To the Editor — Using atomic force microscopy (AFM), Cross et al.1 showed that cancer cells from patients are about 70% less stiff than normal cells. Seven different clinical samples, each consisting of eight normal and eight malignant cells, were measured and from the reported results, one can deduce that three or four force curves were recorded at the same position for each single cell. Cross et al. conclude that nanomechanical analysis may be a potentially useful technique for detecting cancers in the clinic.

The correlation between deformability of cells and diseases is well-known2 and has been evaluated in cell filtration3 and cell 'poking'4 experiments. Developments in AFM5 have allowed cancer detection at the single-cell level by mechanical means; coupled with routine morphological analysis of cells on histological specimens, large differences between cancerous and normal cells can be identified with more certainty. However, for cell stiffness to be used as an indicator of cancer the following considerations should be put forward.

Measurements on a significant number of cells are required for a statistically meaningful set of data. This is particularly important because only a single sample is typically drawn from patients, and as such the possibility of repeating the analysis and the probability of obtaining identical cancerous cells is low. Moreover, factors such as patient diet, medication and smoking habits can also affect the cell stiffness6. To determine the appropriate number of cells to measure, one needs to take into account the stiffness variations in a single cell and also in the population of cancerous cells in the patient sample. The way to do this is to consider cells as heterogeneous objects and to take multiple indentations at different locations on each cell and then calculate Young's modulus — the quantitative measure of the elastic properties of cells — for each cell separately.

Further to this, it would be necessary to study whether consecutive probing of the cell surface during the measurements can influence their elastic properties over time. The static measurements by AFM may not be as sensitive as interference methods7 in detecting rapid, dynamic changes within the cytoskeleton of the cell. However, if the cytoskeleton remodelling has a significant influence, it should propagate to the next probed location. Therefore, one way to check the influence of prolonged poking on the cell is to monitor the change in Young's modulus over time during the measurements.

Several sources of errors that could potentially limit the use of AFM for cancer detection must be pointed out. First, although the Hertz model8 is widely used to estimate Young's modulus with relatively good accuracy, the model assumes the living cell is an infinitely thick, purely elastic, isotropic and homogeneous material even though a cell is a highly heterogeneous material. This assumption makes it extremely difficult to obtain absolute values of Young's modulus. Moreover, the model approximates the real shape of the AFM tip by a cone or a paraboloid. One should remember that indentation depth is determined indirectly from AFM measurements, and therefore, small indentations are burdened with large errors arising from the uncertainty of the contact point, from the noise fluctuations of the baseline or from the change in the contact area. All these factors may influence the values of Young's modulus and given the heterogeneous characteristic of the cell, it is difficult to assume that taking a few measurements at one position (the centre of the cell) will satisfactorily reflect the mechanical properties of the entire cell.

In conclusion, the elasticity measurements of single cells should be carefully processed taking into account the limitations of AFM, and rigorous statistical tests should be used before this method can be reliably used for cancer diagnosis.