Alternative assays.

Delay between initiation and promotion on mouse skin was in 1949 reported by Berenblum and Shubik not to affect tumour yields, and this led to the important concept of the irreversibility of initiation and stimulated the development of multistage models. Subsequent reports have, however, suggested that delay does decrease tumour yields, and this is confirmed by the present study of 2200 mice initiated at 8, 48, or 68 weeks with 10, 30, 100, or 300 ,ug of DMBA and promoted by a standard dose of TPA for 15 weeks, after various delays. However, our data suggest that the decrease in tumour yields is chiefly or wholly due to a reduction, among ageing mice, of the ability to respond to promoters, and not to any substantial loss of initiated cells, for late initiation with immediate promotion also yielded a less rapid response than early initiation with immediate promotion. Interpretation of all such studies is complicated by the few weeks that the skin needs to repair ulceration and other damage induced by the higher doses of DMBA, for if promotion with TPA begins before such repair is complete the tumour yield may be misleadingly increased. Time between initiation and promotion.-When mouse skin is "initiated" with a single dose of DMBA (7,12-dimethyl benz(a)anthracene) and is then "pro-moted" with regular treatment for several weeks or months with croton oil or an active extract of it (Hecker, 1971) such as the phorbol ester TPA (12-0-tetra-decanoyl-phorbol-1 3-acetate), it is well known that "papillomas" will arise. However , there are conflicting claims (see below) in the literature as to whether the quantitative response to this combination of treatments is the same if mice initiated at age 6-10 weeks are kept after initiation for several months before promotion begins. Dose of initiator per unit area.-If the local concentration of initiator on mouse skin is large enough to cause ulcera-tion or skin erosion or inflammation, then this complicates the experimental outcome since (i) skin healing after either mild or severe ulceration has sufficient promoting action § to cause some papillomas to arise without any external promotion, and (ii) ulceration severe enough to remove the basal layer completely will remove some of the initiated cells that are to be studied, and it is unclear what proportion of initiated basal cells can be expected in the subsequent epithelial regrowth over a severely ulcerated area. The likelihood of ulceration depends, of course, …

standard dose of TPA for 15 weeks, after various delays. However, our data suggest that the decrease in tumour yields is chiefly or wholly due to a reduction, among ageing mice, of the ability to respond to promoters, and not to any substantial loss of initiated cells, for late initiation with immediate promotion also yielded a less rapid response than early initiation with immediate promotion. Interpretation of all such studies is complicated by the few weeks that the skin needs to repair ulceration and other damage induced by the higher doses of DMBA, for if promotion with TPA begins before such repair is complete the tumour yield may be misleadingly increased.
Time between initiation and promotion.-When mouse skin is "initiated" with a single dose of DMBA (7,12-dimethyl benz(a)anthracene) and is then "promoted" with regular treatment for several weeks or months with croton oil or an active extract of it (Hecker, 1971) such as the phorbol ester TPA (12-0-tetradecanoyl-phorbol-1 3-acetate), it is well known that "papillomas" will arise. However, there are conflicting claims (see below) in the literature as to whether the quantitative response to this combination of treatments is the same if mice initiated at age 6-10 weeks are kept after initiation for several months before promotion begins.
Dose of initiator per unit area.-If the local concentration of initiator on mouse skin is large enough to cause ulceration or skin erosion or inflammation, then this complicates the experimental outcome since (i) skin healing after either mild or severe ulceration has sufficient promoting action § to cause some papillomas to arise without any external promotion, and (ii) ulceration severe enough to remove the basal layer completely will remove some of the initiated cells that are to be studied, and it is unclear what proportion of initiated basal cells can be expected in the subsequent epithelial regrowth over a severely ulcerated area.
The likelihood of ulceration depends, of course, not only on the strain of mouse and the dose of initiator, but also on the volume and physical nature of the vehicle in which the initiator is applied. Some experiments have used 0-2 ml of acetone, which spreads the initiator over the whole shaved back of the mouse and reduces the likelihood of ulceration, while others have used only about 0 02 ml of liquid paraffin, which confines the initiator to an area less than 0-5 cm2 between the shoulder blades of the mouse. This has the advantage of preventing the animal from scratching the treated area, but at the expense of increasing the likelihood of ulceration.
