Abstract
BACKGROUND:
Inflammatory breast cancer (IBC) is a rare and aggressive type of breast cancer (BC). Physicians have difficulty diagnosing it correctly given its clinical nature. Previous studies have shown that North Africa compared to the United States has a higher proportion of IBC relative to all BC.
PURPOSE:
The purpose of this study was to calculate a corrected IBC incidence rate using the population-based registry of Casablanca and other, local hospital-based data sources and to characterize the clinical presentation and basis of diagnosis of IBC.
METHODS:
We retrieved the Casablanca registry data from 2009–2012, and matched its data with demographic and clinical data from the medical records, logbooks and a local epidemiologic IBC case-control study. A corrected incidence was calculated after accounting for missed cases in the registry.
RESULTS:
The data showed that the incidence of IBC in the registry was significantly underestimated when the population-based and hospital data were combined.
CONCLUSIONS:
Population-based registries should focus on specific measures for verifying the diagnosis of IBC since physicians may miss documenting the disease in medical records. This study may have implications for better reporting and documentation of IBC in hospital- and population-based cancer registries in Morocco and other similar countries.
Introduction
Inflammatory breast cancer (IBC) is a rare but severe form of breast cancer (BC) [17]. The clinical characteristics of IBC include rapid onset of erythema (redness), edema (swelling), and peau d’orange (ridged or dimpled skin) of the breast [15]. It has been reported that IBC is more prevalent in younger women and is more likely to be diagnosed at advanced stages than other kinds of BC [2]. Because IBC is particularly aggressive, there is a shorter disease-specific survival period, with a median survival rate of less than 4 years in the U.S. [2,16]. In most cases, IBC is diagnosed based on clinical characteristics instead of pathological methods. It is often difficult to distinguish IBC from mastitis, a bacterial infection, and other types of locally advanced breast cancer. For all of these reasons, it is difficult to have standardized criteria for IBC diagnosis, given that laboratory-based pathologic confirmation is not possible, and that IBC may be mistaken for other cancers or diseases [6,9,13].
North Africa, including Morocco, has shown a higher proportion of IBC compared to other types of BC than other parts of the world. The proportion of all BC cases that are IBC in North Africa has been reported as 5–10%, while in the United States, it is around 1–2% [4,7,10]. Previous studies that initially suggested the elevated IBC incidence in Morocco may have had biases due to small sample size and lack of population-based recruitment and representation [10,13]. Whether or not the lower overall BC incidence in North Africa compared to western countries plays a role in the higher proportion of IBC is up for debate [12]. However, the possible higher proportion of IBC in Morocco may be due to risk factors that are specific to North Africa and indicate an actual elevated IBC incidence in the region.
Over the past several years, cancer data quality has been improved in Morocco because of the initiation of a population-based cancer registry in Casablanca. In addition, there are other sources such as a hospital-based registry and research studies that are available on IBC in Casablanca. Therefore, this study aimed at retrieving and analyzing the recently available multiple data sources on IBC in Casablanca to better characterize the incidence and clinical presentation of IBC in this population.
Methods
Registry data collection
The present study in Morocco used data from the population-based Casablanca Cancer Registry that covered the period from 2008 through 2012. The data in the registry are based on cancer cases among the ∼4 million residents of the city of Casablanca and the surrounding administrative area. In this study, analysis of existing datasets was conducted in order to calculate a more accurate incidence rate for IBC in Casablanca, by correcting for missing cases in the registry.
An electronic database that contained all de-identified cancer cases recorded in the Casablanca Cancer Registry that were collected by active registration for the five years from 2008 through 2012 was obtained as a CSV file, and imported into SAS. A subset of only BC cases, based on the WHO ICD-10 international topography classification, was extracted for analysis [8]. All variables were cleaned for inconsistencies and basic descriptive data were calculated. A subset of all BC cases that were IBC was also compiled, defined by being coded “T4d” for the TNM classification variable. The randomly generated patient IDs from the CanReg5 software used to collect cases in the registry were recorded. An Excel spreadsheet linking these cases to the case number of the medical records was then used to obtain each patient record. To obtain medical records, a list of patient ID numbers that had IBC according to the registry was given to the medical records staff, who then retrieved the corresponding record. Data were abstracted from each record, including height, weight, clinical symptoms, past surgery, tumor grade, tumor size, imaging that was available, human epidermal growth factor receptor (HER2) status, estrogen and progesterone receptor (ER and PR) status, age at menarche, occupation, and number of children. Our dataset was carefully checked to ensure that each medical record corresponded to its randomly generated CanReg5 ID number and that the recorded data remained de-identified. Estimates of sex-specific population in Casablanca based on data from the 2004 and 2014 Moroccan census were used to calculate incidence rates [1]. These estimates were available for the years 2008–2012, and a linear extrapolation of these data was used to estimate the sex-specific population for 2013.
