Research on the predicting power of the revised Tokuhashi system: how much time can surgery give to patients with short life expectancy?

Metastatic epidural spinal bone tumors are major health problems nowadays, besides being an economic burden [1]. The ever-expanding oncological treatment opportunities extend the lifetime of the patients, and, consequently they affect the incidence of secondary lesions [2].

Since the publication of Patchell et al. [3], it has been accepted in literature and in practice that the primary treatment option for metastatic spine tumor is surgery (supplemented by radiotherapy in accordance with the radiosensitivity of the primary alteration) [4]. The exact surgical methodology (posterior stabilization with or without decompression, debulking, partial or en bloc spondylectomy etc.) depends on the general conditions of the individual [5, 6]. There are patients who could be treated only with non-surgical, palliative or supportive methods (oedema control, prevention of further pathological fractures, etc.) [7].

However, therapeutic decision-making is highly complicated (setting up the indication of the surgery and choosing the exact degree of invasiveness being the most difficult tasks), we have to consider many aspects [8]. Modern medicine requires the implementation of personalized solutions, in which the prognostic predicting scoring systems could help us. The “revised Tokuhashi system” (rTS) [9] perhaps is the most well-known system worldwide, however, the publications discussing its usability are also highly controversial, reporting a total predictive ability about 60% [10‐12]. A highly beneficial feature of the system that besides predicting their life expectancy, it also offers therapeutic options for the patients.

Our aim was to propose new scoring thresholds in the widely used rTS, as new surgical techniques could be used to successfully treat previously inoperable patients.

Medical database at the Department of Neurosurgery, Semmelweis University, Budapest was collected to identify all patients who underwent surgical intervention because of vertebral metastases between December 2007 and September 2015. It contains data on 382 operations performed on 337 patients, extended with the survival data of 329 patients. Only patients over the age of 18 with operated metastasis were included. Excisional surgical methods mean that the whole tumor of the vertebral body was resected (en bloc spondylectomy), and posterior stabilization was made; palliative techniques include posterior stabilization with or without posterior decompression (laminectomy) and partial tumor removal.

We have collected information about each patient, including demographic data, detailed data on the disease and the results of the surgery (see details in Table 1).

Most of these factors are part of the prognosis predictions and the therapeutic option recommendation system discussed in this study. Tokuhashi et al. published their scoring system in 1989 [13] and, after several retro- and prospective analyses, they published the revised version in 2005 (Table 2) [9]. Other systems have been proposed as well, however, the above version has had the greatest effect on the neurosurgeon society. We scored our patients according to the version published in 2005, and our use of the term ‘rTS score’ is grounded on the 2005 version.

The population under study has the following general properties: the sex ratio is balanced, with 199 (59.1%) males and 138 (40.9%) females in it. The typical member is from the elderly generation, as the mean age is 63, but the full age range varies between the minimum of 18 years and the maximum of 88 years. The median OS, that is, the amount of time when 50% of the patients have died is 222 days (CI 95% 175–274). OS was calculated using the Kaplan–Meier (KM) formula. Here we review the general properties of the population from the original study of the rTS [9], in which there were 154 male patients and 92 female patients. Their mean age was 56.5 years (range 15–85 years). The site of lesion is as follows: 55 cervical patients, 142 thoracic patients, and 99 lumbosacral patients. The primary site of the cancer was as follows: lung 48, breast 26, kidney 24, liver 15, prostate 15, rectum 10, stomach 10, thyroid 7, uterus 6, colon 5, stomach 5, osteosarcoma 4, unidentified 34, others 37.

In our first report, we tested the factors that have the most significant impact on survival [14]. Later, we published about the efficiency and prediction accuracy of the four best-known scoring systems used with metastatic spinal tumor patients. The systems are able to separate the patients according to their overall survival periods. However, the rTS showed 60.5%, Tomita 28.8%, modified Bauer 29.5% and the van der Linden 48.6% precision about the prediction of the real survival periods [15].

In this article, we present a plausible explanation of why the rTS has low precision despite of its statistically validated separation feasibility.

As an oncological disease progresses, the chance of developing metastatic bone diseases (including metastatic spine tumors) increases, which means poor prognosis for the patients [2, 16].