Previous experiments. -Berenblum & Shubik (1949) studied female Swiss mice, the shaved backs of which were initiated with "a drop on the end of a fine glass rod" of a 1 5 % solution of DMBA in liquid paraffin. If the volume deposited was about 0 02 ml, each mouse must have received about 300 [kg of DMBA in a small region of the back. These mice were later promoted for 17 weeks with regular croton oil. In one experiment, 100 mice underwent promotion "early" (Weeks 3-20 after initiation), while in another experiment 25 mice underwent promotion "late" (Weeks 43-60 after initiation). In both the "early" and the "late" experimenits, there was almost no intercurrent mortality, and about half the mice developed at least one papilloma, while half developed no papillomas. This apparent similarity of response of the animals promoted during Weeks 3-20 and 43-60 after initiation has generally been taken as evidence that initiation is essentially irreversible, a conclusion which has acted as an important stimulus to the fruitful development of ''multistage models" for cancer induction (e.g. Armitage & Doll, 1961; for review, see Peto, 1977, or XVhittemore & Keller, 1978. Unfortunately, subsequent work has not confirmed the evidence* on which this conclusion rests, and delay between initiation and promotion does appear to decrease the response to the promoter (Roe & Salaman, 1954;Roe et al., 1972;van Duuren et al., 1963van Duuren et al., , 1967van Duuren et al., , 1975van Duuren et al., , 1978Hieger, 1965). However, this does not necessarily imply any progressive loss of initiated cells, as (a) the biological potency of promoters in ageing mouse skin may be less than in the skin of young adult mice, and (b) promotion during the few weeks while the skin is recovering from the ulcers, erosions and other less obvious short-term effects of initiation may cause more papillomas to become visible than would promotion after the skin had repaired itself. Explanations (a) and (b) could account for the observation by van Duuren et al. (1978) of (a) earlier response to early initiation with immediate promotion than to late initiation with immediate promotion, and (b) nmore papillomas with late initiation and late (i.e. immediate) promotion than with early initiation and late promotion.
The present study extends the results of van Dtiuren et al. (1978) in various ways.
To make our results more manageable, wN-e have presented and discussed the parts of our data which bear on these 3 distinct questions in 3 separate, though overlapping, papers. Moreover, for simplicity of presentation we have summarized our principal results graphically, relegating the detail of the data from which these graphs, derive to reduced-size tabular appendices.* After the description of our experimental and statistical methods, this first paper will address itself to the question of the persistence of initiation.
Test animals. Female Swviss mice, randolm bred at the Eppley Institute, were used, and were 8 weeks old at the start of the experiment. All animals w,ere randomized before treatment, and were kept unvaccinated in groups of 10 in plastic cages ith a wrire mesh top in Sanicel bedding, and given WVaynie pelleted diet (Wayne Corp., South Bend, Ind.) and wrater ad libitum,. The animals wiere allowed to die spontaneously or were killed when moribund, and we then attempted to determine wNhether skin tumours had, directly or indirectly, caused that death; in almost all cases, an unequivocal answer was obvious.
Chemicals. DMBA w,as ol)tained from Aldrich Corp., Milwaukee. Wis., and purified and analysed by paper chromatography for purity (which wNas alw,ays more than 990o).
DMBA solutions in acetone were always prepared immediately before use and were checked quantitatively before use; no discrepancies were found. The tumour promoter 12-0-tetradecanoyl-phorbol-13-acetate (TPA) was obtained from Midland Corp., New York, N.Y., and checked for purity before use, as well as regularly during treatment, by the mnethod of Roe et al. (1972). Acetone solutions were prepared monthly and all batches were checked quantitatively both before and after use; again, no discrepancies were found. Fluorescence-free acetone solvent (Fisher Co., Chicago, Ill.) was used.
