Revue Medecine , Volume 13, numéro 8, Octobre 2017
Organized screening for breast cancer is still controversial. According to its supporters, it should enable a decrease of treatments, due to earlier diagnoses. This assumption has never been verified in France. The "Programme de médicalisation des systèmes d'information" (PMSI), is an exhaustive census of the surgical treatments achieved in France. Any change in the number of mastectomies for breast cancer must therefore be found in PMSI.
We searched the PMSI databases and found that no decrease in mastectomies, total or partial, can be reported after the generalization of organized screening.
Keywords: breast cancer; mass screening; mastectomy
Robert Vincent 1, Jean Doubovetzky 2, Annette Lexa 3, Philippe Nicot 4, Cécile Bour 5
1 Hôpitaux Robert Schuman, Département d'information médicale (DIM)- Luxembourg
2 Médecin généraliste, rédacteur senior à la Revue Prescrire, Albi
3 Docteur en toxicologie (Eurotox), Metz
4 Médecin généraliste, expert à la HAS, Panazol
5 Radiologue libéral, présidente de l’Association Cancer Rose, Talange
Organized breast cancer screening was generalized in France in 2004. However, it is still a subject of controversy.
In particular, the postulate stating that organized screening allows earlier diagnosis and detects smaller tumors, leading to less aggressive treatments, has not been evaluated in France.
If this really was the case, in parallel with the generalization of screening, we should see a decline in the heaviest treatments, and in particular of total mastectomies. Since 1997, all surgical operations carried out in France during hospitalization have been recorded as part of the Information Systems Medicalization Program (PMSI) . The aim of our work is to verify whether, the generalization of organized breast cancer screening in France has been accompanied by a reduction in the number of the most mutilating surgical operations, by studying the annual progression of the number of mastectomies for cancer recorded within the framework of the PMSI.
Materials and methods
The number of mastectomies was estimated from data extracted from the ScanSanté PMSI data recovery platform . The research covered all the territories in France and all types of establishments, both public and private.
The ScanSanté queries were carried out on :
- "Total mastectomies for malignant tumors" (root of GHM 09C04);
- “Subtotal mastectomies for malignant tumors" (root of GHM 09C05)  .
The number of total mastectomies was lowered by the number of "prophylactic breast ablations".This means that admissions with a main diagnosis Z40.00 were subtracted from admissions classified in the root 09C04.
The PMSI is generally only considered exhaustive from the year 2000 onwards. Data prior to the year 2000 are therefore potentially underestimated and have not been taken into account.
Raw data on the number of mastectomies have been completed by estimating the ratios [annual number of mastectomies/annual incidence of breast cancer]. These ratios could only be calculated for the four years 2000, 2005, 2010 and 2012, for which the incidence of breast cancer in France is available .
The number of total mastectomies performed in the last 4 years prior to organized screening (period 2000 to 2003) was compared with the number of total mastectomies performed in the last 4 years with organized screening (period 2013 to 2016).
The trend in the annual number of mastectomies was investigated using a graphical method with linear trend line, completed by a Spearman rank correlation test. For comparisons between 2 periods, confidence intervals were determined by bootstrap resampling with 105 replicates. The significance threshold considered is the usual threshold of 0.05.
All the statistical analyses were carried out using R software version 3.0.2.
As shown in Table 1 and Figure 1, the annual number of total mastectomies for cancer tends to increase over the period 2000 to 2016.
This upward trend is statistically significant (p < 0.0002 on Spearman's rank correlation test).
The comparison of the last 4 years without organized screening (period 2000 to 2003) with the last 4 years with organized screening (period 2013 to 2016) shows that during the period 2013 to 2016, on average 1,615 more total mastectomies were performed each year compared to the period 2000 to 2003 (95% CI: 1010-2280).
Click to enlarge
Table 1 and Graph 2 show that the number of mastectomies of all types (total + partial) is also increasing considerably between 2000 and 2016.
This upward trend is statistically significant (p < 10-7 in Spearman's rank correlation test).
A comparison of the last 4 years without organized screening (period 2000 to 2003) with the last 4 years with organized screening (period 2013 to 2016) shows that during the period 2013 to 2016, on average, 13,389 more mastectomies (all types) were performed each year compared to the period 2000 to 2003 (95% CI: 9,610-17,160).
