To the Editor:
The observational study by Trinh et al.(1) concludes that there is a higher recurrence rate when breast cancers are detected during the use of a LNG-IUS and when it is not removed after diagnosis. Although the authors mainly generate several hypotheses around their observations, we would like to warn about inappropriate interpretation of their data and premature conclusions, as such a message was released in a couple of Belgian newspapers recently.(2)
First, there is a problem with the control group in the authors’ study. This control group is matched for ‘all’ LNG-IUS users, including a group with a low risk for relapse, those where the clinician decided to insert a LNG-IUS probably long after breast cancer diagnosis, and a group with a higher risk for relapse, those diagnosed with breast cancer during the use of the LNG-IUS.
Understandably, the high risk for relapse group (more chemotherapy, more ER-negative, more grade 3), also being the LNG-IUS users at breast cancer diagnosis, presented with a poorer outcome than the controls. To proove their point, cases with the LNG-IUS at breast cancer diagnosis should have been matched for controls without LNG-IUS at diagnosis and followed up taking into account demographics like family history, but also prognostic and predictive factors and adjuvant therapy. Furthermore, the authors hypothesize that breast cancers in LNG users are detected in a much more advanced stage based on differences, for example, in nodal involvement between the group with the LNG-IUS at diagnosis and those where the LNG-IUS was inserted after diagnosis. They now define a second control group (subgroup B), which is inappropriate.
The question of whether breast cancers in LNG-users are in a much more advanced stage at diagnosis is, therefore, not proven by this observation, as the control group is likely biased by selection for the use of a LNG-IUS after breast cancer diagnosis to a lower breast cancer relapse risk population. We believe that a case-control study is more appropriate to prove this point.
We also do not know whether breast cancer prognosis is different when the LNG-IUS is removed at diagnosis, an important group which is missing in this study. It is also unclear how long the LNG-IUS was inserted prior to or after the diagnosis of breast cancers respectively in subgroup A and subgroup B. We understand that patients in subgroup A could have had the LNG-IUS for over 14 years, whereas those in subgroup B might have had the LNG-IUS inserted several years after their breast cancer diagnosis. This may, again, imply bias through a healthy user effect that may affect relapse of breast cancer. It may also explain local versus distant metastasis between subgroups and controls. Finally, both Kaplan Mayer curves suppose a much longer mean follow-up than the stated 2.8 and 2.9 years, respectively, for LNG-IUS users and for controls.
Based on these observations, the hypotheses on neither safety nor prognosis of the LNG-IUS in operable breast cancer have been generated. It remains an open question whether low levels of circulating levonorgestrel affect the natural history of breast cancer at diagnosis and follow-up, or whether it interferes with adjuvant anti-estrogen therapy. We continue (3 )to agree there is need for further research and welcome their prospective registration of LNG-IUS. This is an important issue not only because the LNG-IUS is a very popular method of contraception, and 2% of women in Belgium do develop breast cancer during fertile years, but also because the LNG-IUS is now being used in long-term estrogen (4) and tamoxifen users (5) to protect the uterus from bleeding and proliferative endometrial changes.Patrick Nevena
Rudy Van den Broeckeb aDept of Obstetrics and Gynaecology
University Hospital KU
bDept. of Obstetrics and Gynaecology
University Hospital RU
1. Trinh XB, Tjalma WAA, Makar AP, Buytaert G, Weyler J, van Dam PA. Use of the levonorgestrel-releasing intrauterine system in breast cancer patients. Fertility and Sterility 2008; 90: 17-22.
2. Article in Gazet van Antwerpen on October the 16th: ‘Spiraaltje nefast bij borstkanker, onderzoekers vragen bijsturing’: ‘Hormone coil is dangerous in breast cancers patients, researchers do request action to prevent harm’.
3. Neven P. Local levonorgestrel to prevent tamoxifen-related endometrial lesions. Lancet 2000; 356: 1698-9.
4. Chan SS, Tam WH, Yeo W et al. A randomised controlled trial of prophylactic levonorgestrel intrauterine system in tamoxifen-treated women. BJOG 2007; 114: 1510-5.
5. Wildemeersh D, Pylyser K, De Wever W, Pauwels P, Tjalma W. Endometrial safety after 5 years of continuous combined transdermal estrogen and intrauterine levonorgestrel delivery for postmenopausal hormone substitution. Maturitas 2007; 57: 205-9.
Published online in Fertility and Sterility DOI: 10.1016/j.fertnstert.2008.12.067
The Authors Respond:
We thank Neven and colleagues for their letter. We agree that our published data do not provide robust evidence to draw firm conclusions. The abstract clearly states: “Overall, we did not find an increased risk of breast cancer recurrence associated with the use of the LNG IUS. However in a subgroup analysis of women who developed breast cancer while using an LNG IUS and who continued to use the LNG IUS, we found a higher risk of recurrence of borderline statistical significance. Additional research is needed to confirm or refute these findings.”
The remarks that Neven express about our study concern mainly the subgroup analysis that we did and factors which are inevitably inherent to a retrospective study.
To study patients who had their LNG IUS removed at time of diagnosis were beyond the objectives of our study. As Neven pointed out, it is indeed an interesting group. Yet, we believe it is the merit of our study that is has contributed to this interest.
The manuscript clearly state that the subgroup analysis was an additional exploratory analysis. A Cox regression model is nonetheless able to correct for possible confounders. Letting study observations or results influence or change predefined study methodology is not our policy. Instead, we decided to end the study and initiate a prospective study. In this setting we are able to collect data that were not (or not consistently) available in a retrospective setting (e.g., time of insertion of LNG IUS, which was usually lacking in subgroup A patients). The mean follow-up time of the control cohort was 2.8 ± SD 2.0 years with a range of 3 months-8.1 years. For the control cohort it was 2.9 ± SD 2.7 years with a range of 3 months-8.2 years.
Although Neven et al. suggest that it is obvious that the subgroup B patients are a selected group with low risk for recurrence, we are not sure this is completely true. In our study, we only could identify a minority (3/41) of patients who had their LNG IUS inserted >5 year after initial breast cancer diagnosis. Furthermore, 97.4% of all LNG-IUS users (39/41 in subgroup B) in our analysis were premenopausal at time of initial breast cancer diagnosis, and therefore were by definition at “high risk” for recurrence.
Regardless and independent of the control cohort of which Neven et al. question the relevance, subgroup A seems to have unfavorable characteristics. We found this observation noteworthy to be reported.
We conclude with the same last paragraph we wrote in the article: “Because this was a retrospective analysis with a number of limitations, a prospective cohort study is being initiated in Belgium to follow premenopausal breast cancer patients, including data on contraceptive use (e.g. duration and type) and other background risk factors that were missing in this study. This should help clarify whether the results of this study were a result of chance or confounding or a result of a treatment effect.”Xuan Bich Trinh, MDa,b
Amin Makar, MD, PhDc,d
Guy Buytaert, MDe
Joost Weyler, MD, PhDf
Peter van Dam, MD, PhDa
aDepartment of Obstetrics and Gynaecology
General Hospital St. Augustinus
bDepartment of Gynaecologic Oncology
University Hospital Antwerp
cDepartment of Gynaecologic Oncology
The Middelheim Hospital
dDepartment of Gynaecologic Oncology
University Hospital Ghent
eDepartment of Obstetrics and Gynaecology
fDepartment of Epidemiology and Social Medicine
University of Antwerp
Published online in Fertility and Sterility DOI: 10.1016/j.fertnstert.2008.12.068