Treatment.-The chemicals were applied to the small interscapular region of the back (an area which cannot be scratched by the animals) with a precision pipette (Hamilton Ind., London, Ont.) 0-017 ml at a time. An area of about 1 cm2 of the interscapular dorsal skin was regularly shaved with an electrical clipper, care being taken not to erode the skin. The part of this area which wAas actually treated did not have a very sharp boundary, being defined by the spread of the 0-017 ml of acetone solution which was applied, but it was roughly circular with a diameter of about 0 5 cm, and hence with an area certainly less than 0-5 cm2. In such a small area, there is not room for many tumours to develop. All animals had to be in the resting stage of the hair cycle at the time of initiation, since it is well known that the hair cycle affects the * In a stu(ly with as maniy groups as thlis, some lhunidre(ds of (different pairwise comparisons between groups are possible, ancd so all sorts of anomalies cani be expectedi to arise by clhanice. To get reliable answers to our main questions, we slall tlherefore lhave to average some of ouir groups together in ordler to reduce purely random erros. Tabulation of the analys;es leading to sucli combinations of groups resuilts in rather a mass of numbers. efficacy of the initiator (Berenblum et al., 1958; see also Manil et al., 1981). To determine wArhether animals Aere in resting stage, their entire backs were shaved, and 1 week later thev were examined for signs of regrowth. At Week 8, animals then in the resting stage were randomized 5 ways, between delayed initiation and immediate initiation wNith a particular one of the 4 DMBA doses, while all animals not in the resting stage were put into the "delayed initiation" group. (Because of the presumed uniformity of the hair cycle in all animals, it was not felt that the "delayed" group wA-ould thereby be biased, but w-e have no direct check on this supposition.) At Week 48 the survivors of the "delayed" group which wN-ere then in the resting stage Awere likewise randomized 5 wAays, one group plus those not in the resting stage at Week 48 having their DMBA initiation further delayed.
At Week 68 all animals were randomized irrespective of hair cycle, but the treatment of those undergoing regrowth was delayed for a week to allow reversion to the resting stage.
Tumours and other skin lesions-.Animals were checked weekly and gross appearance, behaviour, nutritional status, etc., were recorded. All skin lesions were measured and charted wNeekly, size and gross appearance (apparent nature, shape, colour, necrosis, etc.) being noted. For each tumour we recorded, where relevant, any preceding lesions (e.g. ulceration or hyperplasia), week of appearance, wNeek of apparent progression from one type of tumour to another, and wieek of regression. Particular care was taken to record accurately the week w,hen a tumour first attained a maximal diameter of at least 10 mm.
Histology.-A complete necropsy was performed and all grossly visible lesions, diseases, tumours, etc. were recorded. All skin lesions and tumours, plus selected specimens from other organs, were studied histologically. Formalin-fixed, paraffin-embedded specimens were sectioned, stained with haematoxylin and eosin and other stains as needed, and examined histologically by F. Stenback, using standard criteria for morphology (Stenback, 1969). Most skin tumours that did not regress spontaneously wsrere available for histological confirmation, since fewr mice were lost due to decomposition or cannibalism. Information for lost tumours, for regressing tumours, and for tumours ultimately con-fluent with other surrounding lesions is based upon gross observation, but there wvas never a case where the fact of malignancy or otherwise was uncertain.
Survival.-In general, the groups given 10 or 30 [kg DMBA had, at least until w-ell after the 15-week course of promotion was finished, death rates very similar to the controls which received no DMBA (about 75O% remaining alive at 1 year), but the groups receiving 100 or 300 [g of DMBA suffered much more rapid mortality, so much so that there wvere too few of these animals left for proper study of the effects of a 43or 63-week delay between initiation and promotion. Among the mice receiving 100 or 300 itg of DMBA, inflammation, reddening and superficial erosion of the whole of the small treated area were usually apparent within a few days, and sometimes the whole epithelium, including the basal layer, was lost. These early changes gradually disappeared, and the animals then remained healthy for a month or two, after which, even without promotion with TPA, some tumours began to appear. Ultimately, even in the absence of promotion, many animals initiated wvith 100 or 300 [kg of DMBA developed severe ulcers with raised borders as well as various tumours, and several died of these causes.
Indices of response.-Four separate indices of neoplastic response have been analysed for each comparison, usually with concordant results. (The need for more than one analysis arises because the difference between an irreversible, invasive carcinoma and the sort of transient benign papilloma typically produced by TPA is so extreme that it could not necessarily be assumed that their aetiologies and relationships to treatment would be similar.)

(i) Number of tumours
We may count the total numbers of tumours, irrespective of size or type, arising during the 20 weeks from the beginning of promotion, on mice alive at the end of that 20-week period. (A 20-week period is long enough to include most of the papillomas arising after initiation + 15 weeks' promotion, but is not long enough for significant numbers of deaths from tumours.) (ii-iv) Tumour-bearing animals (ii) Among animals which wN-ere tumourless at the start of a particular period (e.g. at the start of promnotion) wve many count the nutnbers of animals which ultimnately develop one or more tumour s (i.e. tumours of any size or type).