Hospital-based data collection
Over the past 10 years, an epidemiologic IBC case-control study was developed and carried out by our group in Egypt, Tunisia, and Morocco. Data on IBC symptoms were collected, and cases were confirmed based on symptoms and clinical examinations at diagnosis by physicians participating in the case-control study [3,11]. Cases of IBC and their symptom data from the case-control study were recorded for the current study for the years 2010–2015. These cases were patients that were seen at Ibn Rochd Oncology Center, the site where the registry is located in Casablanca and part of University of Hassan II Hospital. Thus, epidemiologic case-control study cases may or may not have been included in the population-based registry, depending on whether or not the place of residence was reported by the case as Grand Casablanca region. Duplicate cases from 2 sources were eliminated by removing cases that had the same registry ID number. Only those cases identified as IBC with a recorded residence within the geographic area covered by the registry were included in the analysis.
Additional information that had been abstracted into an excel file was also available from logbooks concerning patients seen at Ibn Rochd Oncology Center. The original logbooks were paper records taken on site at or near the time of BC diagnosis. Variables in the excel version of the logbooks included date of diagnosis, birth year, sex, area of residence, histology of tumor, TNM classification, referring physician, and referring hospital.
Data analysis
A corrected proportion of IBC compared to all BC was calculated by including the additional number of cases present in the case-control study and logbooks, but not present in the registry. The year 2008 was eliminated in this calculation, because there were no data from either the case-control study or the logbooks. Furthermore, the registry cases for 2008 had considerable missing data, particularly for TNM classification. The years 2014 and 2015 were also omitted from this analysis given that neither registry nor logbook data was available for these years.
Data on documented symptom type from the case-control study were compiled, which showed cases that had already been diagnosed as IBC and also identified BC cases that were previously undiagnosed as IBC. Symptom data from cases already identified as IBC were combined with symptom data from registry case medical records. This was compared to symptom data from cases later identified as IBC in the case-control study to see if there were notable trends that could suggest any kind of bias in the way that cases were originally diagnosed.
The Health Sciences and Behavioral Sciences Institutional Review Board (IRB-HSBS) at the University of Michigan exempted the study from ongoing IRB review per exemption #4 of the 45 CFR 46.101.(b). This study was also exempt from ongoing review by the ethics committee at CHU Ibn Rochd in Casablanca.
Results
Breast cancer
The combined records represented a total of 4,796 cases of female BC in the Grand Casablanca region from 2008–2012. The mean age for these cases was 50.3 ± 0.3 years. The crude average annual incidence rate of female BC for the period 2008--2012 was 47.0 per 100,000 women. Age-specific BC incidence rates were calculated and tabulated in Table 2. Using these incidence rates, an age-standardized rate of 52.9 per 100,000 was calculated, using the World Standard Population from the WHO [17]. There were 3,424 (71.4%) cases that had some sort of staging available (SEER staging, TNM classification, or tumor extension) and 1,372 (28.6%) cases that had no staging of any kind. There were only 1,559 (45.5% of the those with staging) cases that had the TNM classification required to identify IBC, 1,399 (40.9% of those with staging) cases with SEER staging, and 1,789 (52.2% of those with staging) cases with tumor extension.
IBC cases by year from registry and hospital-based sources
IBC cases by year from registry and hospital-based sources
Age-specific female BC incidence rates
A total of 82 IBC cases were identified from all sources combined from the years 2009–2013. The corrected average annual IBC incidence rate was 0.79 per 100,000 women compared to 0.39 per 100,000 women when only using the registry data (calculated using cases from Table 1). This represents a 49.4% underestimation of IBC incidence. After finding an average number of BC cases per year (959) using totals from each year in the registry and an average number of IBC cases per year (16.4), IBC cases account for 1.7% of all BC. For all IBC cases, the mean age was 49.3 ± 2.1 years. A one sample t-test indicated that the age pattern of all BC cases was not statistically different than IBC cases (p = 0.347) when comparing the mean age for IBC and the mean age for all BCs in the registry.