The highly malignant glioblastoma multiforme, which is still an unresolved oncological problem, has a median survival time of 14.6 months [17]. However, our metastatic spinal tumor population has a lower value, with a median OS time of 7.5 months. The above data are identical to what has been published in the literature. Rades et al. [18] found a 2-, 4- and 11-months median OS time for their three subgroups of patients who underwent radiotherapy without surgery, da Silva et al. [19] examined a population with secondary lesions caused by lung carcinoma, and described a median OS value between 8 and 12 months.

Prognosis-predicting scoring systems establish the categorization of patients based on their survival time, which can help doctors choose the optimal type of treatment [20, 21]. This separation ability is also validated by the literature [22], nonetheless, more and more articles are being produced about the feasibility of the systems. Our previous study showed an average predictive capability of 60% for the rTS, which is also consistent with literature data. Zoccali et al. [23] reported an average 63% predictive value for rTS. In their results, it can be seen that the prediction ability of the system is reduced at the patients with a life expectancy shorter than 1 year (more than 12 months: 77.21%, 6–12 months: 55.32% and less than 6 months: 64.10%).

The main question of our study is what the cause of the inaccurate prediction of the system could be. In our conservative category population, a significant increase in survival time was observed, the OS value decreased to 0.15 on day 475, contrary to the 180th day in the literature, which means 295 days of lifetime elongation. The reason for the increase in survival time was found in differences of the treatment options, namely, while all of our patients underwent surgery, literature suggests only supportive (or non-surgical palliative-) therapy for this category. Pelegrini et al. [24] also criticized the recommendation of treatment methods of the rTS, the advantages of surgical treatment were emphasized against conservative options. Oliveira et al. [25] also reported that the therapy cannot be based on the recommendation of the rTS.

Based on our results, we tried to change the conservative category point limits where there was a significant difference in therapy. Instead of the 0–8 points found in the literature, we would change the conservative category points to 0–6.

It could be seen again; the primary tumor type is one of the most important prognostic factors influencing the outcome of the disease. The subgroup of the conservative category examined in detail, which contained patients who did not survive the predicted 6-month period, contained significantly more lung cancer patients. The study prepared by Bollen et. al [26] also highlights the above statement (the primary tumor and the performance status are the most important factors affecting survival). Even more prognosis-predicting scoring systems appear in the literature which focus on metastases caused by one type of primary tumor (so we can get more precise prediction based on more specific information, e.g., histological categorization, molecular subtypes, specific targeted treatment options, etc.). Uei et al. [27] designed a new system for estimating the outcome of spinal metastasis in patients with primary lung cancer, which also takes into account the possibility of tumor antigens and molecular target therapy.

The spread of minimally invasive spine surgery techniques also allows for surgical intervention for patients with poor general condition due to lower surgical load. For example, spinal stabilization procedures could be performed with less blood loss and surgical strain [28, 29], or a metastatic epidural tissue mass causing metastatic epidural spinal cord compression can be removed with minimal access hemi-semi laminectomy [30].

By conserving neurological functions, we can achieve significant increases in the patients’ quality of life and survival time. Several studies have shown that surgery has a good effect on the quality of life of patients [31, 32], de Ruiter et al. [33] reported that invasive open surgery techniques do not negatively affect patient outcomes. By preventing neurological functions, side effects (such as thrombosis, decubitus, infections, etc.) can be avoided. The extension of the survival time of our own population is also attributed to the absence of the above ‘side effects’. To sum up, in determining the therapeutic options for the subcategories of the prognosis-predicting scoring systems, we must take into account the open-access opportunities offered by minimal access surgery techniques, as described by Rao et al. [34].

The treatment of epidural spinal metastases is a major challenge from both oncological and surgical aspects. In doubtful cases, prognosis-predicting systems could help, however, but we would like to emphasize that they only serve as guidance and are of secondary importance compared to a doctor’s years or decades of medical experience.

According to our results, rTS draws a too strict boundary between surgical and other, non-surgical palliative and/or supportive therapeutic options, as it may be worthwhile to also provide surgery for patients who are in poor condition based on the scoring systems. Surgical treatment is unquestionably important for the treatment of pain and may also lead to further improvement in the quality of life of a patient through the improvement/preservation of neurological functions, which may make patients eligible for further oncological treatments. Based on the above, we recommend the revision of the therapeutic recommendation section of the rTS system, also considering the modern spinal surgery techniques.