Tussen de 5 en 10% van de patiënten met een botmetastase krijgt uiteindelijk te maken met een dreigende of bestaande pathologische fractuur. De sterkte van pathologisch bot hangt af van zowel kenmerken van het bot als van de laesie (d.w.z. voor het bot: de grootte, de vorm en de driedimensionale variatie in dichtheid van trabeculair en corticaal bot, en bot-microarchitectuur, en voor de laesies: grootte, vorm, en aard). Natuurlijk is de belasting als externe factor ook van invloed voor het optreden van een fractuur. Progressieve chronische pijn (belastingafhankelijk, palpatoir, nachtelijk) is vaak het eerste symptoom van een pathologische afwijking in het bot, maar is in de praktijk maar van matig voorspellende waarde op een pathologische fractuur (Van der Linden, 2004; Nguyễn, 2021; Ardakani 2022).

Tijdige inschatting van het risico op fracturen is essentieel, aangezien dit een belangrijke overweging is voor de keuze van een lokale behandeling, zoals radiotherapie bij een laag risico en preventieve chirurgie bij een hoog risico. Daarnaast heeft het consequenties voor bijvoorbeeld adviezen rondom belasting, bijvoorbeeld door mobiliseren met krukken te adviseren. Er zijn verschillende scoresystemen in de literatuur beschreven die clinici helpen om een risico-inschatting te doen. Het scoringsysteem van Mirels (2003) bevat vier items (anatomische lokalisatie, mate van corticale aantasting, lytisch of blastisch, en mate van pijn) en wordt sinds de jaren negentig wereldwijd gebruikt. Op basis van later Nederlands onderzoek naar femora wordt inmiddels in Nederland de zogenaamde 3cm regel geadviseerd, die nauwkeuriger blijkt in het voorspellen van een fractuur op basis van een eenvoudig te verkrijgen röntgenfoto (Van der Linden 2004, Van der Wal 2020). Omdat Mirels een overschatting van het risico op daadwerkelijke fractuur geeft werd in de vorige richtlijn geadviseerd dit scoresysteem niet meer te gebruiken. Maar omdat in de praktijk het systeem nog veel gebruikt wordt beschrijven we de uitkomsten tov andere scoringssystemen in dit hoofdstuk 

Omdat zowel over- als onderbehandeling nog steeds voorkomt, wordt er op verschillende vlakken gewerkt aan verbetering van voorspellende modellen om de keuze voor een behandeling (conservatief, radiotherapie of operatief) te ondersteunen.

Description of studies

The quality of the literature review by Nguyễn (2021) was very low. The main methodological limitations are the absence of a systematic search strategy, no comprehensive literature search strategy, the absence of in- and exclusion criteria and no risk of bias assessments. Despite these limitations, Nguyễn (2021) pointed out that the prediction of long bone pathologic fracture remains a clinical challenge. In many cases, conservative treatment will suffice, although some patients with impending fractures are likely to benefit from preventive surgery. However, because of the variation between patients in disease characteristics such as life expectancy or response to irradiation or chemotherapy, it is difficult to establish a general yet accurate predictive tool for pathologic fracture risk assessment (Nguyễn, 2021).

Nguyễn (2021) summarized available instruments for the risk assessment of long bone metastasis fractures (see Table 1).

Table 1. Instruments of the risk assessment of long bone metastasis fractures (from: Nguyễn, 2021)

Results

The impact of using a prognostic instrument

One study assessed the impact of using a scoring system (CT-RA) on treatment planning (Nazarian, 2015) and one study assessed the impact on clinical decision making of using the BOS score (Eggermont, 2022).

Computed Tomography–Rigidity Analysis (CT-RA)

In the study by Nazarian (2015), 124 patients with 149 metastatic lesions were assigned to a treatment plan on the basis of the Mirels score. The Mirels criteria assigned 96 lesions (96/149; 64%) to the high-risk group (Mirels score > 9), whereas the physicians recommended surgery for 64 lesions (64/149; 43%).

Of the 85 patients (57%) that did not undergo prophylactic stabilization, 65 were followed over the following 4-month period. Seven new fractures, all at the lesion sites, were reported during follow-up in 7 different patients (11%). All 7 new fractures were correctly predicted to fracture using the CT-RA method (100% sensitivity). Of the 58 lesions that did not fracture, CT-RA predicted 52 not to fracture (90% specificity).

