancer Registry of Norway database did not provide TNM grading. The TNM system has included Breslow thickness for several years, but this measure has not been routinely registered and is available only from histology reports. In the future, a number of important prognostic factors will be available because the Cancer Registry of Norway established a malignant melanoma clinical registry in 2008, which includes tumor thickness as one of the routinely collected variables.⁶ On the basis of information about Breslow thickness in a subgroup of the pregnant patients and a similar proportion of the nonpregnant control patients, we found no difference in tumor thickness between the groups.² Also, because information was prospectively registered, it is quite unlikely that there would have been systematic misclassification of material associated with the end point. In supposedly egalitarian societies like Norway and the other Nordic countries, public health care is readily available and inexpensive. The aspect of patient or doctor delay is one that might have influenced time from onset of symptoms to diagnosis and therefore tumor thickness. Perhaps diagnostic delay rather than pregnancy-related growth factors could explain the thicker tumors in pregnant women reported in some studies.^7–9 The only characteristic difference in pregnant women in our study was body site distribution, with a poorer prognostic site in 54% of pregnant patients compared with 41% in nonpregnant control patients.

Several background analyses were undertaken to explore our findings² in detail. For example, we performed analyses of factors including educational level at diagnosis and parity. To evaluate parity, we grouped patients according to whether they had had zero, one, or two or more pregnancies before diagnosis, and for education, patients were grouped on the basis of low, medium, or higher level of education. To adjust for sociodemographic parameters, educational level at diagnosis was used in this study. There was no association between outcome and number of pregnancies before diagnosis. However, educational level influenced mortality, with crude HR significantly reduced in women with medium or higher level of education at time of diagnosis. When educational level and tumor body site were incorporated into the multivariable Cox model, the HR for pregnant women at diagnosis was reduced to 1.48 (95% CI, 0.97 to 2.24).

Studies on this topic reporting survival for pregnant and lactating women usually combine these two groups and include the first year postpartum. For comparison, we performed multivariable Cox analyses including both these groups (total number of women, 286; lactation period, 6 months) compared with 4,460 nonpregnant women. This resulted in an HR of 1.33 (95% CI, 0.95 to 1.86). This additionally underscores the conclusion expressed in the letter from Lens et al.¹

In the future, we encourage cross-national collaborative studies to shed more light on whether our standpoint today about malignant melanoma during pregnancy is controversial. Some smaller prospective studies have been published, revealing that malignant nevi do not appear to show significant tumor changes during pregnancy.¹⁰ On the other hand, changing nevi should always promptly be biopsied. Despite our findings of borderline significant elevated risk of cause-specific death in the pregnant group, our additional analyses seem to add to the notion that pregnancy makes no difference in prognosis for patients diagnosed with malignant melanoma.