(iii) Among animals which were without a 10mm tumour at the start of a particular period wre may count the number of animals which eventually developed a 10mm tumour.
(iv) Among animals which were without an apparently malignant tumour at the start of a particular period we may count the number of animnals which eventually developed a malignant tumour. (Every week it was noted whether each tumour seemed malignant, so for histologically malignant tumours we can use the week of first apparent malignancy. Statistical nmethods.-For each of the above 4 indices of response, the fundamental concept underlying the statistical comparison of tumour yields in a fewA, particular treatment groups is to contrast the observed numbers of tumours, 0, in each of these groups with the "'expected" numbers, E, and to calculate their ratios, O/E. "Expected" nurnber has the usual meaning in statistics. That is, if the tumours in the few particular groups being statistically compared were shared out among those groups in proportion to the numbers at i,sk in them, how many would be expected in each group: For exarnple, when comparing the effects of 300, 100, 30 and 10 ,ug of DMBA followed by immediate promotion on the total numbers of tumours on survivors at 20 weeks, wve observed respectively 146/73. For the analyses (ii), (iii), and (iv) of titne to the first ocurrence of some specified type of tumour, differences in longevity must be allowed for -when calculating expected numbers (IARC, 1980). Advantages of using Os and Es include the facts that: (a) Different comparisons can be pooled simply by adding up the corresponding Os and Es: many examples will be found in the appendices to these 3 papers.
(b) P values for the differences between groups in respect of numbers of tumourbearing animals (although not for the total numbers of tumours) can be derived from the differences between Os and Es (as in Peto & Pike, 1973;for discussion, see IARC, 1980).
(c) The ratios O/E provide a useful description of the relative impact of the tumour type of interest on the particular groups being compared and, when the O/E values for the 4 different indices of response (i-iv) all give a similar impression, a satisfactory characterization of the differences between those groups has been achieved.
A fuller discussion of the statistical methods (ii), (iii), and (iv) that we have used for our analyses of time-to-tumour may be found in the annex on statistical methodology for animal experiments to IARC (1980), the methods used being those described for tumours that are observed in a "mortalityindependent" context.
Appendix Table (a) summarizes certain indices of response to promotion in the 8 groups -which wN,ere scheduled to be promoted. (Because promotion lasted for 15 weeks only, the papillomatous response to promotion is adequately characterized by the tumour yield within 20 weeks of starting promotion.)

RESULTS
This first paper now deals only with the persistence of initiation, which was the main question we wished to study. Our second paper (Stenback et al., 1981a) deals with systematic differences between the effects of giving the same initiating dose at different ages, followed by the same interval before promotion. Our third paper (Stenback et al., 1981b) deals with the shape of the dose-response curve. This subdivision helps to make the data manageable, because most of the analysis and discussion in Papers II and III is not relevant to the present question of the persistence of initiation. However, 2 conclusions from those analyses are needed here, namely: (1) Among the groups initiated but not promoted for some time, healing of the ulcerations and erosions caused by DMBA (2) Either initiation, or promotion, or both, are less effective at age 18 months than earlier, for initiation plus immediate promotion yielded only half as many tumours at 68-86 weeks as it had done at 8-26 weeks or 48-66 weeks, when the yields were about equal ( Fig. 1 in Stenback et al., 1981a).
Bearing these conclusions in mind, let us first examine the observations which reproduce and extend the original experiment of Berenblum & Shubik (1949). Fig.  2 summarizes the statistical comparison of Schedules a, b, c, d, and e, in which the mice were all initiated at age 8 weeks and then promoted for 15 weeks from ages 11, 18, 31, 51, and 71 weeks, respectively. (For numerical details, see Appendix Tables b and c). In Fig. 2 it is apparent that the tumour yields are not constant (P=0 0004 for trend), as was also found by Roe et al. (1972)   and insufficient to cause any erosion or minor skin changes, there might be no special synergy with early TPA (unless the proliferative or other effects of TPA 3 weeks after initiation interfere with some hypothetical form of slow repair of the DMBA-induced damage to the DNA in the stem cells).