Table 2 shows the ratios [number of mastectomies / incidence of breast cancer]. These ratios are stable, between 0.38 and 0.41, in the case of total mastectomies.
The data show an increasing trend between 2000 and 2012 for partial mastectomies and mastectomies of all types.
Click to enlarge.
The PMSI is the only comprehensive source of information on surgical activity in France. It is mainly used for invoicing hospital stays. However, previous studies have shown that the PMSI can be used for epidemiological analyses, particularly regarding breast cancer [3,4].
If organized screening had been accompanied by a decrease in the number of mastectomies for cancer performed in France, this decrease should be reflected by an equivalent decrease in the number of mastectomies recorded in the PMSI databases. However this is not consistent with the observed results. On the contrary, there is a statistically significant increase in total mastectomies as well as all mastectomies of all types combined.
The stability over time of the ratios between total mastectomy / breast cancer incidence shows that the increase in the number of total mastectomies is parallel to the increase in the number of diagnosed invasive breast cancers. Thus, in 2012, 8 years after the generalization of organized screening, there still were practiced 4 total mastectomies for every 10 new invasive breast cancers, exactly as in 2000, before the generalization of organized screening.
The increasing trend over time in the ratios of partial mastectomies/ breast cancer incidence and ratio of mastectomies of all types / breast cancer incidence suggests that the increase in interventions involving partial mastectomies for cancer is higher than the increase in the incidence of invasive breast cancer.
For example, in 2012, there were nearly 15 procedures involving mastectomy for 10 new invasive breast cancers, compared to less than 13 procedures for 10 new cancers in 2000.
One of the possible explanations could be the over-treatments associated with over-diagnosis. Indeed, according to the authors, in 30% to 52% of cases, screening leads to the discovery of small or slowly progressing cancerous lesions, which would have remained asymptomatic throughout the patient's life. As a precaution, all lesions are treated and their discovery leads to their surgical removal, most often by partial mastectomy [5,6].
It is also possible that partial mastectomies secondarily followed by total mastectomies are being performed in higher number; in other words, it is possible that a certain number of women who were in the past directly treated by total mastectomy are now undergoing a partial mastectomy prior to undergoing secondarily a total mastectomy. However, this mechanism could only play a very small role, since according to the National Health Insurance Fund, in 2012 only 3% of lumpectomies were followed by mastectomy .
Hospitalization services are better remunerated for total mastectomies (root 09C04) than for partial mastectomies (root 09C05). It may also be speculated that there is an over-coding bias in total mastectomies, related to financial motivation. This interpretation is unlikely, for at least two reasons:
- On the one hand, if there were an abusive increase in coding for "total mastectomy" (09C04), it would be done to the detriment of coding for "partial mastectomy" (09C05), which should therefore be decreasing. However, the data available on ScanSanté show that partial mastectomies are on the contrary increasing, and even faster than total mastectomies.
- On the other hand, overcoding would change the balance between the number of total mastectomies recorded in the PMSI and the number of breast cancers. This would result in an increase in the ratios of total mastectomies/ breast cancer incidence, parallel to the increase in total mastectomies. On the contrary, our results show a stability of these ratios.
Our results are consistent with those found in other countries :
- In the United States, in a 2015 study of 16 million women, a 10% increase in screening activity was associated with a nearly 25% increase in lumpectomies and partial mastectomies (RR 1.24; CI 1.15-1.34), without a decrease in total mastectomies .
- In the United Kingdom, according to the 2013 Marmot report on breast cancer screening, the frequency of mastectomies is increased by about 20% in the screened population compared to the unscreened population .
- For all the randomized controlled trials worldwide that examined this issue in 2013, the Cochrane Collaboration estimates that the number of mastectomies is increased by 20% (RR 1.20; 95% CI 1.08-1.32) and the number of surgical procedures (mastectomies and lumpectomies) is increased by 30% (RR 1.31; 95% CI 1.22-1.42) .
Conclusions for practice
PMSI data show that there has been no reduction in surgical treatment of breast cancer in France since the generalization of organized screening.
On the contrary, both the number of total mastectomies and the number of partial mastectomies continue to increase despite current recommendations to favor conservative surgery whenever possible. Whether organized screening is contributing to this increase through over-diagnosis should be investigated.
The authors declare that they have no conflict of interest in relation to the content of this article.