Assessment of IBC symptoms
An additional 25 cases (107 total) were included from the 2014 and 2015 years of the epidemiologic case-control study data in the assessment of IBC symptoms. Symptoms reported in the medical records included swelling, redness, peau d’orange, breast retraction, and pain. Duration of symptoms was only available in the epidemiologic case-control study data. There were 22 (20.6%) IBC cases that had no medical record from which to abstract symptom data. There were 6 (5.6%) other IBC cases whose medical records contained no description of any specific symptoms, or recording of symptoms was incomplete in the case-control study. Of 79 cases with available symptom data (this included cases that only had “inflammatory signs” documented as a symptom with no further specific symptoms), 57 cases (72.2%) were originally diagnosed as IBC and 22 (27.8%) were later identified as IBC. As shown in Tables 3 and 4, the number of cases with each symptom was tabulated for both originally identified and newly-identified IBC. Table 3 shows counts when “inflammatory signs,” which was noted in some medical records as the only symptom data, was treated as no symptoms documented (method 1). With this method, cases that had only “inflammatory signs” documented in the medical record were not recorded as having any symptoms. Table 4 shows counts when “inflammatory signs” was treated as including redness, swelling, and peau d’orange, the hallmark symptoms of IBC (method 2). Thus, the counts and percentages for these 3 symptoms are higher in Table 4. Significant p-values for the chi-square testing were obtained when using this second method with each of the 3 main symptoms.
Recorded symptoms for originally and newly identified IBC – method 1
Recorded symptoms for originally and newly identified IBC – method 1
a Chi-square test.
Recorded symptoms for originally and newly identified IBC – method 2
a Chi-square test.
This study revealed three important observations. First, the study revealed that IBC incidence was underestimated in Casablanca by at least 49.4% for the period from 2009–2013. This was demonstrated by the addition of inflammatory status for cases from other local sources to the population-based registry. We recognized that the corrected incidence is likely still not an accurate measure of IBC incidence as will be discussed below. However, quantifying the underestimation of IBC incidence serves to demonstrate the need for more complete data in the registry and confirms that IBC is being underestimated when not including data sources outside of the registry. Second, the medical records for BC cases were not of sufficient quality to reflect the true IBC incidence in the registry. We found that 28.6% of cases did not have staging data of any kind. This is likely a source of missing IBC cases and another factor in the underestimation of IBC incidence from the registry alone. Third, a significant proportion of medical records for confirmed IBC cases included non-specific symptoms. On the other hand, some cases contained symptoms indicative of IBC but had no confirmed diagnosis in the medical record. This highlights the importance of clear documentation of IBC symptoms in the medical records in identifying IBC cases [4,11].
It is crucial to have a higher proportion of cases with available staging data in the registry in order to effectively assess the burden of specific subtypes of BC such as IBC. This is evident in other population-based cancer registries in developing countries as shown in our previous studies from the Gharbiah Cancer Registry in Egypt where no coding exists to easily identify IBC [4]. The Casablanca cancer registry also has a significant number of BC cases with missing staging data, therefore limiting its ability to accurately estimate IBC incidence [14]. Given the lack of staging data and specific registration coding for IBC, many cancer registries in developing countries are only able to report approximate proportions of IBC instead of accurate incidence. While still not capturing the entirety of the IBC burden in Casablanca, our study improves upon an IBC proportion by reporting a corrected incidence estimate. It is also important to note that our study does not assess the completeness of the registry for the purpose of calculating an incidence rate for all BC, but rather assesses the completeness of the registry for the purpose of identifying IBC. The corrected IBC incidence rate does not affect the incidence rate for all BC in the registry.
Even with the higher corrected IBC incidence rate from this study, we argue that there is still significant underestimation of IBC in Casablanca for a number of potential reasons. First, the registry does not contain “death certificate only” cases due to lack of access to mortality records. Without these cases, there is likely an underestimation of all types of BC including IBC. Second, there could be incomplete referral to Ibn Rochd Oncology Center from other departments within University of Hassan II Hospital. It is likely that a number of patients are seen for surgery or treatment in another department, such as obstetrics/gynecology without being seen at Ibn Rochd where information on staging is more likely to be documented. This highlights the issue of relying on medical records for case ascertainment, because IBC was not adequately documented in the medical records. It also shows the need for a specific ICD-O histology code 8530 to be used in the registry and when diagnosing IBC as done with the Surveillance, Epidemiology, and End Results (SEER) registry in the US [9]. Third, other diagnosing hospitals and laboratories covered by the registry that reported cases of BC reported very few cases of IBC, if any. This was due to the small percentage of cases from these diagnostic sources that had TNM classification or staging data of any kind and the clinical nature of diagnosis of IBC. Lastly, IBC has unique clinical characteristics which make it challenging to diagnose. Differing clinical criteria for diagnosis has led to difficulty in assessing the burden of IBC as well as comparing rates and proportions of IBC across studies and geographic locations that also have differing levels in the quality of medical records [3,11]. This was recently shown to play a role in the US, with some researchers finding potential IBC underestimation due to lack of standardized criteria when comparing clinical IBC diagnosis with documentation of IBC in SEER data [9]. This shows the difficulty of obtaining an accurate measure of the proportion of IBC to all BC even in a high resource setting like the US. It is likely that underestimation due to misdiagnosis of IBC is also occurring in Morocco. All of these reasons for underestimation underline the importance for physicians to accurately document IBC symptoms and staging in medical records.