What is more interesting, however, is the apparenit further reduction in response among animals promoted at 71-86 weeks of age, as compared with animals treated at 51-66 weeks or earlier. The magnitude and timing of this decrease in response at 71-86 weeks is similar to the 50%0 reduction in response reported in our accompanying paper ( Fig. 1 in Stenback et al., 1 981a) when initiation (with promotion starting 3 weeks later) takes place in animals aged 68 weeks rather than at 8 or 48 weeks of age. The grouips of animals concerned in these 2 comparisons ( Fig.  2 and ibid., Fig. 1) are with one exception (Group a) completely different. Thus the observed decreases in response to promotion at 71-86 weeks in both of these age relationships are statistically fairly independent of each other. However, the fact that the 95% confidence intervals are in each case rather wide for the groups promoted at 71-86 weeks means that chance factors may have made the 2 relationships appear more similar than thev wotuld have in an even larger study.* Despite this reservation, the most economical explanation for the observed similarity between these 2 patterns is that the one factor which is common to Fig. 2 and ibid., Fig. 1 i.e. promotion at 71-86 weeks rather than earlier, explains the decrease in effect in both graphs. If it is provisionally accepted that promotion at about 18 months of age is about 50%0 less potent than pronmotion during the first year or so of adult life, our data offer no evidence whatever for any substantial loss of initiated cells as initiated mice grow older. The foregoing conclusions are consistent with our findings (Fig. 3; for numerical details, see Appendix Tables b and c that, among animals initiated at 48 weeks, promotion at 51-66 weeks produces twice the tumour inicidence of promotion at 71-86 weeks. The postulated 50%0 decrease in the efficacy of TPA promotion at 71-86 weeks wouild contribute to the difference seen in Fig. 3. Another contribution to it would be provided by our previous postulate that, if TPA is given promptly (3 weeks) after DMBA, the net promotional effect is greater (owing, perhaps, to aggravation of the promotional effects of wound healing, or perhaps to interference with slow DNA repair). Between them, the hypotheses invoked previously are supported by, and account naturally for, the statistically significant difference seen in Fig. 3.
Thus far we have supposed that the numbers of initiated cells produced by a given dose of DMBA does not vary much with age, and that there is no material loss of initiated cells as animals age. We have then inferred that the promotional * Tlbe approximate stu(ly sizes to date in wlmieb the effects of (lelaying promotbon liav-e beeni studied are 120 (1949), 50 (1954), 30 (1963), 60() (1965), 80 (1967), 310 (1972), 560 (1975/8) andl now 2200 mice (1980). All tie larger stuclies liave shown a (decrease in response if promotioni is delayed, buit only thie first study, wbrere no (lecrease was e-idlent, is widely kniown. effect of TPA does decrease somewhat as the mice get past 1 year of age and on to the age of 18 months or so. Finally, we have inferred that if IDMBA is given only 3 weeks before TPA, we get a greater tumour yield than if it had been much earlier. If these hypotheses are indeed correct, we should find no material dependence of tumour rates on time since initiation, except for an excess rate among those initiated just previously. Figs 4 and 5 summarize the relevant comparisons between animals promoted at a fixed age (which eliminates all artefacts due to the age-dependence of promotion). Fig. 5 does show the expected increase in effect when initiation is just before promotion, buit Fig. 4 does not.* Among our many comparisons, this is the only result which is not concordant with our hypotheses. Moreover, since the predicted increase in tumour yield when initiation jtust precedes promotion is clearly seen in Fig. 5, the lack of it in Fig. 4 is not strong evidence against our general concluision that promotion rather than initiation varies with age, especially since the direction of the not indicate any loss of with time. These findings may be explained in 2 main ways: (a) "Single" hypothesis: promotioin late in life is less effective than promotion early in life, ancd (b) "Double" hypothesis: initiation late in life is less effective than initiation early in life, and among animals initiated early in life there is a substantial loss of initiated cells between ages 12 and 18 months. Logically, it is clear that there can be no wavy, from experiments such as oturs, of distinguishing between the "single" and the "double" hypotheses above. However, the "double" hypothesis requires the existence of 2 separate mechanisms which by chance yield similar quantitative * This may merely reflee t te fact tbat there were almost nio high-(lose survivors left at 80 vWeeks in Grouip e for comparison with tile high-dose mice in Group hi, wliich meanis that the few survivors in Group e are rather select and that the O/E ratio in tlihe larger grouip is biasedt tows aids unit.v. INITIATION AND PROMIOTION, I effects. We therefore consider the "single" hypothesis, that there is a decrease in the efficacy of promotion as animals age, to be a more economical, and hence more plausible, explanation for our findings. Supporting evidence for it could perhaps be obtained by studying the biochemical or histological response of young adult and of very old adult mouse skin to TPA (for details and references, see Stenback et al., 1981a). TPA is known to induce enormous increases in certain enzymes such as ornithine decarboxylase and plasminogen activator, as well as in "primitive stem cells" (or "dark" cells). If the inducibility of cells from old skin was found to be less, this would strengthen our hypothesis. Conversely, it is now possible to measure the numbers of polyaromatic hydrocarbon molecules bound per 106 DNA bases in the epithelial basal layer after initiation, and it might be useful to study the agedependence of this binding. Doses lower than those we have studied should probably be studied, as the flat dose-response relationship we have observed (Stenback et al., 1981ib) suggests that we are saturating the target cells. Our data therefore provide no direct evidence of the reversibility of initiation to be expected if non-saturating doses (perhaps of -0-jug DMBA in 1/60 ml acetone, or 1 ,ug DMBA in 0-2 ml acetone) were used for initiation.
