The extensive validation of the CCI as a powerful predictor of clinical outcome combined with its simplicity and widespread use in oncology have led to the adoption of the Charlson Index as the prototypical comorbidity index in our department. In addition to validity, our criteria for the use of a comorbidity index focus on simplicity in design, consistency in scoring, and ease of use. It is our contention that many of the commonly used comorbidity indices, such as the ICED, CIRS, and KFI have failed to achieve widespread use because they remain complicated, cumbersome to use, and poorly accessible for use in clinical research. Given the adaptability of the CCI for the inclusion of additional variables, such as age, the CCI also demonstrates marked potential for modification into cancer specific comorbidity indices. We have, therefore, developed a Charlson Comorbidity Calculator based on a Microscoft Excel File to improve the collection of comorbidity data in our department.
Comorbid illness has demonstrated increasing importance as a prognostic factor for survival and treatment-related outcomes in oncology. It confounds the results of clinical trials because the lack of a standardized measurement has resulted in the failure to adjust for comorbidity in statistical analysis of outcomes data [1
]. It also limits the applicability of clinical research to large segments of the oncology population because protocol designs tend to exclude older and sicker patients [21
]. Recent reviews consistently identify the CCI, ICED, CRIS and KFI as validated and acceptable measurements of comorbidity and recommend their use in clinical research. Although the ICED, CIRS, and KFI obtain superior prognostic power in some series, the CCI consistently demonstrates statistical validity, particularly in terms of prognostic validity, and remains the most structurally simple, easy to use and well-defined of the comorbidity indices. The ICED and CIRS, for example, both require coding manuals and training courses to be used effectively. The KFI has required extensive modification for use in oncology because it was originally designed to assess comorbidity in diabetic patients. Recent modifications of the KFI for use in oncology, such as those applied by Piccirillo in a head and neck cancer specific modification of the KFI (available in electronic calculator format at http://oto.wustl.edu/clinepi/calc.html
) also require training courses for effective use [12
]. By contrast, the Charlson Index is intuitive, requiring users to select a condition from a defined list, rather than searching for disease value or specific information about disease severity. In our department, the cumbersome requirements for use of the ICED, KFI, and CIRS would reduce compliance with collection of comorbidity data. Furthermore, the increased training requirements and intricacies of these indices may increase variability between scores, as it is unlikely that a single staff member would be responsible for the collection of all data. It is, therefore, our belief that the Charlson Index represents the optimal balance between ease of use and prognostic ability and has, therefore, become the method of choice for the collection of comorbidity data in our department. Accordingly, we developed the CCI calculator to improve compliance with the collection of comorbidity data and as a quality assurance tool to ensure that such data is collected correctly and uniformly.
The use of comorbidity data in clinical research is at an important crossroads, with necessity of its use becoming imperative as electronic capabilities for its assessment become more feasible. As the US population gets older, the use of comorbidity data in clinical trials will only increase in relevance. Current estimates indicate that the elderly will comprise 20% of the population by the year 2030 [23
]. Studies of older oncology patients also suggest that the elderly shoulder the majority of cancer burden, with risk rates 11 times greater than those of younger patients, with over 50% of all cancer-related mortality [24
]. The rise of comorbidity with increasing age is a theme common to most retrospective studies of comorbidity. In this light, determining the effect of comorbidity on cancer-related survival and treatment-related complications has become increasingly important.
Furthermore, evidence to suggest that comorbidity and performance status represent independent prognostic factors is accumulating. Extermann, et al, for example, examined the relationship between comorbidity and performance status. Both Charlson and CIRS-G were found to have little or no correlation with ECOG performance status, activities of daily living (ADL), or instrumental activities of daily living (IADL). More recently, Repetto, et al, found that among 269 elderly cancer patients with a reported ECOG performance score of <2, 13% had 2 or more comorbidities, 9.3% had ADL limitations, and 37.7% had IADL limitations. Although a statistical correlation between ECOG performance status, number of comorbidities, and comprehensive geriatric assessment was identified in univariate analysis, only comorbidity, ADL limitation and IADL limitation maintained statistical significance in multivariate analysis. Firat, et al also found CIRS-G and Karnofsy performace status to be independent predictors of outcome in their analysis of prognostic factors in 112 patients enrolled on 4 RTOG trials of stage III lung cancer [11
Without widespread integration of comorbidity data into clinical research, an increasing number of elderly patients, and their physicians, will be left with treatment recommendations and outcomes data that lack relevance for their age and level of comorbidity.
Concurrently, electronic medical records (EMR) and data collection systems are becoming increasingly common and easy to use, with EMR use among European countries approaching 60% to 90% [27
]. The EMR ultimately promises increased physician efficiency and improved clinical outcomes for patients. Contemporary EMR systems have improved outcomes by reducing errors with the use of electronic prescribing systems and improving preventative care with automated reminder systems [28
]. The MS Excel CCI Calculator provided with this manuscript, for example, could easily be integrated into an EMR for aid in data collection. Such integration would eventually provide an enormous data pool on which to base future research on the prognostic importance of CCI.
To our knowledge, this is the first electronic data collection system offered for the Charlson Comorbidity Index. The simplicity of the index itself, coupled with the simplicity of MS Excel and the Visual Basic programming language, have resulted in a robust electronic CCI calculator that functions well across both Windows and Macintosh platforms. The latest version of the calculator, which is provided with this manuscript, has performed without error consistently on the first (WH), second (RR) and third (SN) authors' Windows-based PCs.
The major limitations of the CCI calculator lie in the limitations known to comorbidity indices and to the index itself. These include lack of understanding as to the relative importance of various individual conditions on mortality, treatment-related complications and quality of life. Furthermore, failure to include some conditions with particular relevance to cancer patients, such as non-malignant hematopoietic disorders and thromboembolic disorders, as well as uncertainty as to whether a few specific diseases or the overall disease burden is more important for prognosis, remain important considerations limiting use of the CCI [2
]. Additionally, the CCI has a tendency to underscore comorbidity because it is limited to 19 conditions and because it excludes the primary malignant condition. For example, in a patient with localized prostate cancer, history of COPD and myocardial infarction, the CCI score calculated by a urologist would exclude prostate cancer from the calculation resulting in a score of 2. The same patient might receive a score of 3 by a cardiologist because myocardial infarction, as opposed to prostate cancer, was excluded from the calculation. Another limitation of the CCI lays in the frequent use of grouped CCI scores, or CCI grades, rather than the use of scores as continuous variables. Within an elderly cohort in whom comorbidity is likely to be high, the CCI will have reduced utility if it lacks the ability to distinguish between a score of 2, representing mild to moderate comorbidity, and a score of 8, representing severe comorbidity. With this limitation in mind, we recommend the use of CCI score as a continuous variable.
Despite its limitations, the general oncology literature supports the use of CCI as a prognostic variable in clinical research. It should be emphasized that the CCI is not meant to replace clinical experience and its use in clinical decision-making should be considered investigational. With additional research, CCI methodological limitations can be addressed and the index modified to improve upon its utility. In an effort to improve our understanding of the CCI and identify areas of the index in need of improvement, we are currently investigating the effect of score thresholds on treatment decision-making among prostate cancer experts. We believe that dissemination of the MS Excel CCI Macro will lead to increased use of the CCI for clinical research purposes as well as modification of the CCI to increase its validity and clinical utility. Ultimately, we hope that the comorbidity indices, such as the CCI, will see widespread use in clinical research and eventual integration into EMRs as a result of these efforts.