After obtaining the CT-RA results, the physicians changed the treatment for 36 patients (29%). Seven of those who did not undergo fixation ended up suffering fractures (19%). Although all seven of these fractures were correctly predicted by CT-RA, the physicians were influenced by subjective criteria, considering pain and the primary source of metastasis before opting for preventive stabilization. Only five of the seven lesions were predicted as being at high risk of fracture on the basis of the Mirels scores (Nguyễn; 2021).

The certainty of the evidence regarding the clinical impact of using CT-RA versus the Mirels score is very low. The certainty of the evidence started at “low” because of the pre-post study design and was downgraded to very low because of very serious study limitations and very serious imprecision (only one small study with insufficient power).

The 30mm (axial length)

The study of Van der Linden prospectively followed up patients taking part in a randomised trial of radiotherapy in order to predict pathologic fractures. In 102 patients with 110 femoral lesions, 14 fractures occurred during follow-up. The risk factors studied were increasing pain, the size of the lesion, radiographic appearance, localisation, transverse/axial/circumferential involvement of the cortex and a modified scoring system of Mirels. Axial cortical involvement >30 mm (p = 0.01) was most predictive for fracture. Their results indicate that most conventional risk factors overestimate the actual occurrence of pathological fractures of the femur. The assessment of axial cortical involvement provides a simple, objective tool using simple radiographs in order to decide which treatment is appropriate.

The BOne Strength (BOS) score

In the study by Eggermont (2022), 39 patients with 42 metastatic lesions were assigned to a treatment plan on the basis of the BOS score. BOS scores were requested by eighteen different physicians, of which ten requested more than one BOS score. The BOS score assigned 20 lesions (20/42; 48%) to the high-risk group, whereas the physicians estimated a high risk for 14 lesions (14/42; 33%).

For five femurs (5/42; 12%), both the initial estimation by the physician and the BOS score indicated a high fracture risk. These five patients were all scheduled for elective surgery, and in two patients, a pathological fracture occurred shortly prior to the scheduled surgery. In one case, the patient refused surgery but changed her mind after the BOS score indicated a high fracture risk.

For six femurs (6/42; 14%), both the initial estimation by the physician and the BOS score indicated a low fracture risk. One patient was treated with multiple fractions of radiotherapy because of pain complaints, four patients received a single dose, and one patient was not treated at all.

For 31 femurs (31/42); 74%), a discrepancy existed between the physician’s estimation and the BOS score. For 22 femurs (22/42; 52%), the initial fracture risk was assessed as low, but the BOS score indicated a high risk (n = 15) or moderate risk (n = 7). In 18 of these cases (18/22; 82%), the treatment plan was adapted based on the BOS score (1x elective surgery and 17x higher dose of radiotherapy. One patient who received multiple fractions of radiotherapy developed a fracture.

Regarding the impact of using the BOS score, physicians stated the BOS score had a decisive effect on the treatment decision. It was mentioned that treatment decisions were better substantiated. Another theme was “reassurance”. If the BOS score was in correspondence with the clinical fracture risk estimation, it gave both physicians (16 times) and patients (6 times) additional confirmation that the most appropriate treatment decision was made.

Regarding the theme “shared decision making”, it was mentioned the BOS score helped to open the conversation between the physician and patient.

The certainty of the evidence regarding the clinical impact of using the BOS score is very low. The certainty of the evidence started at “low” because of the pre-post study design and was downgraded to very low because of very serious study limitations and very serious imprecision (only one small study with insufficient power).

No studies were found that described and compared the impact of other scoring systems.

Comparative prognostic value

Nguyễn (2021) described three comparative studies: one prospective study comparing a cortical invasion of more than 30 mm in the axial plane (axial cortical involvement (ACI) > 30 mm), detected on a frontal and lateral X-ray or CT-scan, with the Mirels score, in 102 patients (Van der Linden 2003). The ACI > 30 mm was more specific of the risk of fracture than the Mirels score (Table 2). In a validation study on the axial cortical involvement using a new prospective set of 100 patients, all treated with radiotherapy for pain, again, the predictive value of the ACI as simple tool over Mirels was shown (van der Wal, 2020).

The second prospective study (Eggermont, 2020) compared the BOS Score with ACI in 45 patients. The CT-FEA model showed a better prognostic value with higher sensitivity and higher specificity.

Tatar (2014) proposed use of three-dimensional CT scans in order to obtain a more precise assessment of cortical involvement. In their retrospective multivariate analysis, based on 37 patients, a circumferential involvement ≥30% was the only predictive parameter for pathological fracture. Table 2) shows results from the prospective studies.