In 1949, Berenblum & Shubik concluded that initiation was irreversible and that initiated cells persisted in the skin. Their evidence for this was comparison of 2 experiments in which there was the same apparent effect of promotion at 3 as at 43 weeks after initiation. The findings of Roe et al. (1972), van Duuren et al. (1978) and the present study show conclusively that Berenblum & Shubik's quantitative findings were not reproducible, but the present study suggests that their conclusion was substantially correct, at least to within a factor of 2.
Initiation is probably more or less irreversible, and the effects of initiation with subtumorigenic doses of DMBA are probably more or less independent of age. However, (i) the efficacy of promoters appears to decrease in ageing mice, (ii) if DMBA doses large enough to kill skin cells are given the promotional effects of the resultant healing processes are greater in older than in younger mice, and (iii) if promotion is administered before the skin has been given a chance to settle down again after initiation, the tumour yield will be greater.
These conclusions explain naturally why it was that, when Peto et al. (1975) tested on mouse skin the carcinogenicity of repeated subtoxic doses of benzo(a) pyrene (a chemical which resembles DMBA much more than TPA in its likely mode of action) they found, after correcting by life-table methods for intercurrent mortality, that the neoplastic response was independent of the age (10, 25, 40, or 55 weeks) at which treatment began. It appears to be the promotional, rather than the initiating, processes which vary with age, and which account for most or all of the anomalies which we and others have observed.
Whether analogous promnotional processes govern human tumour induction is not known. It should be emphasized that although the little papillomas that can be elicited rapidly by TPA have been referred to as "tumours", both by us and by many others who have studied initiation and promotion on mouse skin, their biological nature is very different from that of an invasive carcinoma. In particular, many of them are not autonomous and may either shrink or disappear completely when regular treatment with TPA ceases (or even, in some instances, while it continues). There is obviously no necessary analogy between the aetiologies of such lesions and of invasive carcinomas. But we note that, in our analyses where there were a sufficient number of carcinomas for statistical stability, the observed effects of the various delays of initiation and promotion on invasive carcinomas have generally been rather similar to the effects on papillomas.  2. The entry "T/M = x + y in a/b/c" indicates that c animals underwent initiation, b survived to the start of promotion, and a of these survived a further 20 weeks thereafter, and that among these 20-week survivors an average of x new tumours per mouse arose during the 20 weeks after the start of promotion. y is an estimate of the standard error of x. Overall, 2200 mice were initiated, 1701 started promotion and 1082 were alive 20 weeks thereafter.
3. Finally, the crop of new tumours elicited within 20 weeks of the start of promotion is further characterized by listing in brackets, for eaclh new tumour-bearing survivor 20 weeks after starting promotion, the number of new tumours that arose since promotion began. ("3 x 1, 2" means 3 animals had one each and one animal had 2, etc.) 300 ,ug initiation 69% T/M = 2-00 + 0-27 in 73/80/80 (19 x 1, 9 x 2, 9 x 3, 6 x 4, 5 x 5, 6, 6, 7, 14)  Total numbers of tumours (including second and subsequent tumours) during the 20 weeks from the start of promotion, among animals promoted at a fixed age (51-66 or 71-86 weeks, as indicated) following initiation at various previous ages, excluding animals which died less than 20 weeks after starting promotion.
Abbreviations andl footnotet as in Appendix