- Manuel des Groupes Homogènes de Malades - Version 2016 de la classification. Bulletin officiel No 2016/5 bis Fascicule spécial.
- Binder-Foucard F, Belot A, Delafosse P, Remontet L, Woronoff AS, Bossard N. Estimation nationale de l’incidence et de la mortalité par cancer en France entre 1980 et 2012. Partie 1 – Tumeurs solides. Saint-Maurice (Fra) : Institut de veille sanitaire, 2013. 122 p. (Page 55 du document)
- Trombert Paviot B, Gomez F, Olive F, Polazzi S, Remontet L, Bossard N et al. Identifying Prevalent Cases of Breast Cancer in the French Case-mix Databases. Med. 2011;50(2):124-30. doi : https://doi.org/10.3414/ME09-01-0064
- Quantin C, Benzenine E, Hägi M, Auverlot B, Abrahamowicz M, Cottenet J et al. Evaluation de l’intérêt de l’utilisation des données du PMSI pour l’estimation de l’incidence du cancer du sein dans deux départements français. Rev Epidemiol Sante Publique 2010 ; 58 (Suppl. 1) : S14.
- Jørgensen KJ, Gøtzsche PC. Overdiagnosis in publicly organised mammography screening programmes: systematic review of incidence trends. BmJ. 2009 Jul 9;339:b2587.
- Bleyer A, Welch HG. Effect of three decades of screening mammography on breast-cancer incidence. N Engl J Med 2012;367:1998-2005.
- Améliorer la qualité du système de santé et maîtriser les dépenses : proposition de l’Assurance Maladie pour 2015. Rapport au Ministre chargé de la sécurité Sociale et au Parlement sur l’évolution des charges et des produits de l’Assurance maladie au titre de 2015 (loi du 13 août 2014) ; page 62.
- Harding C, Pompei F, Burmistrov D, Welch HG, Abebe R, Wilson R. Breast Cancer Screening, Incidence, and Mortality Across US Counties. JAMA Intern Med. 2015 Sep;175(9):1483–9.
9- Baum M. Harms from breast cancer screening outweigh benefits if death caused by treatment is included. BMJ 2013 ; 346 : f385. doi : https://doi.org/10.1136/bmj.f385
10- Gøtzsche PC, Jørgensen KJ. Screening for breast cancer with mammography. Cochrane Database of Systematic Reviews 2013 ; 4;(6) : CD001877. doi: https://doi.org/10.1002/14651858.CD001877.pub5
 Developed progressively from 1982 onwards, the PMSI aims to describe hospitalizations by classifying them into groups with similar characteristics, the Homogeneous Groups of Patients (HGP). Since 1995, public establishments have been required to provide their data. In 1997, the PMSI was extended to private establishments. It is generally considered that data became quasi-exhaustive from the year 2000 onwards.
 The PMSI data are publicly available to all, health care actors and the general public, on a national platform with open access, managed by the Agence technique de l'information sur l'hospitalisation (ATIH). This platform is called ScanSanté and can be consulted at: https://www.scansante.fr/
 At the end of a patient's hospitalization, the reason for admission, co-morbidities and procedures performed during the stay are coded to enable their electronic processing. Each hospital stay is then classified in a GHM root, based on the reason for admission and any surgical procedures performed. Thus, the root of GHM 09C04 corresponds, by definition, to hospital stays with an admission reason related to breast cancer and including a total mastectomy surgery. Similarly, the root of GHM 09C05 corresponds, by definition, to hospital stays with an admission reason related to breast cancer and involving a partial mastectomy.
Overall, there is as much surgery in 2016 as there were in 2000. In the official recommendations, there should be an increasing tendency to favor conservative surgery whenever possible. It is certain that, despite this declared willingness to favor conservative surgery, the annual number of total mastectomies is not decreasing. What we have also shown is that the number of total mastectomies in relation to the incidence of invasive cancers has not decreased either.
There are two possible explanations:
- either surgery is not more conservative in 2016 than in 2000 (i.e. recommendations favoring conservative surgery whenever possible are not followed; this negates the expected benefit of screening),
- or total mastectomies are performed for non-invasive tumors (over-treatment associated with over-diagnosis) and these additional total mastectomies cause the loss of the benefit of a general trend towards more conservative surgery.
READ MORE :