Table 2. Comparative predictive values obtained from prospective studies (from: Nguyễn, 2021).

A systematic review of the literature was performed to answer the following question:

What is the impact on clinical decision-making and patient outcomes of using a prognostic instrument for the identification of patients with bone metastases that are at risk for fractures?

Timing: at the moment of diagnosing osseal metastasis.

Setting: first, second or third line.

About prognostic research in general

There are currently few examples of the implementation of prognostic research resulting in improved patient care. The methodological quality of prognostic research is often low (Kent 2020).

For clinical use, externally validated prediction models should be translated into simple clinical prediction rules and help clinical decision making. Those rules guide the choice for treatment by providing information on the likely outcome of an individual (given different interventions), whereas prognostic rules inform the likely prognosis of an individual given just one treatment or care pathway. In a final stage, clinical prediction and decision rules, as well as single prognostic determinants can be tested in intervention studies (e.g. randomised clinical trials) to determine the impact of using the rule on patient outcomes and the cost-effectiveness of care or the effects of intervening on the prognostic determinant. Randomised and non-randomised impact studies can also play a role in describing the pragmatic ability of clinical rules to be adopted, change practice and improve outcomes.

This literature review focuses primarily on the clinical application of prognostic information and, secondary, confirmatory studies (prediction model external validation and investigation of causal relationships) and excludes exploratory studies (descriptive, association, and model development studies). The clinical application of prognostic information can include the development of clinical prediction or decision rules, and studies that seek to determine whether those rules do make a difference to outcomes when applied in treatment settings.

Relevant outcome measures

The guideline development group considered outcome measures reflecting the impact of using a prognostic tool on clinical decision making and patient outcomes as critical outcome measures for decision making; and outcome measures reflecting theoretical effects and the prognostic value of using a prognostic instrument, such as true positives, true negatives, false positives, and false negatives as important outcome measures for decision making.

The working group did not define the outcome measures listed above but used the definitions used in the studies.

Clinical and/or patient important differences

The working group agreed that sensitivity should outweigh specificity: overtreatment and to prevent a fracture (operating on a patient who would eventually not get a fracture) would be less serious than undertreatment (not operating on a patient and fracture occurs), while surgery also contributes to stability. This is also preferable from a patient perspective, patients are often afraid to move for fear of a fracture.

Search and select (Methods)

On 11 July 2022, the Embase and Ovid/Medline databases were searched with relevant search terms for systematic reviews on the prognosis of fractures in patients with bone metastases. The literature search yielded 670 unique hits. The detailed search strategy is depicted under the tab Methods. Studies were selected based on the following criteria:

• The patient population consists of patients with bone metastases other than vertebral metastases (for example long bones, such as the femur);
• The scope of the study is preferably the clinical application of prognostic information or external model validation;
• The design is a systematic literature review.

Seven studies were initially selected based on title and abstract screening. After reading the full text, five studies were excluded (see the table with reasons for exclusion under the tab Methods), and two studies were included (Benca 2016; Nguyễn, 2021). Although the review by Nguyễn (2021) was not a systematic literature review, it was decided to include this review for this guideline module (deviation from protocol). Despite its limitations, the guideline working group did not believe that conducting a new systematic literature review into this topic would be of substantial additional value that would outweigh the efforts and costs involved in a systematic literature review on this topic.

After the search date, in November 2022, an article was published about a Dutch prospective pilot study that evaluates the added clinical value of the Bone Strength Score (BOS) towards treatment-related decision making (Eggermont, 2022). Based on the relevance of this article for this guideline module, it was added to the body of evidence.

Results

Systematic reviews

Both Nguyễn (2021) and Benca (2016) were included in the analysis of the literature, but Nguyễn (2021) provided the most relevant information, included more recent literature (up to 2021) and included studies relevant for long bone metastases in general while Benca (2016) included only studies applicable for femoral metastases. Benca (2016) was therefore mainly used for cross-checking the information provided by Nguyễn (2021). Important study characteristics of Nguyễn (2021) are summarized in the evidence tables. The assessment of the risk of bias is summarized in the table of quality assessment.

Individual study not included in Nguyễn (2021)

One original study was included (Eggermont 2022). The original research on which this study was based (Eggermont 2020), was included in Nguyễn (2021).  Eggermont 2022 was included because of additional information on effect on decision making.