The lack of clear documentation of specific symptoms and confirmed diagnoses of cases with symptoms indicative of IBC has an effect on the number of IBC cases that can be identified in the registry. This is not unusual given the structure of the healthcare system in Morocco and the lack of complete investigation in some cases to obtain staging information. No clear patterns were seen in the assessment of symptom data of IBC cases comparing originally and newly identified IBC. However, when “inflammatory signs” was treated as a symptom for each of the three main symptoms of IBC, significant differences in the symptom type were seen for redness, swelling, and peau d’orange. These main IBC symptoms should help medical practitioners in diagnosing IBC, as these are hallmark symptoms of the disease. However, it could also indicate the need for further investigation if a BC case does not clearly have all of these symptoms and still may be IBC. It is important to note that the sample size of this study is small and the significance of these findings should not be overstated. It is also probable that “inflammatory signs” in some cases did not truly mean all three of the symptoms were present in every case. Lastly, it is possible that cases with no symptoms available but still diagnosed as IBC actually had some of the diagnostic symptoms, but these were undocumented. This would also contribute to underestimation of IBC incidence.
The corrected IBC incidence in this study indicates underestimation of IBC incidence in Casablanca if the estimate is based solely on data from the population-based registry. The additional cases from other local sources in this study (hospital logbooks and epidemiologic case-control study data), which increased this incidence estimate from 0.39 to 0.79 per 100,000 women, produced an IBC incidence in Casablanca lower than that seen in the US [16]. This study also found that IBC for a smaller percentage of all BC than has been seen in previous studies in North Africa. However, these results must be taken with caution. Casablanca has only begun population-based cancer registration relatively recently, and it is unwise to assume these results reflect the true IBC incidence for reasons outlined earlier. Previous hospital-based studies used only proportions of IBC to other BC and may not have had the same issues with missing staging data and having to rely on medical records for case ascertainment. Before a definitive explanation for the elevated proportion of IBC in North Africa can be found, there is a need for more population-based studies that examine IBC incidence instead of proportions. The improvement of existing and future population-based registries in developing countries will be essential in calculating more accurate IBC incidence estimates that better reflect the burden of IBC in each region.
This study has notable strengths. First, the Casablanca Cancer Registry is the most comprehensive cancer registry in Morocco. It has the largest number of cancer cases in Morocco, and the population-based nature of the registry allows for an estimate of BC incidence. The registry was reviewed and validated by international experts and deemed to be of acceptable quality [18]. The size of the population covered under the registry and the quality of the data were unique for a region that has only recently begun widespread cancer surveillance efforts. It was also crucial to the success of this study that there were on-site staff trained in cancer registration and epidemiology. This facilitated the research team’s effort to understand and communicate the importance of population-level research when other clinicians may not have seen the utility in conducting a study that did not have an immediate direct benefit to patients.
This study also has some limitations. While the registry is a great source of quality data, the high percentage of missing data describing TNM staging made it difficult to identify IBC cases without investigating other data sources. Although additional cases were identified through another case-control study and oncology center logbooks, we did not capture the entirety of IBC incidence in Casablanca. This is due to the fact that the case-control study had cases only from a limited number of years (part of 2010 through 2015) and the logbooks were under review by registry staff at the time of the study and in the process of being abstracted into an electronic file. The fact that the epidemiologic case-control study was only up and running for part of 2010 may have led to the underestimation of cases for that year, and thus the underestimation of the average annual incidence. The small sample size of IBC cases was an issue when attempting to calculate the corrected incidence, however this is to be accepted when dealing with a rare cancer such as IBC. One final limitation was the availability of paper medical records when trying to ascertain case-specific data and confirm IBC diagnosis. A transition from paper to electronic medical records, while logistically difficult and time consuming, would be helpful both to clinicians in their current work and for future population level health research.
Conclusion
In this study, we assessed the burden of IBC in Casablanca using a population-based registry and other hospital-based sources. We also documented the underestimation of IBC incidence and evaluated the quality of the medical records used to identify IBC for the registry. Population-based registries should focus on specific measures for verifying the diagnosis of IBC since physicians may miss documenting the disease in medical records. Oncologists should also receive training to accurately diagnose and document IBC in medical records. This study may have implications for better reporting and documentation of IBC in hospital- and population-based cancer registries in Morocco and other similar countries. This study demonstrates the usefulness of population-based cancer surveillance in developing settings such as Morocco while also highlighting some potential areas for improvement.
Footnotes
Acknowledgements
The project described was supported by Grant Number R25CA112383 from the National Cancer Institute. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health. Funding for this project was also provided by the Office of Global Public Health at the University of Michigan.